WO2022012257A1 - 通信的方法及通信装置 - Google Patents

通信的方法及通信装置 Download PDF

Info

Publication number
WO2022012257A1
WO2022012257A1 PCT/CN2021/100637 CN2021100637W WO2022012257A1 WO 2022012257 A1 WO2022012257 A1 WO 2022012257A1 CN 2021100637 W CN2021100637 W CN 2021100637W WO 2022012257 A1 WO2022012257 A1 WO 2022012257A1
Authority
WO
WIPO (PCT)
Prior art keywords
learning model
channel
communication device
applicable
channel learning
Prior art date
Application number
PCT/CN2021/100637
Other languages
English (en)
French (fr)
Inventor
王婷
何高宁
冯奇
卢建民
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2022012257A1 publication Critical patent/WO2022012257A1/zh
Priority to US18/153,876 priority Critical patent/US20230155702A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/391—Modelling the propagation channel
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/382—Monitoring; Testing of propagation channels for resource allocation, admission control or handover
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413—MIMO systems
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413—MIMO systems
    • H04B7/0417—Feedback systems
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413—MIMO systems
    • H04B7/0456—Selection of precoding matrices or codebooks, e.g. using matrices antenna weighting
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H04L25/024—Channel estimation channel estimation algorithms
    • H04L25/0254—Channel estimation channel estimation algorithms using neural network algorithms
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W40/00—Communication routing or communication path finding
    • H04W40/02—Communication route or path selection, e.g. power-based or shortest path routing
    • H04W40/18—Communication route or path selection, e.g. power-based or shortest path routing based on predicted events
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Definitions

  • the present application relates to the field of communication, and more particularly, to a method of communication and a communication device.
  • Massive Multiple-Input Multiple-Output technology
  • network devices can allocate limited power to data streams that can be efficiently transmitted through precoding, while reducing the amount of communication between multiple terminal devices. It is beneficial to improve signal quality, realize space division multiplexing, and improve spectrum utilization.
  • the terminal device can determine the precoding matrix based on the downlink channel measurement, and through feedback, the network device can determine the precoding for data transmission based on the precoding matrix fed back by the terminal device, thereby improving signal transmission performance.
  • a method for information transmission based on artificial intelligence is known.
  • the terminal device and the network device are jointly trained to obtain a first channel learning model and a second channel learning model, wherein the first channel learning model
  • the communication system including the first channel learning model and the second channel learning model is called an AI communication system.
  • the AI-based information transmission method includes: the terminal device obtains the information to be fed back, the terminal device processes the information to be fed back through at least a first channel learning model to obtain the information that needs to be fed back through the air interface, and the terminal device then passes The feedback link feeds back the information that needs to be fed back through the air interface to the network device, the network device receives the information fed back by the terminal device, and the network device processes the feedback information at least through the second channel learning model, so as to obtain the terminal device side to be fed back Information.
  • the first channel learning model and the second channel learning model obtained by offline training are directly applied in the process of online information transmission. Changes in the communication environment may cause the first channel learning model and the second channel learning model to be inapplicable, thereby affecting the performance of information transmission based on the information transmission method of the AI based on the first channel learning model and the second channel learning model.
  • the present application provides a communication method, which can judge the applicability of a channel learning model being used.
  • a communication method may include: a first communication device determining whether a first channel learning model is applicable, where the first channel learning model is used to determine first channel information based on target channel information, the The data volume of a channel information is less than the data volume of the target channel information; in the case that it is determined that the first channel learning model is not applicable, the first communication device sends a first message, and the first message is used to indicate the first channel Learning models do not apply.
  • the first channel information is used to determine the second channel information through the second channel learning model, and the second channel information is the same as or similar to the target channel information.
  • the first communication device can judge the applicability of the first channel learning model without the assistance of the second communication device, so the interaction of signaling can be reduced and the applicability of the channel learning model can be judged complexity.
  • the first communication device determines whether the first channel learning model is applicable, including: the variation of the long-term statistical characteristics of the target channel of the first communication device is greater than or equal to the first In the case of a preset threshold value, it is determined that the first channel learning model is not applicable; or the first communication device determines that the first channel learning model is not applicable when the variation of the long-term statistical characteristics of the target channel is less than the first preset threshold value.
  • the channel learning model applies.
  • the first communication device determines the applicability of the first channel learning model according to the long-term statistical characteristics of the target channel, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the first channel can be reduced.
  • the processing burden of the communication device is not limited to the above technical solution.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device determining the first channel learning model according to the received first scheduling information Whether the model is applicable.
  • the first scheduling information is sent by the second communication apparatus according to the second channel information, and the second channel information is determined according to the first channel information and the second channel learning model.
  • the first communication device determines the applicability of the first channel learning model according to the first scheduling information, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the first communication device can be reduced in size processing burden.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device determining whether the first channel learning model is applicable according to data transmission performance .
  • the transmission performance of the data may include the transmission performance of the first data and/or the transmission performance of the second data.
  • the first data is sent by the first communication device according to the target channel information
  • the second data is sent by the second communication device according to the second channel information.
  • the first communication device determines the applicability of the first channel learning model according to the data transmission performance, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the first communication device can be reduced. deal with the burden.
  • the determining, by the first communication apparatus, whether the first channel learning model is applicable includes: determining, by the first communication apparatus, the A channel learning model is not applicable; or, the first communication device determines that the first channel learning model is applicable when the scene in which it is located has not changed; wherein, the scene includes at least one of the following: indoor static, outdoor static , low-speed motion, high-speed motion, suburban, town, macro station, micro station, in-vehicle scenario, car-to-other device scenario, scenarios defined in the 3rd generation partnership project (3GPP) protocol.
  • 3GPP 3rd generation partnership project
  • the first communication device determines the applicability of the first channel learning model according to whether the scene in which it is located changes, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the number of times can be reduced.
  • the processing burden of a communication device it is easier to implement whether the first channel learning model is applicable according to whether the scene in which it is located changes.
  • the first communication device determining whether the first channel learning model is applicable includes: the performance index of the first communication device in the first channel learning model is smaller than the second In the case of a preset threshold, it is determined that the first channel learning model is not applicable, and the performance index includes continuity and/or authenticity; or, the performance index of the first communication device in the first channel learning model is greater than or equal to the first channel learning model. In the case of two preset thresholds, it is determined that the first channel learning model is applicable.
  • the first communication device determines the applicability of the first channel learning model according to the performance index of the first channel learning model, and does not need to restore the first channel information, so it does not need to perform a large number of calculations, so it can be reduced Processing burden of the first communication device.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device is based on an error between the target channel information and the second channel information It is determined whether the first channel learning model is applicable, the second channel information is determined according to the first channel information and the second channel learning model, and the second channel learning model corresponds to the first channel learning model.
  • the first communication device can determine whether the first channel learning model is applicable according to the error between the target channel information and the second channel information, that is, whether the first channel learning model is applicable is determined by performing channel learning model training. With the assistance of the second communication device, the first communication device can determine whether the channel learning model is applicable, so this method is simple and fast. Determining whether the first channel learning model is applicable while considering the characteristics of the target channel information and the second channel information at the same time can ensure that the first communication device and the second communication device obtain the same or similar characteristics of the channel information, which is helpful for subsequent The performance of data transmission is improved when the channel information is used for data transmission.
  • the method further includes: the first communication apparatus receiving first indication information, where the first indication information is used to instruct the first communication apparatus to perform channel learning model training .
  • the first communication device without the active learning capability can judge the applicability of the first channel learning model in the case of receiving the first indication information.
  • the second communication device can instruct the first communication device to determine whether the first channel learning model is applicable through signaling, so that the second communication device can notify the first communication device in time when it finds that the first channel learning model is not applicable.
  • the communication device performs verification, so that the delay in determining whether the first channel learning model is applicable can be reduced.
  • the first channel learning model can be updated in time to avoid communication performance degradation caused when the first channel learning model is not applicable.
  • the first indication information is also used to indicate one or more of the following:
  • the resources used to transmit the first message, the content of the first message, the form of sending the first message, and the training parameters of the channel learning model, the training parameters include at least one of the following: time for channel learning model training, channel learning Configuration information of the reference signal for model training.
  • the method further includes: the first communication device sending a first request message, where the first request message is used to request one or more of the following: perform channel learning Model training, sending the first message and the first indication information.
  • the first communication device can determine whether the first channel learning model is applicable through a signaling request, so that when the first communication device finds that the first channel learning model is not applicable, it can timely request the second communication device for verification to reduce the delay in determining whether the first channel learning model is applicable.
  • the first channel learning model can be updated in time, so as to avoid communication performance degradation caused when the first channel learning model is not applicable.
  • the first message is further used to indicate one or more configuration parameters used to update the second channel learning model, where the second channel learning model is used to update the second channel learning model according to the The first channel information determines the second channel information.
  • the first communication device may determine one or more configuration parameters for updating the first channel learning model and the second channel learning model, and use them for updating One or more configuration parameters of the second channel learning model are sent to the second communication device, so that the second communication device can update the second channel learning model. Further, the performance of the first communication device and the second communication device to transmit information based on the updated first channel learning model and the second channel learning model is relatively good.
  • the method further includes: the first communication device receiving a second message, where the second message is used to indicate one or more methods for updating the first channel learning model. Multiple configuration parameters.
  • the second communication device may determine one or more configuration parameters for updating the first channel learning model and the second channel learning model, and use them for updating One or more configuration parameters of the first channel learning model are sent to the first communication device, so that the first communication device can update the first channel learning model. Further, the performance of the first communication device and the second communication device to transmit information based on the updated first channel learning model and the second channel learning model is relatively good.
  • the method further includes: the first communication device determines, according to the first parameter, one or more configuration parameters for updating the second channel learning model; wherein, The first parameter includes at least one of the following: a cell identifier of a cell where the first communication device is located, a scene where the first communication device is located, a type of the first communication device, and a geographic location where the first communication device is located.
  • the first communication device may determine the first channel learning model and/or the second channel learning model according to the first parameter, and further determine the configuration parameters for updating the second channel learning model, that is, the channel learning model is not performed.
  • a new channel learning model can be determined after training, which reduces the processing complexity of the first communication device.
  • a communication method may include: a second communication device receives a first message; the second communication device determines, according to the first message, that a first channel learning model is not applicable, and the first channel learning model is not applicable.
  • the model is used to determine first channel information based on target channel information, and the dimension of the first channel information is smaller than the dimension of the target channel information; the second communication device sends first indication information, where the first indication information is used to instruct channel learning Model training.
  • the first channel information is used to determine the second channel information through the second channel learning model, and the second channel information is the same as or similar to the target channel information.
  • the first communication device can judge the applicability of the first channel learning model without the assistance of the second communication device, so the interaction of signaling can be reduced and the applicability of the channel learning model can be judged complexity.
  • the first communication device without the active learning capability can judge the applicability of the first channel learning model.
  • the first indication information is also used to indicate one or more of the following:
  • the resources used to transmit the first message, the content of the first message, the form of the first message, and the training parameters of the channel learning model, the training parameters include at least one of the following: the time of the channel learning model training, the channel learning model Configuration information of the training reference signal.
  • the method further includes: the second communication device receives first request signaling, where the first request signaling is used to request one or more of the following: The channel learning model is trained, and the first message and the first indication information are sent.
  • the first communication device can determine whether the first channel learning model is applicable through a signaling request, so that when the first communication device finds that the first channel learning model is not applicable, it can timely request the second communication device for verification to reduce the delay in determining whether the first channel learning model is applicable.
  • the second communication device receives the request signaling, it can find out in time that the first channel learning model and/or the second channel information is not applicable, and the first communication device can train and update the channel learning model in time, so as to avoid the first channel learning model and/or the second channel information. The communication performance is degraded when the first channel learning model and/or the second channel information is not applicable.
  • the first message is further used to indicate one or more configuration parameters used to update the second channel learning model, where the second channel learning model is used to update the second channel learning model according to the The first channel information determines the second channel information.
  • the first communication device may determine one or more configuration parameters for updating the first channel learning model and the second channel learning model, and use them for updating One or more configuration parameters of the second channel learning model are sent to the second communication device, so that the second communication device can update the second channel learning model. Further, the performance of the first communication device and the second communication device to transmit information based on the updated first channel learning model and the second channel learning model is relatively good.
  • the method further includes: the second communication device sending a second message, where the second message is used to indicate one or more methods for updating the first channel learning model. Multiple configuration parameters.
  • the second communication device may determine one or more configuration parameters for updating the first channel learning model and the second channel learning model, and use them for updating One or more configuration parameters of the first channel learning model are sent to the first communication device, so that the first communication device can update the first channel learning model. Further, the performance of the first communication device and the second communication device to transmit information based on the updated first channel learning model and the second channel learning model is relatively good.
  • the method further includes: the second communication device determines one or more configuration parameters for updating the first channel learning model according to the first parameter; wherein, The first parameter includes at least one of the following: a cell identifier of a cell where the first communication device is located, a scene where the first communication device is located, a type of the first communication device, and a geographic location where the first communication device is located.
  • the second communication device may determine the first channel learning model and/or the second channel learning model according to the first parameter, and further determine the configuration parameters for updating the first channel learning model, that is, the channel learning model is not performed.
  • a new channel learning model can be determined after training, which reduces the processing complexity of the second communication device.
  • a communication method may include: a first communication device determining whether a first channel learning model is applicable, where the first channel learning model is used to determine the first channel information based on target channel information, and the first channel learning model is used to determine first channel information based on target channel information.
  • the data volume of a channel information is less than the data volume of the target channel information; in the case that it is determined that the first channel learning model is not applicable, the first communication device sends a first message, the first message is used to indicate that the first message is used to update the first channel.
  • Configuration parameters of a two-channel learning model where the second channel learning model is used to determine second channel information based on the first channel information.
  • the first channel information is used to determine the second channel information through the second channel learning model, and the second channel information is the same as or similar to the target channel information.
  • the first communication device can judge the applicability of the first channel learning model without the assistance of the second communication device, so the interaction of signaling can be reduced and the applicability of the channel learning model can be judged complexity.
  • the above solution may be suitable for sending information to the second communication device in time when the first communication device finds that the first channel learning model is not applicable, so as to reduce the delay in determining whether the channel learning model is applicable.
  • the second communication device receives the information, it can timely find out that the first channel learning model and/or the second channel information is not applicable, and the first communication device can train and update the channel learning model in time, so as to avoid the first channel learning model. Degradation of communication performance caused when the learning model and/or the second channel information is not applicable.
  • the first communication device may determine one or more configuration parameters for updating the first channel learning model and the second channel learning model, and use them for updating the second channel learning model.
  • One or more configuration parameters of the channel learning model are sent to the second communication device so that the second communication device can update the second channel learning model. Further, the performance of the first communication device and the second communication device to transmit information based on the updated first channel learning model and the second channel learning model is relatively good.
  • the first communication device determines whether the first channel learning model is applicable, including: the amount of change of the long-term statistical characteristics of the first communication device in the target channel is greater than or equal to In the case of the first preset threshold, it is determined that the first channel learning model is not applicable; or, the first communication device determines that the first channel learning model is less than the first preset threshold when the amount of change of the long-term statistical characteristics of the target channel is less than the first preset threshold.
  • a one-channel learning model applies.
  • the first communication device determines the applicability of the first channel learning model according to the long-term statistical characteristics of the target channel, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the first channel can be reduced.
  • the processing burden of the communication device is not limited to the above technical solution.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device determining the first channel learning model according to the received first scheduling information Whether the model is applicable.
  • the first scheduling information is sent by the second communication apparatus according to the second channel information, and the second channel information is determined according to the first channel information and the second channel learning model.
  • the first communication device determines the applicability of the first channel learning model according to the first scheduling information, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the first communication device can be reduced in size processing burden.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device determining whether the first channel learning model is applicable according to data transmission performance .
  • the transmission performance of the data may include the transmission performance of the first data and/or the transmission performance of the second data.
  • the first data is sent by the first communication device according to the target channel information
  • the second data is sent by the second communication device according to the second channel information.
  • the first communication device determines the applicability of the first channel learning model according to the data transmission performance, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so the first communication device can be reduced. deal with the burden.
  • the determining, by the first communication apparatus, whether the first channel learning model is applicable includes: determining, by the first communication apparatus, the A channel learning model is not applicable; or, the first communication device determines that the first channel learning model is applicable when the scene in which it is located has not changed; wherein, the scene includes at least one of the following: indoor static, outdoor static , low-speed motion, high-speed motion, suburban, town, macro station, micro station, in-vehicle scenario, car-to-other device scenario, scenarios defined in the 3GPP protocol.
  • the first communication device determines the applicability of the first channel learning model according to whether the scene in which it is located changes, and does not need to restore the first channel information, so it does not need to perform a large amount of calculations, so the number of The processing burden of a communication device.
  • the first communication device determining whether the first channel learning model is applicable includes: the performance index of the first communication device in the first channel learning model is smaller than the second In the case of a preset threshold, it is determined that the first channel learning model is not applicable, and the performance index includes continuity and/or authenticity; or, the performance index of the first communication device in the first channel learning model is greater than or equal to the first channel learning model. In the case of two preset thresholds, it is determined that the first channel learning model is applicable.
  • the first communication device determines the applicability of the first channel learning model according to the performance index of the first channel learning model, and does not need to restore the first channel information, so it does not need to perform a large amount of calculation, so it can reduce the Processing burden of the first communication device.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device is based on an error between the target channel information and the second channel information It is determined whether the first channel learning model is applicable, the second channel information is determined according to the first channel information and the second channel learning model, and the second channel learning model corresponds to the first channel learning model.
  • the first communication device can determine whether the first channel learning model is applicable according to the error between the target channel information and the second channel information, that is, whether the first channel learning model is applicable is determined by performing channel learning model training. With the assistance of the second communication device, the first communication device can determine whether the channel learning model is applicable, so this method is simple and fast. Determining whether the first channel learning model is applicable while considering the characteristics of the target channel information and the second channel information at the same time can ensure that the first communication device and the second communication device obtain the same or similar characteristics of the channel information, which is helpful for subsequent The performance of data transmission is improved when the channel information is used for data transmission.
  • the first communication device determining whether the first channel learning model is applicable includes: the first communication device determining the first channel according to whether the first indication information is received Whether the learning model is applicable, the first indication information is used to instruct the first communication device to perform channel learning model training.
  • the first indication information is further used to indicate training parameters of the channel learning model, where the training parameters include at least one of the following: time for channel learning model training, configuration information of reference signals for channel learning model training.
  • the second communication device can instruct the first communication device to determine whether the first channel learning model is applicable through signaling, so that the second communication device can notify the first communication device in time when it finds that the first channel learning model is not applicable.
  • the communication device performs verification, so that the delay in determining whether the first channel learning model is applicable can be reduced, and the complexity of determining whether the channel learning model is applicable by the first communication device can be reduced at the same time.
  • the first channel learning model can be updated in time to avoid communication performance degradation caused when the first channel learning model is not applicable.
  • a communication method may include: a second communication device determining whether a second channel learning model is applicable, where the second channel learning model is used to determine the second channel information according to the first channel information, the The first channel information is determined according to the first channel learning model and the target channel information, and the data volume of the first channel information is smaller than the data volume of the target channel information; in the case where it is determined that the second channel learning model is not applicable, the The second communication device sends first indication information, where the first indication information is used to instruct channel learning model training; the second communication device receives a first message, where the first message is used for instructing to update the second channel learning model one or more configuration parameters.
  • the first channel information is used to determine the second channel information through the second channel learning model, and the second channel information is the same as or similar to the target channel information.
  • the second communication device can judge the applicability of the second channel learning model without the assistance of the first communication device, so the interaction of signaling can be reduced and the applicability of the channel learning model can be judged complexity.
  • the above solution may be suitable for sending information to the first communication device in time when the second communication device finds that the second channel learning model is not applicable, so as to reduce the delay in determining whether the channel learning model is applicable.
  • the first communication device receives the information, it can timely find that the second channel learning model and/or the first channel information is not applicable, and the first communication device can timely train and update the channel learning model, thereby avoiding the second channel learning Degradation of communication performance caused when the model and/or the first channel information is not applicable.
  • the second communication device may send the first indication information to the first communication device, and receive the first message sent by the first communication device; further, according to the first message Update the second channel learning model. Further, the performance of the first communication device and the second communication device to transmit information based on the updated first channel learning model and the second channel learning model is relatively good.
  • the first indication information is further used to indicate training parameters of the channel learning model, where the training parameters include at least one of the following: time for channel learning model training, configuration information of reference signals for channel learning model training.
  • the second communication device determines whether the second channel learning model is applicable, including: the amount of change of the long-term statistical characteristics of the target channel by the second communication device is greater than or equal to In the case of the third preset threshold, it is determined that the second channel learning model is not applicable; or the second communication device determines that the second channel learning model is not applicable when the variation of the long-term statistical characteristics of the target channel is less than the third preset threshold. Two-channel learning model is applicable.
  • the second communication device determines the applicability of the second channel learning model according to the long-term statistical characteristics of the target channel, and does not require the first communication device to feed back the target channel information, so the feedback overhead of the first communication device can be reduced, and There is no need to perform a large amount of calculation, and the processing load of the second communication device can be reduced.
  • the second communication device determining whether the second channel learning model is applicable includes: the second communication device determining whether the second channel learning model is suitable according to the first scheduling information Be applicable.
  • the first scheduling information is sent by the second communication apparatus according to the second channel information, and the second channel information is determined according to the first channel information and the second channel learning model.
  • the second communication device determines the applicability of the second channel learning model according to the first scheduling information, and does not require the first communication device to feed back the target channel information, so the feedback overhead of the first communication device can be reduced, and no need Performing a large number of calculations can reduce the processing load of the second communication device.
  • the second communication device determining whether the second channel learning model is applicable includes: the second communication device determining whether the second channel learning model is applicable according to data transmission performance .
  • the transmission performance of the data may include the transmission performance of the first data and/or the transmission performance of the second data.
  • the first data is sent by the first communication device according to the target channel information
  • the second data is sent by the second communication device according to the second channel information.
  • the second communication device determines the applicability of the second channel learning model according to the data transmission performance, and does not require the first communication device to feed back the target channel information, so the feedback overhead of the first communication device can be reduced, and no A large amount of computation can reduce the processing burden of the second communication device.
  • the method further includes: the second communication device receiving a third message, where the third message is used to indicate a scene where the first communication device is located, the scene includes At least one of the following: indoor stationary, outdoor stationary, low-speed motion, high-speed motion, suburban, urban, macro station, micro station, vehicle-mounted scene, vehicle-to-other equipment scene, and a scene defined in the 3GPP protocol; the second communication device determines the first Whether the two-channel learning model is applicable includes: the second communication device determines that the second channel learning model is not applicable when the scene changes; or, the second communication device determines when the scene does not change This second channel learning model applies.
  • the second communication device determines the applicability of the second channel learning model according to the scene where the first communication device is located, and does not require the first communication device to feed back target channel information, so the feedback overhead of the first communication device can be reduced , and does not need to perform a large amount of calculation, which can reduce the processing burden of the second communication device.
  • determining whether the first channel learning model is applicable according to whether the scene in which the first communication device is located changes is easier to implement.
  • the second communication device determining whether the second channel learning model is applicable includes: the second communication device is based on an error between the target channel information and the second channel information It is determined whether the second channel learning model is applicable.
  • the second communication device can determine whether the first channel learning model is applicable according to the error between the target channel information and the second channel information, that is, whether the second channel learning model is applicable is determined by performing channel learning model training. With the assistance of the first communication device, the second communication device can determine whether the channel learning model is applicable, so this method is simple and fast. Determining whether the second channel learning model is applicable while considering the characteristics of the target channel information and the second channel information at the same time can ensure that the first communication device and the second communication device obtain the same or similar characteristics of the channel information, which is helpful for subsequent The performance of data transmission is improved when the channel information is used for data transmission.
  • a communication apparatus including various modules or units for performing the method in the first aspect and any possible implementation manner of the first aspect.
  • a communication apparatus including various modules or units for performing the method in the second aspect and any possible implementation manner of the second aspect.
  • a communication device comprising a transceiver unit and a processing unit: the processing unit is configured to determine whether a first channel learning model is applicable, and the first channel learning model is configured to determine the first channel information based on target channel information, The data amount of the first channel information is smaller than the data amount of the target channel information; in the case that the first channel learning model is determined to be inapplicable, the transceiver unit is configured to send a first message, where the first message is used to indicate A configuration parameter of a second channel learning model for determining second channel information based on the first channel information is updated.
  • the processing unit is specifically configured to: determine whether the first channel learning model is applicable according to whether the variation of the long-term statistical characteristics of the channel is greater than a first preset threshold.
  • the processing unit is specifically configured to: determine the first channel when the variation of the long-term statistical characteristics of the target channel is greater than or equal to a first preset threshold The learning model is not applicable; or, when the variation of the long-term statistical characteristics of the target channel is less than the first preset threshold, it is determined that the first channel learning model is applicable.
  • the processing unit is specifically configured to: determine whether the first channel learning model is applicable according to the data transmission performance.
  • the processing unit is specifically configured to: determine that the first channel learning model is not applicable when the scene in which it is located changes; or, in the scene in which it is located In the case of no change, it is determined that the first channel learning model is applicable, wherein the scene includes at least one of the following: indoor stationary, outdoor stationary, low-speed motion, high-speed motion, suburban, urban, macro station, micro station, vehicle scene , vehicle to other equipment scenarios, scenarios defined in 3GPP protocol.
  • the processing unit is specifically configured to: determine the first channel learning model when the performance index of the first channel learning model is less than a second preset threshold Not applicable, the performance index includes continuity and/or authenticity; or, when the performance index of the first channel learning model is greater than or equal to the second preset threshold, it is determined that the first channel learning model is applicable.
  • the processing unit is specifically configured to: determine whether the first channel learning model is applicable according to the error between the target channel information and the second channel information, and the second channel The information is determined according to the first channel information and a second channel learning model, where the second channel learning model corresponds to the first channel learning model.
  • the processing unit is specifically configured to: determine whether the first channel learning model is applicable according to whether first indication information is received, and the first indication information is used to indicate the The first communication device performs channel learning model training.
  • a communication device comprising a transceiver unit and a processing unit: the processing unit is configured to determine whether a second channel learning model is applicable, and the second channel learning model is configured to determine the second channel information according to the first channel information , the first channel information is determined according to the first channel learning model and the target channel information, and the data volume of the first channel information is less than the data volume of the target channel information; in the case where it is determined that the second channel learning model is not applicable , the transceiver unit is used to send first indication information, the first indication information is used to instruct channel learning model training; the transceiver unit is also used to receive a first message, the first message is used to instruct to update the second One or more configuration parameters for the channel learning model.
  • the processing unit is specifically configured to: determine the second channel when the variation of the long-term statistical characteristics of the target channel is greater than or equal to a third preset threshold The learning model is not applicable; or, when the variation of the long-term statistical characteristics of the target channel is less than the third preset threshold, it is determined that the second channel learning model is applicable.
  • the processing unit is specifically configured to: determine whether the second channel learning model is applicable according to the first scheduling information.
  • the processing unit is specifically configured to: determine whether the second channel learning model is applicable according to data transmission performance.
  • the transceiver unit is further configured to receive a third message, where the third message is used to indicate a scene in which the first communication device is located, and the scene includes at least one of the following : indoor stillness, outdoor stillness, low-speed motion, high-speed motion, suburban, town; the processing unit is specifically used to: determine that the second channel learning model is not applicable when the scene changes; or, when the scene does not change In the case of , it is determined that the second channel learning model is applicable.
  • the processing unit is specifically configured to: determine whether the second channel learning model is applicable according to an error between the target channel information and the second channel information.
  • a communication apparatus including a processor.
  • the processor is coupled to the memory and can be used to execute instructions in the memory, so as to implement the method in any possible implementation manner of the first aspect and the third aspect.
  • the communication device further includes a memory.
  • the communication device further includes a communication interface, and the processor is coupled to the communication interface.
  • the communication device is a first communication device.
  • the communication interface may be a transceiver, or an input/output interface.
  • the communication device is a chip configured in the first communication device.
  • the communication interface may be an input/output interface.
  • the transceiver may be a transceiver circuit.
  • the input/output interface may be an input/output circuit.
  • a communication apparatus including a processor.
  • the processor is coupled to the memory and can be used to execute instructions in the memory, so as to implement the method in any of the possible implementation manners of the second aspect and the fourth aspect.
  • the communication device further includes a memory.
  • the communication device further includes a communication interface, and the processor is coupled to the communication interface.
  • the communication device is a second communication device.
  • the communication interface may be a transceiver, or an input/output interface.
  • the communication device is a chip configured in the second communication device.
  • the communication interface may be an input/output interface.
  • the transceiver may be a transceiver circuit.
  • the input/output interface may be an input/output circuit.
  • a processor comprising: an input circuit, an output circuit and a processing circuit.
  • the processing circuit is configured to receive a signal through the input circuit and transmit a signal through the output circuit, so that the processor executes the method in any one of the possible implementations of the first aspect to the fourth aspect.
  • the above-mentioned processor may be one or more chips
  • the input circuit may be input pins
  • the output circuit may be output pins
  • the processing circuit may be transistors, gate circuits, flip-flops and various logic circuits, etc. .
  • the input signal received by the input circuit may be received and input by, for example, but not limited to, a receiver
  • the signal output by the output circuit may be, for example, but not limited to, output to and transmitted by a transmitter
  • the circuit can be the same circuit that acts as an input circuit and an output circuit at different times.
  • the embodiments of the present application do not limit the specific implementation manners of the processor and various circuits.
  • a twelfth aspect provides a processing apparatus including a processor and a memory.
  • the processor is configured to read the instructions stored in the memory, and can receive signals through the receiver and transmit signals through the transmitter, so as to execute the method in any possible implementation manner of the first aspect to the fourth aspect.
  • processors there are one or more processors and one or more memories.
  • the memory may be integrated with the processor, or the memory may be provided separately from the processor.
  • the memory can be a non-transitory memory, such as a read only memory (ROM), which can be integrated with the processor on the same chip, or can be separately set in different On the chip, the embodiment of the present application does not limit the type of the memory and the setting manner of the memory and the processor.
  • ROM read only memory
  • the relevant data interaction process such as sending indication information, may be a process of outputting indication information from the processor, and receiving capability information may be a process of receiving input capability information by the processor.
  • the data output by the processor can be output to the transmitter, and the input data received by the processor can be from the receiver.
  • the transmitter and the receiver may be collectively referred to as a transceiver.
  • the processing device in the twelfth aspect above may be one or more chips.
  • the processor in the processing device may be implemented by hardware or by software.
  • the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor, implemented by reading software codes stored in a memory, which can Integrated in the processor, can be located outside the processor, independent existence.
  • a thirteenth aspect provides a computer program product, the computer program product comprising: a computer program (also referred to as code, or instructions), which, when the computer program is executed, causes the computer to execute the above-mentioned first aspect to The method in any possible implementation manner of the fourth aspect.
  • a computer program also referred to as code, or instructions
  • a fourteenth aspect provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program (which may also be referred to as code, or instructions), when it is run on a computer, so that the above-mentioned first aspect to The method in any of the possible implementations of the fourth aspect is performed.
  • a computer program which may also be referred to as code, or instructions
  • a communication system including the aforementioned first communication device and second communication device.
  • FIG. 1 shows a schematic diagram of a communication system suitable for this embodiment of the present application.
  • Figure 2 shows a schematic diagram of the structure of the neural network.
  • Figure 3 shows a schematic diagram of the structure of a symmetric neural network.
  • FIG. 4 shows a schematic diagram of the operation performed by the neural network.
  • FIG. 5 shows a schematic diagram of operations performed by a one-dimensional convolutional neural network.
  • FIG. 6 shows a schematic diagram of operations performed by a two-dimensional convolutional neural network.
  • FIG. 7 shows a schematic diagram of operations performed by the excitation layer of the convolutional neural network.
  • Figure 8 shows a schematic diagram of the operation of the pooling layer of the convolutional neural network.
  • Figure 9 shows a schematic diagram of the structure of a convolutional neural network.
  • FIG. 10 shows a schematic flowchart of a communication method provided by an embodiment of the present application.
  • FIG. 11 shows a schematic diagram of the location where the terminal device is located.
  • FIG. 18 shows a schematic diagram of a communication apparatus provided by an embodiment of the present application.
  • FIG. 19 shows a schematic block diagram of a communication apparatus provided by another embodiment of the present application.
  • FIG. 20 shows a schematic diagram of a chip system provided by an embodiment of the present application.
  • LTE Long Term Evolution
  • FDD frequency division duplex
  • TDD time division duplex
  • UMTS universal mobile telecommunication system
  • WiMAX worldwide interoperability for microwave access
  • 5G mobile communication system fifth generation
  • 5G fifth generation
  • NR new radio access Technology
  • 6G sixth generation
  • the 5G mobile communication system may include a non-standalone (NSA, NSA) and/or an independent network (standalone, SA).
  • NSA non-standalone
  • SA independent network
  • the technical solutions of the embodiments of the present application can also be applied to satellite communication systems, high altitude platform station (HAPS) communication and other non-terrestrial network (NTN) systems, as well as various mobile systems integrated with satellite communication systems.
  • HAPS high altitude platform station
  • NTN non-terrestrial network
  • the technical solutions provided in this application can also be applied to machine type communication (MTC), Long Term Evolution-machine (LTE-M), and device to device (D2D) networks.
  • M2M Machine to Machine
  • IoT Internet of Things
  • the IoT network may include, for example, the Internet of Vehicles.
  • vehicle to X vehicle to X
  • V2X vehicle and vehicle Infrastructure
  • V2I vehicle to pedestrian
  • V2N vehicle to network
  • the network device may be any device with a wireless transceiver function.
  • the device includes but is not limited to: evolved Node B (evolved Node B, eNB), radio network controller (radio network controller, RNC), Node B (Node B, NB), base station controller (base station controller, BSC) , base transceiver station (base transceiver station, BTS), home base station (for example, home evolved NodeB, or home Node B, HNB), baseband unit (baseband unit, BBU), wireless fidelity (wireless fidelity, WiFi) system Access point (AP), wireless relay node, wireless backhaul node, transmission point (TP) or transmission and reception point (TRP), etc.
  • evolved Node B evolved Node B
  • RNC radio network controller
  • Node B Node B
  • BSC base station controller
  • base transceiver station base transceiver station
  • BTS home base station
  • home base station for example, home evolved NodeB, or home Node B, HNB
  • It can also be 5G, such as NR , a gNB in the system, or, a transmission point (TRP or TP), one or a group of (including multiple antenna panels) antenna panels of a base station in a 5G system, or, it can also be a network node that constitutes a gNB or a transmission point, For example, a baseband unit (BBU), or a distributed unit (distributed unit, DU), etc., or a base station in a future communication system, etc.
  • BBU baseband unit
  • DU distributed unit
  • a gNB may include a centralized unit (CU) and a DU.
  • the gNB may also include an active antenna unit (AAU).
  • CU implements some functions of gNB
  • DU implements some functions of gNB.
  • CU is responsible for processing non-real-time protocols and services, implementing radio resource control (RRC), and packet data convergence protocol (PDCP) layer function.
  • RRC radio resource control
  • PDCP packet data convergence protocol
  • the DU is responsible for processing physical layer protocols and real-time services, and implementing the functions of the radio link control (RLC) layer, medium access control (MAC) layer, and physical (PHY) layer.
  • RLC radio link control
  • MAC medium access control
  • PHY physical layer.
  • AAU implements some physical layer processing functions, radio frequency processing and related functions of active antennas.
  • the higher-layer signaling such as the RRC layer signaling
  • the network device may be a device including one or more of a CU node, a DU node, and an AAU node.
  • the CU can be divided into network devices in an access network (radio access network, RAN), and the CU can also be divided into network devices in a core network (core network, CN), which is not limited in this application.
  • the network equipment provides services for the cell, and the terminal equipment communicates with the cell through transmission resources (for example, frequency domain resources, or spectrum resources) allocated by the network equipment, and the cell may belong to a macro base station (for example, a macro eNB or a macro gNB, etc.) , can also belong to the base station corresponding to the small cell, where the small cell can include: urban cell (metro cell), micro cell (micro cell), pico cell (pico cell), femto cell (femto cell), etc. , these small cells have the characteristics of small coverage and low transmission power, and are suitable for providing high-speed data transmission services.
  • a macro base station for example, a macro eNB or a macro gNB, etc.
  • the small cell can include: urban cell (metro cell), micro cell (micro cell), pico cell (pico cell), femto cell (femto cell), etc.
  • these small cells have the characteristics of small coverage and low transmission power, and are suitable for providing high-speed data transmission
  • a terminal device may also be referred to as user equipment (user equipment, UE), an access terminal, a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, Terminal, wireless communication device, user agent or user equipment.
  • user equipment user equipment
  • UE user equipment
  • an access terminal a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, Terminal, wireless communication device, user agent or user equipment.
  • the terminal device may be a device that provides voice/data connectivity to the user, such as a handheld device with a wireless connection function, a vehicle-mounted device, and the like.
  • some examples of terminals can be: mobile phone (mobile phone), unmanned aerial vehicle, tablet computer (pad), computer with wireless transceiver function (such as notebook computer, PDA, etc.), mobile internet device (mobile internet device, MID) ), virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving, remote medical ), wireless terminals in smart grid, wireless terminals in transportation safety, wireless terminals in smart city, wireless terminals in smart home, Cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication capabilities, Computing equipment or other processing equipment connected to a wireless modem, in-vehicle equipment, wearable equipment, terminal equipment in a 5G network or terminal equipment in a future evolved public land mobile network
  • wearable devices can also be called wearable smart devices, which is a general term for the intelligent design of daily wear and the development of wearable devices using wearable technology, such as glasses, gloves, watches, clothing and shoes.
  • a wearable device is a portable device that is worn directly on the body or integrated into the user's clothing or accessories.
  • Wearable device is not only a hardware device, but also realizes powerful functions through software support, data interaction, and cloud interaction.
  • wearable smart devices include full-featured, large-scale, complete or partial functions without relying on smart phones, such as smart watches or smart glasses, and only focus on a certain type of application function, which needs to cooperate with other devices such as smart phones. Use, such as all kinds of smart bracelets, smart jewelry, etc. for physical sign monitoring.
  • the terminal device may also be a terminal device in an internet of things (Internet of things, IoT) system.
  • IoT Internet of things
  • IoT is an important part of the development of information technology in the future. Its main technical feature is to connect items to the network through communication technology, so as to realize the intelligent network of human-machine interconnection and interconnection of things.
  • IoT technology can achieve massive connections, deep coverage, and terminal power saving through, for example, narrow band (NB) technology.
  • NB narrow band
  • terminal equipment can also include sensors such as smart printers, train detectors, and gas stations.
  • the main functions include collecting data (part of terminal equipment), receiving control information and downlink data of network equipment, and sending electromagnetic waves to transmit uplink data to network equipment. .
  • FIG. 1 shows a schematic diagram of a communication system 100 suitable for the method provided by this embodiment of the present application.
  • the communication system 100 may include at least one network device, such as the network device 101 in the 5G system as shown in FIG. 1 ; the communication system 100 may also include at least one terminal device, as shown in FIG. 1 .
  • Terminal devices 102 to 107 may be mobile or stationary.
  • Each of the network device 101 and one or more of the end devices 102 to 107 may communicate over a wireless link.
  • Each network device can provide communication coverage for a specific geographic area and can communicate with terminal devices located within that coverage area. For example, the network device may send configuration information to the terminal device, and the terminal device may send uplink data to the network device based on the configuration information; for another example, the network device may send downlink data to the terminal device. Therefore, the network device 101 and the terminal devices 102 to 107 in FIG. 1 constitute a communication system.
  • D2D or V2X technology can be used to realize direct communication between terminal devices.
  • D2D or V2X technology can be used for direct communication between terminal devices 105 and 106 and between terminal devices 105 and 107 .
  • Terminal device 106 and terminal device 107 may communicate with terminal device 105 individually or simultaneously.
  • the terminal devices 105 to 107 can also communicate with the network device 101, respectively. For example, it can communicate directly with the network device 101, as shown in the figure, the terminal devices 105 and 106 can directly communicate with the network device 101; it can also communicate with the network device 101 indirectly, as in the figure, the terminal device 107 communicates with the network device via the terminal device 106. 101 Communications.
  • FIG. 1 exemplarily shows a network device, a plurality of terminal devices, and communication links between the communication devices.
  • the communication system 100 may include multiple network devices, and the coverage of each network device may include other numbers of terminal devices, such as more or less terminal devices.
  • the communication system 100 may include multiple network devices, and the multiple network devices may perform coordinated multipoint transmission, for example, multiple network devices may cooperate or cooperate to communicate with one terminal device. This application does not limit this.
  • Each of the above communication devices may be configured with multiple antennas.
  • the plurality of antennas may include at least one transmit antenna for transmitting signals and at least one receive antenna for receiving signals.
  • the transmitting antenna and the receiving antenna may be the same or different. For example, in the same situation, one antenna can be used for both transmission and reception; in different situations, the transmission antenna and the reception antenna are different antennas.
  • each communication device additionally includes a transmitter chain and a receiver chain, which can be understood by those of ordinary skill in the art, all of which may include multiple components (eg, processors, modulators, multiplexers) related to signal transmission and reception. , demodulator, demultiplexer or antenna, etc.). Therefore, the network device and the terminal device can communicate through the multi-antenna technology.
  • the wireless communication system 100 may further include other network entities such as a network controller, a mobility management entity, and the like, which are not limited in this embodiment of the present application.
  • network entities such as a network controller, a mobility management entity, and the like, which are not limited in this embodiment of the present application.
  • the MIMO technology is usually used to increase the system capacity, that is, multiple antennas are used simultaneously at the transmitting end and the receiving end.
  • the use of multiple antennas combined with space division multiplexing can double the system capacity, but in fact, due to the use of multiple antennas, it also brings about the problem of interference enhancement, so it is often necessary to process the signal to suppress the interference. impact.
  • This method of interference suppression through signal processing can be implemented at the receiving end or at the transmitting end.
  • the signal to be sent can be preprocessed, and then sent through the MIMO channel. This sending method is precoding.
  • the elements on the line) are the singular values of the channel matrix H, and these singular values can usually be arranged in descending order, and the superscript "T" indicates the transposition operation.
  • the MIMO channel needs to be known, so the MIMO channel needs to be estimated.
  • the network device can obtain downlink channel information according to the measured uplink channel information, so as to perform precoding matrix calculation and downlink transmission.
  • the network device performs precoding matrix calculation and downlink transmission according to the channel state information fed back by the terminal device.
  • the terminal device and the network device obtain a first channel learning model and a second channel learning model through joint training, wherein the first channel learning model is set in the terminal.
  • the device side and the second channel learning model are set on the network device side, and the communication system including the first channel learning model and the second channel learning model is called an AI communication system.
  • the AI-based information transmission method includes: the terminal device obtains the information to be fed back, the terminal device processes the information to be fed back through at least a first channel learning model to obtain the information that needs to be fed back through the air interface, and the terminal device then uses the feedback The link feeds back the information that needs to be fed back through the air interface to the network device, the network device receives the information fed back by the terminal device, and the network device processes the feedback information at least through the second channel learning model to obtain the information to be fed back by the terminal device. information.
  • the first channel learning model and the second channel learning model obtained by offline training are directly applied in the process of information online transmission.
  • the change may cause the first channel learning model and the second channel learning model to be inapplicable, thereby affecting the performance of information transmission based on the information transmission method of the AI based on the first channel learning model and the second channel learning model.
  • an embodiment of the present application provides a communication method, so as to realize the judgment of the applicability of the channel learning model being used.
  • the channel learning model is first introduced and described below.
  • the channel learning model mentioned in the embodiments of the present application may be a model or algorithm for channel acquisition, or may be a model or algorithm for determining channel information, or may be a model or algorithm related to a channel, or
  • the model or algorithm, etc. applied in the communication system is not limited in this embodiment of the present application.
  • the channel learning model may be a machine learning algorithm, etc., for example, may be one or more of the following machine learning algorithms: decision tree algorithm, naive Bayesian algorithm, support vector machine algorithm, random forest algorithm, artificial neural network Algorithms, boosting and bagging algorithms, expectation maximization (EM) algorithms, deep learning.
  • machine learning algorithms decision tree algorithm, naive Bayesian algorithm, support vector machine algorithm, random forest algorithm, artificial neural network Algorithms, boosting and bagging algorithms, expectation maximization (EM) algorithms, deep learning.
  • EM expectation maximization
  • the channel learning model may be a neural network (NN) model, for example, may be at least one of the following neural networks: convolutional neural network, fully connected neural network, deep neural network, feedforward neural network , Feedback Neural Networks, Radial Basis Neural Networks, Hopfield Networks, Markov Chains, Boltzmann Machines, Restricted Boltzmann Machines, Autoencoders, Sparse Autoencoders, Variational Autoencoders Machines, Denoising Autoencoders, Deep Belief Networks, Deconvolutional Networks, Deep Convolutional Inverse Graph Networks, Generative Adversarial Networks, Recurrent Neural Networks, Long Short-Term Memory, Neural Turing Machines, Deep Residual Networks, Echo State Networks, Extreme Learning Machines, Support Vector Machines.
  • NN neural network
  • the channel learning model may also be: a principal component analysis algorithm, a matrix eigenvalue decomposition algorithm, a matrix eigenvector decomposition algorithm, or a matrix singular value decomposition algorithm.
  • the channel learning model may be an auto-encoder (auto-encoder, AE) model or the like.
  • the channel learning model may be a model or algorithm for realizing channel dimension reduction and/or channel recovery (or channel reconstruction).
  • the channel learning model is further introduced and explained below by taking the channel learning model as a neural network model as an example.
  • the neural network is mainly composed of an input layer, a hidden layer and an output layer.
  • Figure 2 shows a basic neural network.
  • the network can be called a two-layer neural network. Since the input layer has not undergone any transformation, it can not be regarded as a separate layer.
  • each neuron in the input layer of the network represents a feature, and the dimension of the input layer can also be called the input dimension.
  • the number of output layers can represent the number of classification labels, and the dimension of the output layer can also be called the output dimension.
  • the number of hidden layers and hidden layer neurons can be set to positive integers.
  • Figure 3 shows a symmetric neural network.
  • the neural network can also be called auto-encoding.
  • the neural network includes encoding (from M dimension to D dimension), namely f: c n (high-dimensional data) ⁇ z n (low-dimensional data); decoding (from D dimension to M dimension), namely f -1 : z n (low-dimensional data) ⁇ (high-dimensional data); where M can be greater than D.
  • the average approximation error can be used as the loss function to train the neural network, and there can be many specific training algorithms, such as back-propagation algorithm, gradient descent algorithm, etc., which are not limited in this application.
  • the average training error can be expressed as:
  • c n is an M-dimensional channel vector
  • z n is a D-dimensional channel vector.
  • the network device can use the decoding equation f -1 of the neural network to restore the low-dimensional data into high-dimensional data.
  • the channel matrix may be a complex number of A*B*S dimensions.
  • A is the number of antenna ports of the network device
  • B is the number of antenna ports of the terminal device
  • S is the number of subcarriers.
  • M can be A*B*S (the real part and the imaginary part are input independently, or the complex number is input) or A*B*S*2 (the real part and the imaginary part are jointly input).
  • A 64
  • B is 1, and S is 1, then M is 64 or 128.
  • D can be a positive integer, such as 2, 4, 5, 6, 8, 16, 32, etc.
  • the neural network shown in FIG. 3 is a 4-layer neural network structure, and the number of neurons in each layer can be gradually reduced. For example, if M is 64 and D is 4, the 64-dimensional high-dimensional channel information can be reduced to 4-dimensional channel information.
  • the encoding and decoding equations may be symmetric or asymmetric, that is, the two may adopt the same structure or different structures.
  • the value (activation value) of each hidden layer neuron/output layer neuron is calculated by the neurons in the previous layer, after operations (such as weighted summation, weighted summation plus bias, etc.) and nonlinear transformation obtained.
  • the nonlinear transformation function also called activation function
  • the activation function is the Ranh function and the Relu function as an example.
  • Sigmoid function It is a sigmoid function common in biology, also known as the sigmoid growth curve. In information science, the sigmoid function is often used as the threshold function of neural networks due to its mono-increasing and inverse-function mono-increasing properties. Mapping variables between 0 and 1, the sigmoid function can be represented by the following formula:
  • Tanh function It is one of the hyperbolic functions. In mathematics, the Tanh function is derived from the basic hyperbolic function, hyperbolic sine function and hyperbolic cosine function. The Tanh function can be represented by the following formula:
  • Relu function used to calculate the output of hidden layer neurons.
  • the Relu function can be represented by the following formula:
  • a variant of the Relu activation function which can be a leaky linear rectification function (leakly Relu, LRelu), a leaky random linear rectification function (random leaky Relu, RRelu), or a parameter linear rectification function (parameter Relu, PRelu), etc.
  • the LRelu function can be expressed by the following formula:
  • the PRelu function can be represented by the following formula:
  • ⁇ is a variable that can be learned by backpropagation.
  • the RRelu function can be expressed by the following formula:
  • each transformation can include weighted summation plus bias and nonlinear transformation operations.
  • H g(X*W+b).
  • W is a weight matrix or a weight vector, referred to as a weight
  • b is a bias vector or a bias matrix, referred to as a bias
  • g() is an activation function.
  • X is a matrix of 1*x and the number of neurons in the hidden layer is h
  • W can be a matrix of dimension x*h
  • b can be a matrix of dimension 1*h.
  • h 1 g 1 (x 1 *w 1,1,1 +x 2 *w 1,1,2 +b 1,1 ),
  • h 2 g 1 (x 1 *w 1,2,1 +x 2 *w 1,2,2 +b 1,2 ),
  • h 50 g 1 (x 1 *w 1,50,1 +x 2 *w 1,50,2 +b 1,50 ).
  • the function g 1 may be an activation function, such as a Sigmoid function, a Tanh function or a Relu function.
  • y 1 g 2 (h 1 *w 2,1,1 +h 2 *w 2,1,2 +...+h 50 *w 2,1,50 +b 2,1 ),
  • y 2 g 2 (h 1 *w 2,2,1 +h 2 *w 2,2,2 +...+h 50 *w 2,2,50 +b 2,2 ),
  • y 3 g 2 (h 1 *w 2,3,1 +h 2 *w 2,3,2 +...+h 50 *w 2,3,50 +b 2,3 ),
  • y 4 g 2 (h 1 *w 2,4,1 +h 2 *w 2,4,2 +...+h 50 *w 2,4,50 +b 2,4 ).
  • the function g 2 may be an activation function, such as a Sigmoid function, a Tanh function or a Relu function.
  • the configuration parameters of the channel learning model involved in this step can be transformation algorithm, weight vector, weight matrix, bias vector, bias matrix, activation function, input layer dimension, output layer dimension, hidden layer number, hidden layer One or more of the number of neurons in the layer, etc.
  • the channel learning model is further introduced and explained below by taking the channel learning model as a convolutional neural network model as an example.
  • the neural network is first applied to the image, so the image is used as an example below.
  • applying neural networks to channel learning is similar, except that the input signal is changed from an image to channel data.
  • each pixel of the image is often connected to each neuron in the fully connected layer, while the convolutional neural network only connects each hidden layer neuron with A local area of the image is connected, thereby reducing the number of parameter training. For example, if a 1024 ⁇ 720 image is used with a 9 ⁇ 9 receptive field, only 81 weight parameters are required. The same is true for general vision, when viewing an image, the focus is more on the part.
  • each neuron In the convolutional layer of the convolutional neural network, the weights corresponding to each neuron are the same, that is, it can be considered that each neuron only focuses on one feature.
  • the neuron can be a filter, such as a Sobel filter dedicated to edge detection, so it can be considered that each filter of the convolutional layer will have a data feature that it focuses on, such as mean, variance, amplitude, Phase, vertical edge, horizontal edge, color, texture, etc. All the neurons in the convolutional layer can be considered as the feature extractor set of the entire data after adding up, so the convolution is the inner product of a set of fixed weights and the data in different windows.
  • weight sharing On the one hand, the repeating unit is able to identify the feature regardless of its position in the viewing field. On the other hand, weight sharing enables more efficient feature extraction because the number of free variables to be learned is greatly reduced. By controlling the scale of the model, convolutional networks can generalize well to channel problems.
  • Convolutional neural network can include data input layer (input layer), convolution calculation layer (convolution layer), Relu excitation layer (Relu layer), pooling layer (pooling layer), fully connected layer (fully connected layer, FC layer) .
  • Data input layer The processing of this layer is mainly to preprocess the original data, which can include one or more of the following:
  • De-Meaning Centering all dimensions of the input data to 0, the purpose is to pull the center of the sample back to the origin of the coordinate system.
  • Normalization Normalize the amplitude of the input data to the same range, that is, reduce the interference caused by the difference in the value range of the data in each dimension. For example, suppose the input data has two-dimensional features (for example, feature A 1 and feature B 1 ), the range of data of feature A 1 is 0 to 10, and the range of data of feature B 1 is 0 to 10000. If it is problematic to directly use the data of these two features, it is a better practice to The amplitude of the data of the two features is normalized, that is, the data of the feature A 1 and the feature B 1 can be changed into the range of 0 to 1.
  • PCA Principal component analysis
  • whitening is used to normalize the amplitudes on each feature axis of the input data.
  • the configuration parameters involved in this step may be one or more of the preprocessing operation algorithm, the dimension of the input data, and the value range of the input data.
  • Convolutional computing layer This layer is the most important layer of the convolutional neural network, and it is also the source of the name of the "convolutional neural network”.
  • Depth You can control the depth of the output unit, that is, the number of filters, or the number of neurons connected to the same area. Also known as: depth column (depth column);
  • Stride also known as stride, that is, the length of the window sliding at a time.
  • Zero-padding that is, padding zeros around the input unit to change the overall size of the input unit, thereby controlling the spatial size of the output unit.
  • W 1 the size (width or height) of the input unit
  • F the receptive field
  • S 1 the stride
  • P the number of zero padding
  • K the depth, that is, the depth of the output unit.
  • the calculation result is not an integer, it means that the existing parameters cannot fit the input exactly.
  • the stride setting may be inappropriate. In this case, you can fill with zeros or reset the steps.
  • Figure 5 shows an example of a one-dimensional convolution.
  • the convolution includes a weight parameter, the weight parameter may be an F-dimensional vector or matrix, and the element value of the weight vector or the weight matrix may be an integer, a real number, or a complex number.
  • the weight parameter shown in Figure 5 is a 3-dimensional vector: [1 0-1].
  • the calculation method of the data output by the hidden unit can be the same as the calculation method of the ordinary neural network output data described above.
  • W 1 can be a matrix of w 1 *w 2
  • F can be a matrix of f 1 *f 2
  • S 1 can be a matrix of s 1 *s 2
  • P can be p 1 *p 2
  • the matrix of , K can be a matrix of k 1 *k 2 .
  • w 1 , f 1 , s 1 , p 1 , and k 1 represent the number of rows of the matrix, respectively
  • w 2 , f 2 , s 2 , p 2 , and k 2 represent the number of columns of the matrix, respectively.
  • Figure 6 shows an example of a two-dimensional convolution.
  • the convolution includes a weight parameter, which may be an F-dimensional vector or matrix, and the element value of the weight vector or weight matrix may be an integer, a real number, or a complex number.
  • the weight parameter shown in Figure 6 is a 3*3-dimensional matrix:
  • the calculation method of the data output by the hidden unit can be the same as the calculation method of the ordinary neural network output data described above.
  • the output data of the first hidden unit in the first row is: -8; the output data of the second hidden unit in the second row is: 8.
  • the configuration parameters involved in this step can be the size (width or height) of the input unit; receptive field; stride;
  • Excitation layer Non-linear mapping of the output results of the convolutional layer.
  • the excitation function used by the convolutional neural network can generally be a Relu function.
  • the Relu function is characterized by fast convergence and simple gradient calculation.
  • the calculation formula is also very simple (formula (4)), that is, for the negative value of the input, the output is all 0, and for the positive value of the input, the output is as it is.
  • the first input value of the first line is 0.77, then take max(0,0.77) to get the output value of 0.77; the second value of the second line is -0.11, then take max(0,-0.11 ) to get the output value of 0, and so on to get all the output results.
  • the configuration parameter involved in this step may be an activation function algorithm, and the activation function may be one or more of the activation functions described above.
  • pooling layer is sandwiched between consecutive convolutional layers to compress the amount of data and parameters and reduce overfitting.
  • the main role of the pooling layer is data compression. Therefore, pooling is downsampling, the purpose is to reduce the feature map.
  • the pooling operation is to adjust the size of the image. For example, an image of a dog can be recognized as a photo of a dog when it has been shrunk twice, which means that the most important features of the dog are still preserved in this image. only. That is to say, the information removed during image compression is only some irrelevant information, while the remaining information is the feature with scale invariance and the feature that can best express the image.
  • the initial input channel contains a lot of information and many features, but some information is not very useful or repeated for channel learning tasks, so this kind of redundant information can be removed, and the most important The features are extracted, which is also a major role of the pooling operation. Therefore, the pooling operation prevents overfitting to a certain extent and is more convenient for optimization.
  • the operation performed by the pooling layer generally adopts one or more of the following: 1) Maximum pooling (max pooling): that is, taking the maximum value. For example, if an N 1 -dimensional input value is pooled into 1 output value, the output value is the maximum value among the N 1 values. 2) Mean pooling (mean pooling): that is, taking the mean value. For example, if N 1 -dimensional input values are pooled into 1 output value, the output value is the average of N 1 values. 3) Gaussian pooling: borrow the method of Gaussian blurring. 4) Trainable pooling: that is, the training function f accepts N 1 points as input, and outputs N 2 points, and N 2 is less than N 1 to achieve dimensionality reduction. 5) Overlapping pooling. 6) Empty pyramid pooling.
  • the input matrix is 4*4 and the output matrix is 2*2 as an example, in this process, 4 input values need to be converted into one output value .
  • the maximum value in each 2*2 window is used as the output value. For example, if the maximum value in the first 2*2 window of the input matrix is 6, then the first element of the output matrix is 6, and so on, to get all the output results.
  • the pooling operation will keep the depth size unchanged. If the size of the input unit of the pooling layer is not an integer multiple of 2, it can generally be supplemented by a multiple of 2 by zero-padding, and then pooled.
  • the receiving unit size is: W 1 *H 1 *D 1 ; two hyperparameters are required: spatial extent F 1 , stride S 2 .
  • the configuration parameters involved in this step can be one or more of the spatial range, stride, pooling algorithm, input size (or receiving unit size), output size, etc.
  • Fully connected layer All neurons between the two layers have weight connections, usually the fully connected layer is at the tail of the convolutional neural network. That is, it is the same as the connection method of the traditional neural network neurons.
  • the output value of each fully connected layer neuron is obtained by performing operations (such as weighted summation, weighted summation plus bias, etc.) and nonlinear transformation on the output value of the previous layer of neurons .
  • a convolutional neural network may include one or more convolutional layers, one or more excitation layers, one or more pooling layers, one or more fully connected layers, and convolutional layers
  • the order of arrangement of layers, excitation layers, pooling layers, and fully connected layers is not limited.
  • Figure 9 shows a schematic diagram of the structure of a convolutional neural network.
  • the configuration parameters of the channel learning model involved in this step may be one or more of the number of layers, the sequence of the layers, and the structure of the channel learning model.
  • the first communication apparatus shown in the following embodiments may be replaced by a component (eg, a chip or a system of chips, etc.) configured in the first communication apparatus.
  • the second communication device shown in the following embodiments may also be replaced by a component (eg, a chip or a system of chips, etc.) configured in the second communication device.
  • the embodiments shown below do not specifically limit the specific structure of the execution body of the method provided by the embodiment of the present application, as long as the program that records the code of the method provided by the embodiment of the present application can be executed to execute the method provided by the embodiment of the present application.
  • the method can be used for communication.
  • the execution body of the method provided by the embodiment of the present application may be the first communication device or the second communication device, or, the first communication device or the second communication device can call the program and execute the program. functional module.
  • the first communication apparatus mentioned in the following embodiments may be a terminal device, or may be a component (such as a chip or a chip system, etc.) configured in the terminal device.
  • the second communication apparatus may be a network device, or may be a component (such as a chip or a chip system, etc.) configured in the network device.
  • the first communication apparatus may be a network device, or may be a component (such as a chip or a chip system, etc.) configured in the network device.
  • the second communication apparatus may be a terminal device, or may be a component (such as a chip or a chip system, etc.) configured in the terminal device.
  • the first communication apparatus may be a terminal device, or may be a component (such as a chip or a chip system, etc.) configured in the terminal device.
  • the second communication apparatus may be a terminal device, or may be a component (such as a chip or a chip system, etc.) configured in the terminal device.
  • the first communication apparatus may be a network device, or may be a component (such as a chip or a chip system, etc.) configured in the network device.
  • the second communication apparatus may be a network device, or may be a component (such as a chip or a chip system, etc.) configured in the network device.
  • first channel learning model mentioned in the following embodiments is deployed on the side of the first communication device
  • second channel learning model is deployed on the side of the second communication device.
  • the channel learning model mentioned in the following embodiments may refer to the first channel learning model and/or the second channel learning model if it is not clearly specified whether it is the first channel learning model or the second channel learning model.
  • the channel learning model training mentioned in the embodiments of the present application may also be referred to as channel learning training for short.
  • the channel learning model training may include at least one of the following: determining a first channel learning model, determining a second channel learning model, determining first channel information, determining second channel information, and the like.
  • one or more rows and one column or more columns in the table may be used in practical applications, for example, at least one row and at least one column.
  • FIG. 10 is a schematic flowchart of a communication method 200 provided by an embodiment of the present application, shown from the perspective of device interaction.
  • the method 200 shown in FIG. 10 may include S210 to S240. Each step in the method 200 is described in detail below.
  • the first communication apparatus determines whether the first channel learning model is applicable.
  • the first channel learning model is used to determine the first channel information based on the target channel information, and the data volume of the first channel information is smaller than that of the target channel information, so it can also be said that the first channel learning model is used to compress the target channel information to Obtain first channel information.
  • the data amount of the channel information may refer to the dimension of the channel information.
  • the number of antenna ports of the transmitting end (for example, it may be the first communication device or the second communication device) is A 2
  • the number of antenna ports of the receiving end (for example, it may be the first communication device or the second communication device) is A 3
  • the target channel information between the transmitting end and the receiving end may be a 3 -dimensional matrix of A 2 *A
  • the data amount of the target channel information may be represented by A 2 *A 3 . If the elements in the matrix of the target channel information are complex numbers, and the real part and the imaginary part of each element are expressed separately, the data amount of the target channel information can also be expressed as A 2 *A 3 *2.
  • the data amount of the first channel information may be represented by B 2 .
  • the data amount of the channel information may also refer to the amount of information included in the channel information, and the like.
  • the target channel information may be regarded as the input of the first channel learning model, and the first channel information may be regarded as the output of the first channel learning model.
  • the data volume of the target channel information may be the input information dimension, and the data volume of the first channel information may be the output information dimension.
  • the first channel information is used to obtain the second channel information through the second channel learning model, and the data amount of the second channel information and the target channel information is the same or similar.
  • the second channel information may be used for data transmission, for example, the second communication apparatus may determine scheduling information for data transmission, or determine precoding for data transmission, etc., according to the second channel information.
  • the first channel learning model corresponds to the second channel learning model. Therefore, determining whether the first channel learning model is applicable by the first communication device can be understood as the first communication device determining the first channel learning model and the second channel learning model. Whether the learning model is applicable. That is, when the first communication device determines that the first channel learning model is not applicable, it may determine that the second channel learning model is also not applicable; when the first communication device determines that the first channel learning model is applicable, it may determine that the second channel learning model is applicable. Channel learning models are also applicable.
  • the target channel information may be downlink channel information; for another example, when the first communication apparatus is a network device, the target channel information may be uplink channel information.
  • the target channel information may be uplink channel information, or the target channel information may be uplink channel information and downlink channel information, and the first communication apparatus may be based on the part of the uplink and downlink channels. For reciprocity, determine whether the first channel learning model and/or the second channel learning model is applicable according to the uplink channel information and the downlink channel information, or determine a new first channel learning model and/or the second channel learning model.
  • the target channel information may be downlink channel information, or the target channel information may be uplink channel information and downlink channel information, and the second communication device may be based on the part of the uplink and downlink channels. Reciprocity, determining whether the first channel learning model and/or the second channel learning model is applicable, or determining a new first channel learning model and/or the second channel learning model according to the uplink channel information and the downlink channel information.
  • the first communication device may periodically determine whether the first channel learning model is applicable. For example, the first communication device starts a timer after determining whether the first channel learning model is applicable for the i-th time. Further, in the case of the timer timeout, the first communication device determines whether the first channel learning model is suitable for the i+1-th time. Be applicable.
  • the overhead of interactive signaling between the first communication device and the second communication device can be reduced, and the first communication device can periodically determine whether the first channel learning model is applicable, which can avoid the situation that the first channel learning model is not applicable resulting in a decrease in communication performance.
  • the first communication apparatus may determine whether the first channel learning model is applicable in the case of receiving the first indication information.
  • the method 200 may further include S230: the second communication apparatus sends the first indication information; correspondingly, in S230, the first communication apparatus receives the first indication information.
  • the second communication device can instruct the first communication device to determine whether the first channel learning model is applicable through signaling, so that the second communication device can notify the first communication device in time when it finds that the first channel learning model is not applicable
  • the device performs verification, so that the delay in determining whether the first channel learning model is applicable can be reduced.
  • the first channel learning model can be updated in time to avoid communication performance degradation caused when the first channel learning model is not applicable.
  • the first indication information is used to instruct the first communication apparatus to perform channel learning model training, that is, it may instruct the first communication apparatus to determine whether the first channel learning model is applicable.
  • the first indication information can also be used to indicate one or more of the following:
  • the first message is used to indicate that the first channel learning model is not applicable, and the training parameters may include at least one of the following: time for channel learning model training, configuration information of reference signals for channel learning model training.
  • the resource used for transmitting the first message may be a resource on a physical uplink shared channel (physical uplink shared channel, PUSCH)/physical downlink shared channel (physical downlink shared channel, PDSCH), or may be a physical uplink control channel (physical uplink control channel) channel, PUCCH)/physical downlink control channel (physical downlink control channel, PDCCH), or may be other specific resources, for example, the resource used for transmitting the first message may be resource 1 or resource 2.
  • the content of the first message may include one or more of the following: a rank (rank) value, a channel quality index (CQI), first channel information, a result of channel learning model training (for example, indicating that the first channel learning whether the model is applicable and/or used to update one or more parameters of the second channel learning model).
  • rank a rank
  • CQI channel quality index
  • first channel information a result of channel learning model training (for example, indicating that the first channel learning whether the model is applicable and/or used to update one or more parameters of the second channel learning model).
  • the form of sending the first message may include one or more of the following: periodic sending, semi-persistent sending, aperiodic sending, differential value feedback, absolute value feedback, and relative value feedback.
  • Periodic sending the first communication device sends the first message periodically, for example, sending the first message with a period T.
  • Semi-persistent sending the first communication apparatus continues to send the first message within a period of time, for example, the first communication apparatus continues to send the first message within 10s after receiving the first indication information.
  • Aperiodic sending For example, when the first communication apparatus receives the first indication information, it sends the first message.
  • Training parameters of the channel learning model that is, the parameters used by the first communication device to train the first channel learning model and/or the second channel learning model.
  • the training time of the channel learning model may be periodic or aperiodic.
  • the training time of the channel learning model indicated by the first indication information may be a training period.
  • the cycle of the channel learning model training can be indicated by indicating the number of subframes, the number of time slots or the number of radio frames. For example, if the number of subframes indicated by the first indication information is 7, the cycle of the channel learning model training is 7 subframes. frame; or can indicate the period of channel learning model training by indicating milliseconds, seconds, minutes or hours, for example, if the time indicated by the first indication information is 5 seconds, the period of channel learning model training is 5 seconds.
  • the period of channel learning model training may be related to the first parameter. For example, in some scenarios (for example, high-speed motion), the channel environment between the first communication device and the second communication device changes rapidly, and the channel learning model can be trained in a smaller cycle; for example, in some scenarios (For example, in an indoor scene), if the channel environment between the first communication device and the second communication device changes slowly, the channel learning model can be trained with a larger period.
  • the training time of the channel learning model indicated by the first indication information may be a training period and the number of periods.
  • other parameters used for channel model training may be pre-configured by high-level signaling, and the first indication information may indicate the time of channel learning model training, so as to trigger channel learning model training.
  • the method 200 may further include S240: the first communication device sends the first request message; correspondingly, in S240, the second communication device receives the first request message.
  • the first request message is used to request one or more of the following: channel learning model training, sending a first message, and first indication information.
  • the first communication device can determine whether the first channel learning model is applicable through a signaling request, so as to request the second communication device for verification in time when the first communication device finds that the first channel learning model is not applicable , so as to reduce the delay in determining whether the first channel learning model is applicable.
  • the first channel learning model can be updated in time, so as to avoid communication performance degradation caused when the first channel learning model is not applicable.
  • This embodiment of the present application does not limit the method for the first communication apparatus to determine whether the first channel learning model is applicable.
  • the first communication apparatus may use one or more of the following implementation manners to determine whether the first channel learning model is applicable.
  • the first communication device and/or the second communication device may adjust the first channel learning model in time to improve the accuracy and applicability of the channel learning model, thereby improving the communication performance.
  • the following provides a way for the first communication device to determine whether the first channel learning model and/or the second channel learning model is applicable.
  • the way for the first communication device to determine whether the channel learning model is applicable can be used as an independent embodiment, and can also be implemented with other In combination with examples, specifically, the embodiments of the present application do not limit this.
  • One or more of the following ways of determining whether the channel learning model is applicable may be used alone or in combination, which is not specifically limited in this embodiment of the present application.
  • channel learning model mentioned in the embodiments of this application may also refer to whether the channel learning model matches, whether the channel learning model is accurate, whether the channel learning model is outdated, or whether the channel learning model is wrong, etc.
  • the first communication apparatus may determine whether the channel learning model is applicable according to the long-term statistical characteristics of the target channel. For example, when the first communication device determines that the long-term statistical characteristics of the target channel change greatly, it indicates that the channel characteristics or the channel environment between the first communication device and the second communication device have changed greatly, so the first communication device can determine The channel learning model does not apply.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the change amount of the long-term statistical characteristic of the target channel is greater than or equal to the first preset threshold.
  • the first preset threshold may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the long-term statistical characteristics of the target channel may include at least one of the following: rank value, large-scale characteristics, channel covariance matrix, channel correlation matrix, coherence time, coherence bandwidth, and the like.
  • the large-scale characteristics of the channel can be one or more of the following: delay spread, Doppler spread, Doppler shift, average channel gain, and average Delay (average delay), angle of arrival (angle of arrival, AOA), angle of arrival spread (AAS), angle of departure (angle of departure, AOD), angle of departure (angle of departure spread, ADS), spatial RX parameters, and spatial correlation.
  • the first communication device may determine the long-term statistical characteristics of the target channel according to the signal from the second communication device, and further determine whether the first channel learning model is applicable according to the long-term statistical characteristics of the target channel.
  • the signal from the second communication device may be a reference signal or a data signal
  • the reference signal may be a demodulation reference signal (DMRS), a channel state information reference signal (CSI-RS), Phase tracking reference signal (PTRS), tracking reference signal (TRS), synchronization signal and physical broadcast channel (physical broadcast channel) PBCH block (synchronization signal and PBCH channel, SSB), channel sounding reference Signal (sounding reference signal, SRS), etc.
  • the data signal may be a signal transmitted on PDSCH, PDCCH, PUSCH or PUCCH.
  • the terminal device may determine the long-term statistical characteristics of the target channel according to the reference signal and/or the data signal sent by the network device.
  • the reference signal may be at least one of CSI-RS, DMRS, PTRS, TRS, and SSB
  • the data signal may be at least one of a signal transmitted on PDSCH and a signal transmitted on PDCCH.
  • the target channel may refer to a downlink channel.
  • the first communication device when the variation of the long-term statistical characteristics of the target channel is greater than or equal to (or greater than) the first preset threshold, it indicates that the channel characteristics or the channel environment between the first communication device and the second communication device have occurred. There is a large change, so the first communication device can determine that the first channel learning model is not applicable; when the long-term statistical characteristics of the target channel are less than (or less than or equal to) the first preset threshold, it indicates that the first communication device and the first The channel characteristics or the channel environment between the two communication devices are relatively stable, so the first communication device can determine that the first channel learning model is applicable.
  • the first communication apparatus may determine that the first channel learning model is not applicable when it is determined that the rank value changes greatly. For example, the first communication apparatus may determine that the first channel learning model is not applicable when it is determined that the variation of the rank value is greater than or equal to the preset threshold #1 (an example of the first preset threshold).
  • the preset threshold #1 may be R 1 , where R 1 is a positive integer. For example, R 1 is 2, that is, when the change of the rank value is greater than or equal to 2, the first communication apparatus may determine that the first channel learning model is not applicable.
  • the path of the target channel increases, and the rank value changes.
  • the first channel learning model may no longer be applicable.
  • the larger the rank value the more complex the applicable channel learning model may be, for example, the higher the number of layers of the channel learning model may be.
  • the first communication device changes from indoor to outdoor, the statistical characteristics of the channel also change, and the first communication device may determine that the first channel learning model may not be applicable.
  • the first communication apparatus may determine that the first channel learning model is not applicable when it is determined that the Doppler frequency shift changes greatly.
  • Doppler shift may mean that when the mobile station moves in a certain direction at a constant rate, due to the difference in propagation distance, the phase and frequency will change, and this change is usually called Doppler shift.
  • Doppler shifts reveal how the properties of waves change in motion.
  • the first communication apparatus may determine that the first channel learning model is not applicable when the variation of the Doppler frequency shift is greater than or equal to the preset threshold #2 (an example of the first preset threshold).
  • the preset threshold #2 may be F 2 , where F 2 is a real number.
  • the first communication apparatus may determine that the first channel learning model is not applicable.
  • the Doppler frequency shift can reflect the moving speed of the first communication device. When the first communication device changes from walking to vehicle, and the moving speed of the first communication device increases, the first channel learning model may no longer be applicable.
  • the Doppler frequency shift is related to the moving speed of the first communication device, the moving direction of the first communication device and the included angle between the incident direction of the radio wave.
  • the formula for calculating the Doppler shift can be as follows:
  • v is the moving speed
  • ⁇ is the wavelength
  • ⁇ is the angle between the moving direction and the incident direction of the radio wave.
  • the first communication device may determine whether the first channel learning model is applicable according to the variation of the speed of its own motion. For example, the first communication device may determine that the first channel learning model is not applicable when the variation of the speed of movement is greater than the preset threshold #3 (an example of the first preset threshold).
  • the preset threshold #3 may be S 3 , where S 3 is a real number, and the unit of S 3 may be m/s, or km/h, that is, when the variation of the speed of the first communication device is greater than or equal to S 3 , the first communication apparatus may determine that the first channel learning model is not applicable. For example, when the first communication device changes from walking to being in a vehicle, and the change in speed is large, the first channel learning model may no longer be applicable.
  • the first communication device can determine whether the channel learning model is applicable according to the long-term statistical characteristics of the target channel, that is, determining whether the first channel learning model is applicable without performing channel learning model training, which can reduce the first communication
  • the processing complexity of the device and at the same time, it can determine whether the channel learning model is applicable without the assistance of the second communication device, and the signaling interaction is reduced, so this method is simple and fast.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the received first scheduling information.
  • the first scheduling information may be determined by the second communication apparatus according to the second channel information, and the second communication apparatus may send the first scheduling information to the first communication apparatus.
  • the first scheduling information may include at least one of the following: a modulation and coding scheme (modulation and coding scheme, MCS) indication, a transport block (transport block size, TBS) indication, a rank indication, an antenna port indication, and the like.
  • the first scheduling information may be downlink control information (downlink control information, DCI) of the physical layer, or may be scheduling information in higher-layer signaling.
  • DCI downlink control information
  • the MCS indication can be used to indicate the modulation mode (or modulation order) and code rate of data transmission.
  • the MCS indication can use the indication methods in the following tables (Table 1 and Table 2): By indicating the MCS index (MCS index) in the first column of the following table, the modulation order and target code rate in the second column can be determined (target code rate).
  • modulation order 2 represents quadrature phase shift keying (QPSK)
  • modulation order 4 represents 16 quadrature amplitude modulation (16QAM, quadrature amplitude modulation)
  • modulation order 6 represents 64QAM
  • modulation order 6 means 256QAM.
  • the MCS indication may also adopt other indication manners, which are not specifically limited in this application.
  • the modulation mode indication and the code rate indication may be indicated separately.
  • the transport block indication can be used to indicate the bit size of the data transmission.
  • the layer number indication can be used to indicate the number of layers or streams of data transmission.
  • the number of layers or streams can also correspond to the number of codewords. For example, if the number of layers or streams is less than or equal to 4, it corresponds to the transmission of one codeword; if the number of layers or streams is greater than 4, it corresponds to the transmission of two codewords, etc. .
  • the antenna port indicates the antenna port of the DMRS that can be used to indicate data.
  • the terminal device can determine the number of antenna ports according to the antenna port indication.
  • the antenna port indication can be indicated in the following table (Table 3 to Table 4):
  • the DMRS port can be determined by indicating the value of the first column of the following table, where the DMRS port includes the DMRS antenna port number.
  • the number of DMRS antenna ports can be determined according to the number of DMRS antenna port numbers.
  • the number of DMRS antenna ports may correspond to the number of layers (number of streams) of data.
  • the DMRS port is 0, 1, that is, the number of DMRS antenna ports is 2, and the number of data layers (the number of streams) is 2; for example, in the case of a codeword in Table 4, the value is When it is 10, the DMRS port is 0 to 3, that is, the number of DMRS antenna ports is 4, and the number of data layers (streams) is 4; for example, when the two code words in Table 4 are set to 1, the DMRS port is 0 , 1, 2, 3, 4, 6, that is, the number of DMRS antenna ports is 5, and the number of data layers (streams) is 5.
  • the antenna port indication can also be indicated in other tables, such as Table 7.3.1.2.2-3 to Table 7.3.1.2.2-4 in 3GPP technical specification (TS) 38.212, or
  • the antenna port indication may adopt the indication manner in other tables, which is not limited in this application.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the difference between the measured CQI information and the first scheduling information.
  • the CQI information measured by the first communication apparatus may reflect an appropriate modulation method and/or code rate of the data. Therefore, the first communication apparatus may determine whether the first channel learning model is applicable according to the difference between the measured CQI information and the first scheduling information.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the first scheduling information and the second scheduling information.
  • the second scheduling information is determined by the first communication apparatus according to the target channel.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the difference value between the first scheduling information and the second scheduling information is greater than or equal to a preset difference threshold.
  • the preset difference threshold may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the scheduling information of the above data can reflect the quality of the channel. Therefore, the first communication apparatus can determine the appropriate scheduling information (ie, the second scheduling information) of the data according to the measured target channel (or actual channel).
  • the second communication device may determine the second channel information according to the second channel learning model and the first channel information fed back by the first communication device, and may determine data scheduling information (ie, the first scheduling information) according to the second channel information.
  • the difference between the second scheduling information determined by the first communication apparatus and the first scheduling information indicated to the first communication apparatus by the second communication apparatus is large, it indicates that the difference between the second channel information and the target channel information is large, That is, the first channel learning model is not applicable. Therefore, the first communication apparatus may determine whether the first channel learning model is applicable according to at least one of the following manners.
  • the first communication device determines that the first channel learning model is not applicable when the difference between the first scheduling information and the second scheduling information is greater than or equal to a preset difference threshold; the first communication device determines that the first channel learning model is not applicable; When the difference value of the two scheduling information is smaller than the preset difference threshold, it is determined that the first channel learning model is applicable.
  • the preset difference threshold may be at least one of the difference threshold of modulation order, the difference threshold of code rate, the difference threshold of MCS index (index) value, the difference threshold of rank value, the difference threshold of TBS, and the difference threshold of antenna port. .
  • the difference threshold of the modulation order can be N 3 orders, where N 3 is a real number, for example, N 3 is 1, 2, 3, 4, 1/2, 3/ 2, 5/2 etc.
  • the first communication apparatus determines The first channel learning model does not apply.
  • the first communication device determines that the first channel learning model is applicable.
  • the modulation method determined by the first communication device and suitable for data transmission is quadrature phase shift keying (QPSK) (the order is 2).
  • QPSK quadrature phase shift keying
  • the modulation mode indicated by the scheduling information is 64-phase quadrature amplitude modulation (QAM) (the order is 4)
  • the first communication device determines that the first channel learning model is not applicable.
  • the first communication device determines that the first channel learning model is applicable.
  • the first communication apparatus determines that the first channel learning model is not applicable.
  • the first communication apparatus may determine the first channel learning model when the determined modulation order is lower than the modulation order order indicated by the first scheduling information and the difference value is greater than or equal to the modulation order difference threshold Not applicable. That is to say, the difference value between the order of the determined modulation scheme and the order of the modulation scheme indicated by the first scheduling information is greater than or equal to (greater than) the difference threshold of the modulation order, and the When the order is lower than the order of the modulation scheme indicated by the first scheduling information, it is determined that the first channel learning model is not applicable.
  • the first communication apparatus may determine that the first channel learning model is applicable when the determined order of the modulation scheme is higher than or equal to the order of the modulation scheme indicated by the first scheduling information.
  • the first communication device determines that the first channel learning model is applicable.
  • the difference threshold of the modulation order is 2
  • the modulation method determined by the first communication device suitable for data transmission is QPSK (order 2)
  • the modulation indicated in the first scheduling information received by the first communication device is QPSK (order 2).
  • the mode is 64QAM (order 4)
  • the first communication apparatus may determine that the channel learning model is not applicable.
  • the modulation order of the data transmitted by the second communication device to the first communication device based on the second channel information determined according to the channel learning model is relatively high, and the first communication device may not be able to receive the data correctly.
  • the code rate difference threshold may be M 1 , where M 1 is a real number, for example, M 1 is 200/1024, 250/1024, 300/1024, 500/1024, and so on.
  • the first communication device determines that the first channel learning model is not applicable .
  • the first communication device determines that the first channel learning model is applicable. .
  • the first communication device determines that the code rate suitable for data transmission is 400/1024, and the code rate indicated in the first scheduling information is 658/1024, then the first communication The apparatus may determine that the first channel learning model is not applicable.
  • the first communication apparatus may determine that the first channel learning model is not applicable when the determined code rate is lower than the code rate difference value indicated by the first scheduling information and is greater than or equal to the code rate difference threshold.
  • the difference value between the determined code rate and the code rate indicated by the first scheduling information is greater than or equal to (greater than) the code rate difference threshold, and the determined code rate value is lower than the code rate indicated by the first scheduling information.
  • the code rate it is determined that the first channel learning model is not applicable.
  • the first communication apparatus may determine that the first channel learning model is applicable when the determined code rate is higher than or equal to the code rate indicated by the first scheduling information.
  • the first communication apparatus may determine that the channel learning model is not applicable. In this case, the code rate of the data transmitted by the second communication device to the first communication device based on the second channel information determined according to the first channel learning model is relatively high, and the first communication device may not be able to receive the data correctly.
  • the difference threshold of MCS index may be P 1 , and P 1 is an integer, for example, P 1 is 1, 2, 3, 4, 5, 6, 8, 10 and so on.
  • the first communication device determines that the first channel learning model is not applicable .
  • the first communication device determines that the first channel learning model is applicable. For example, if the difference threshold of MCS index values is 4, the MCS index value suitable for data transmission determined by the first communication device is 4, and the MCS index value indicated in the first scheduling information is 10, then the first communication device can determine The first channel learning model does not apply.
  • the first communication apparatus may determine that the first channel learning model is not applicable when the determined MCS index is lower than the MCS index indicated by the first scheduling information and the difference value is greater than or equal to the MCS index difference threshold.
  • the difference between the determined MCS index of the first communication device and the MCS index indicated by the first scheduling information is greater than or equal to (greater than) the difference threshold of the MCS index, and the exact value of the MCS index is lower than the first scheduling information.
  • the first communication apparatus determines that the first channel learning model is not applicable.
  • the first communication apparatus may determine that the first channel learning model is applicable when the determined MCS index is higher than or equal to the MCS index indicated by the first scheduling information.
  • the first communication device may determine that the channel learning model is not applicable. In this case, the MCS index of the data transmitted by the second communication device to the first communication device based on the second channel information is relatively high, and the first communication device may not be able to receive the data correctly.
  • the difference threshold of TBS may be Q, where Q is an integer, for example, Q is 32, 64, 128, 256, 612, 1024 and so on.
  • the first communication apparatus determines that the first channel learning model is not applicable. For another example, when the difference between the TBS determined by the first communication apparatus and the TBS indicated by the first scheduling information is less than (less than or equal to) the difference threshold of TBS, the first communication apparatus determines that the first channel learning model is applicable.
  • the first communication apparatus may determine that the first channel learning model does not Be applicable.
  • the first communication apparatus may determine that the first channel learning model is not applicable when the determined TBS is lower than the TBS indicated by the first scheduling information and the difference value is greater than or equal to the TBS difference threshold.
  • the first communication apparatus determines that the first channel learning model is not applicable.
  • the first communication apparatus may determine that the first channel learning model is applicable when the determined TBS is higher than or equal to the TBS indicated by the first scheduling information.
  • the first communication device can It is determined that the first channel learning model is not applicable. In this case, the TBS of the data transmitted by the second communication device to the first communication device based on the second channel information determined according to the first channel learning model is too large, and the first communication device cannot correctly receive the data.
  • the first communication device determining whether the first channel learning model is applicable according to the difference between the first scheduling information and the second scheduling information as an example.
  • the first communication device Whether the first channel learning model is applicable may also be determined according to the difference value of the multiple items in the first scheduling information and the second scheduling information.
  • the first communication device may be greater than or equal to a code rate difference threshold when the difference between the determined code rate and the code rate indicated by the first scheduling information is greater than or equal to the code rate difference threshold, and the difference between the determined TBS and the TBS indicated by the first scheduling information is greater than or equal to When it is equal to the TBS difference threshold, it is determined that the first channel learning model is not applicable.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the difference between the first scheduling information and the second scheduling information is greater than or equal to a preset difference threshold.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the similarity between the first scheduling information and the second scheduling information is less than a preset similarity threshold.
  • the preset similarity threshold value may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the first scheduling information, the second scheduling information, and the first mapping relationship, and the first mapping relationship is used to indicate whether the scheduling information is applicable to the channel learning model. relation.
  • Table 5 shows an example of the first mapping relationship.
  • the first mapping relationship may be one or more lines in the following table.
  • the first mapping relationship may be predefined by a protocol, or may be notified by the second communication device to the first communication device through signaling. Specifically, this application implements The example does not limit this.
  • the first communication device determines that the first channel learning model is applicable according to Table 5; for another example, if the first communication device determines The code rate is 2/3, and the code rate indicated by the first scheduling information is 3/4, then the first communication apparatus may determine, according to Table 5, that the first channel learning model is not applicable.
  • the first communication device may determine whether the channel learning model is applicable according to the first scheduling information sent by the second communication device, that is, if the channel learning model training is not performed, determine whether the first channel learning model is applicable.
  • the processing complexity of the first communication device is reduced.
  • whether the first channel learning model is applicable is determined according to the scheduling information determined by the first communication device and the second communication device, and the communication performance based on the first channel learning model can also be guaranteed.
  • the method of determining whether the channel learning model is applicable is relatively simple and fast.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the data transmission performance.
  • the performance of data transmission can reflect the similarity between the second channel information and the target channel information, and can further reflect the performance of the first channel learning model and/or the second channel learning model, that is, the first channel learning model and/or the second channel learning model. Whether the channel learning model is applicable.
  • the data transmission performance may include the transmission performance of the first data and/or the transmission performance of the second data.
  • the first data is sent by the first communication device according to the target channel information
  • the second data is sent by the second communication device according to the second channel information.
  • the first communication device is a terminal device and the second communication device is a network device
  • the first data may be uplink data
  • the second data may be downlink data.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the performance of correctness of data transmission, or the performance of acknowledgement (ACK)/negative acknowledgement (NACK) of data transmission.
  • ACK acknowledgement
  • NACK negative acknowledgement
  • the performance of the correctness of data transmission can be measured by the correctness rate of data transmission within a period of time.
  • the performance of ACK/NACK for data transmission can be measured by the ratio of ACK/NACK received or sent over a period of time.
  • the unit of the time length of a period of time may be milliseconds (ms), seconds (s), minutes (min), hours (h), days, months, and the like.
  • the unit of the time length of a period of time may also be a time slot (slot), a subframe (subframe), a radio frame (frame), a transmission time interval (transmission time interval, TTI), and the like.
  • a period of time may refer to T units of length of time, and T may be a real number.
  • a period of time may refer to time lengths such as 10ms, 20ms, 0.5s, 1s, 10s, etc., or may refer to 10 time slots, 20 subframes, 10 radio frames, and so on.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the relationship between the correct rate of data transmission within a period of time, or the ratio of received or sent ACKs and the preset threshold #4.
  • the preset threshold #4 may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the first communication apparatus determines that the correct rate of data transmission within a period of time, or the ratio of receiving or sending ACKs is less than (or less than or equal to) the preset threshold #4, and determines that the first channel learning model is not applicable. For another example, the first communication apparatus determines that the first channel learning model is applicable when the correct rate of data transmission within a period of time, or the ratio of receiving or sending ACKs is greater than or equal to (or greater than) the preset threshold #4.
  • the first communication device determines the correct rate of data transmission within 10s, or if the ratio of receiving or sending ACK is less than 90%, determines the first communication device.
  • the channel learning model does not apply.
  • the first communication apparatus determines whether the first channel learning model is applicable according to the number of data transmission failures within a period of time, or the relationship between the number of received or sent NACKs and the preset threshold #5.
  • the preset threshold #5 may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the first communication device determines that the first channel learning model is not applicable when the number of data transmission failures within a period of time, or the number of received or sent NACKs is greater than (or greater than or equal to) a preset threshold #5. For another example, the first communication device determines that the first channel learning model is applicable when the number of data transmission failures within a period of time, or the number of NACKs received or sent is less than or equal to (or less than) a preset threshold #5.
  • the first communication device determines the number of failed data transmissions within 10s, or if the number of NACKs received or sent is greater than 5, the first communication device determines The channel learning model does not apply.
  • the performance of data transmission may also refer to at least one of performances such as throughput, throughput rate, and spectral efficiency.
  • the first communication apparatus may determine that the first channel learning model is not applicable when the performance of data transmission is less than (or, less than or equal to) the preset threshold #6.
  • the preset threshold #6 may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the first communication device can determine whether the first channel learning model is applicable according to the data transmission performance, that is, it can determine whether the first channel learning model is applicable without performing channel learning model training, which can reduce the first communication The processing complexity of the device.
  • the way to determine whether the first channel learning model is applicable according to the data transmission performance is to measure the accuracy of the first channel learning model based on the final communication performance, so the communication performance based on the first channel learning model can also be guaranteed, that is, This method helps to improve communication performance.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the scene in which it is located changes.
  • the scene may be at least one of the following: indoor stillness, outdoor stillness, low-speed motion, high-speed motion, suburban, town, macro station, micro station, vehicle scene, V2X scene, scene defined by 3GPP protocol, etc.
  • the channel of the first communication device and the second communication device is a direct path; and in some scenarios, the channel of the first communication device and the second communication device is a non-direct path.
  • there are few reflectors between the first communication device and the second communication device, and correspondingly, the channel between the first communication device and the second communication device is simple; There are many reflectors between the communication device and the second communication device, and correspondingly, the channel between the first communication device and the second communication device is complicated.
  • the first channel learning model may not be applicable. For example, if the first communication device moves from indoor to outdoor, the first communication device may determine that the first channel learning model is not applicable; for another example, if the first communication device moves from a macro station to a micro station, the first communication device may determine The first channel learning model does not apply.
  • the first communication device can determine whether the first channel learning model is applicable according to the scene in which it is located, that is, determining whether the first channel learning model is applicable without performing channel learning model training, which can reduce the first communication The processing complexity of the device.
  • the method of determining whether the first channel learning model is applicable according to the scene in which it is located may not depend on the assistance of the second communication device, and the first communication device can independently determine whether the first channel learning model is applicable, so this method is simple and fast.
  • the first communication device when the first communication device is a network device and the second communication device is a terminal device, the first communication device may determine whether the first channel learning model is applicable according to whether the scene in which the second communication device is located changes. .
  • the first communication apparatus may also receive scene information sent by the second communication apparatus.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the performance index of the first channel learning model.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the difference or similarity between the target channel information and the first channel information. For example, the first communication apparatus may compress the target channel information according to the first channel learning model to obtain the first channel information. By comparing the characteristics of the target channel information and the first channel information, the error of the first channel learning model is further determined, thereby determining whether the first channel learning model is applicable.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the performance index of the first channel learning model is smaller than the second preset threshold.
  • the second preset threshold may be predefined by a protocol, may be indicated by the second communication apparatus to the first communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the performance indicator of the first channel learning model is continuity (CT).
  • CT continuity
  • CT can be expressed by formula (1):
  • k represents the data points u i, k for the u i is set
  • K 1 represents a set of data points most neighboring points u i is set in the mapping space (the set point, but not necessarily in correspondence of u i v i K space mapping a set of nearest neighbors )
  • Data representing points u i, k of the mapping point v i, v i in the set of k for , where L is the number of data points.
  • the value range of CT is 0 to 1. If the value of CT is relatively small, it indicates that the adjacent points in the target channel information (high-dimensional data points) are not adjacent to the first channel information (low-dimensional data points). . It can also be said that in this case, the first channel information determined according to the first channel learning model does not retain the effective features of the target channel information, so the performance of the first channel learning model is not good, that is, the first channel learning model is not applicable. . If the value of CT is relatively large (close to 1), it indicates that after the data points in the target channel information are mapped to the first channel information, adjacent features are maintained. It can also be said that in this case, the first channel information determined according to the first channel learning model retains the effective features of the target channel information, so the performance of the first channel learning model is better, that is, the first channel learning model is suitable.
  • the first communication device may determine that the first channel learning model is applicable when the value of CT is determined to be greater than or equal to the preset threshold #7; and when it is determined that the value of CT is less than the preset threshold #7, determine the first The channel learning model does not apply.
  • the performance metric of the first channel learning model is trustworthiness (TW).
  • TW trustworthiness
  • TW represents the set of u i The u i, k of the mapping point v i, k if v i is also set middle. TW can be expressed by formula (2):
  • k represents the data points u i, k of the mapping point v i, v i in the set of k for The ordering in is k, Indicates the wrong neighboring points, that is, the set of these points in v i In, i is set but not u In, r (i, k) indicates mapping point v i, k in the set of data points corresponding to u i, u i for the k , where L is the number of data points.
  • TW ranges from 0 to 1. If the value of TW is relatively small, it indicates that the adjacent points in the first channel information (low-dimensional data points) are not adjacent in the target channel information (high-dimensional data points). It can also be said that in this case, the first channel information determined according to the first channel learning model does not retain the effective features of the target channel information, so the performance of the first channel learning model is not good, that is, the first channel learning model is not applicable. . If the value of TW is relatively large (close to 1), it indicates that the data points in the first channel information are adjacent in the target channel information. It can also be said that in this case, the first channel information determined according to the first channel learning model retains the effective features of the target channel information, so the performance of the first channel learning model is better, that is, the first channel learning model is suitable.
  • the first communication device may determine that the first channel learning model is applicable when the value of TW is determined to be greater than or equal to the preset threshold #8; and when it is determined that the value of TW is less than the preset threshold #8, determine that the first The channel learning model does not apply.
  • the first communication device may determine whether the first channel learning model is applicable according to the performance index of the first channel learning model, that is, whether the first channel learning model is applicable by performing channel learning model training. This method does not depend on the assistance of the second communication device, and the first communication device can determine whether the channel learning model is applicable, so this method is simple and fast.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to an error between the target channel information and the second channel information.
  • the first communication device determines the second channel information according to the first channel information and the second channel learning model, and further, according to the target channel information and the first channel information The error of the two channel information determines whether the first channel learning model is applicable.
  • both the first channel learning model and the second channel learning model are deployed on the side of the first communication device.
  • the first communication device may predetermine configuration parameters of the first channel learning model and configuration parameters of the second channel learning model; further, determine the first channel learning model and the second channel learning model.
  • the second communication device may predetermine the configuration parameters of the first channel learning model and the configuration parameters of the second channel learning model; further, send the configuration parameters of the first channel learning model and the configuration parameters of the second channel learning model to the first communication device; correspondingly, the first communication device determines the first channel learning model and the second channel learning model according to the configuration parameters of the first channel learning model and the configuration parameters of the second channel learning model.
  • the solution for the first communication device to determine whether the first channel learning model is applicable may also be applied to the first communication device to determine whether the first channel learning model and the second channel learning model are applicable, or, It can be applied to the first communication device to determine whether the second channel learning model is applicable. Specifically, this application does not limit this.
  • the first communication device may determine that the first channel learning model is not applicable when the error between the target channel information and the second channel information is greater than or equal to the preset threshold #9; when the error between the target channel information and the second channel information is less than the predetermined threshold #9; When threshold #9 is set, it is determined that the first channel learning model is applicable.
  • This embodiment of the present application does not limit the calculation method of the error between the target channel information and the second channel information.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to a mean square error (mean square error, MSE) between the target channel information and the second channel information.
  • MSE mean square error
  • the t-th channel measurements obtained in the first transmit port and a 2 a 3 a first data channel between the receiving port is expressed as (an example of target channel information)
  • the channel data recovered by the first communication device according to the second channel learning model is expressed as (an example of the second channel information)
  • the error between the target channel information and the second channel information can be expressed as:
  • a 2 indicates the number of sending ports, 0 ⁇ a 2 ⁇ A 2 -1;
  • a 3 indicates the number of receiving ports, 0 ⁇ a 3 ⁇ A 3 -1;
  • T indicates the number of channel measurements, that is, the number of training samples, 0 ⁇ t ⁇ T-1.
  • the MSE between the target channel information and the second channel information can also be calculated by the following formula:
  • the first communication apparatus may determine whether the first channel learning model is applicable according to a normalized mean square error (NMSE) between the target channel information and the second channel information.
  • NMSE normalized mean square error
  • the NMSE between the target channel information and the second channel information can be calculated by the following formula:
  • F represents the F norm.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to a normalized mean correlation error (NMCE) between the target channel information and the second channel information.
  • NMCE normalized mean correlation error
  • the NMCE between the target channel information and the second channel information can be calculated by the following formula:
  • the correlation matrix H of the real downlink channel is used H H eigenvectors of primary and secondary (or right singular vector of H) as an estimate of the downlink channel, a very good performance can be achieved.
  • the core point of the channel learning model lies in the characterization of the eigenvectors of the downlink channel correlation matrix. Based on this, the following two loss function forms are proposed, and the loss function can be used as an index to measure the first channel learning model.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to a normalized mean correlation singular error (NMCSE) between the target channel information and the second channel information.
  • NMCSE normalized mean correlation singular error
  • NMSCE can be expressed by the following formula:
  • the primary and secondary feature vectors of the target channel are used as labels (denoted as Heig ); at this time, the following loss function form can be used:
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the correspondence between one or more of the above-mentioned metrics of the channel learning model and the error threshold. For example, when the performance of the metric is greater than or equal to a certain threshold, the first communication device determines that the first channel learning model is not applicable; or, when the performance of the metric is less than a certain threshold, the first communication device determines that the first channel learning model is not applicable Model applies.
  • the first communication device can determine whether the first channel learning model is applicable according to the error between the target channel information and the second channel information, that is, whether the first channel learning model is applicable is determined by performing channel learning model training. With the assistance of the second communication device, the first communication device can determine whether the channel learning model is applicable, so this method is simple and fast. Determining whether the first channel learning model is applicable while considering the characteristics of the target channel information and the second channel information at the same time can ensure that the first communication device and the second communication device obtain the same or similar characteristics of the channel information, which is helpful for subsequent The performance of data transmission is improved when the channel information is used for data transmission.
  • the first communication apparatus sends a first message, where the first message is used to indicate whether the first channel learning model is applicable.
  • the second communication apparatus receives the first message, and determines whether the first channel learning model is applicable according to the first message.
  • the first communication device when the first communication device determines that the first channel learning model is not applicable, the first communication device sends a first message, where the first message is used to indicate that the first channel learning model is not applicable.
  • the second communication device when the second communication device receives the first message, when the first message is used to indicate that the first channel learning model is not applicable, the second communication device may determine that the first channel learning model is not applicable according to the first message.
  • the first channel learning model corresponds to the second channel learning model
  • the second communication device determines that the first channel learning model is not applicable according to the first message, that is, it may also determine that the second channel learning model is not applicable.
  • the first communication apparatus may send the first message when it is determined that the first channel learning model is not applicable.
  • This embodiment of the present application does not limit the manner in which the first message indicates that the first channel learning model is not applicable.
  • the first message may be a bool variable. For example, if the first message is 0, it means that the first channel learning model is not applicable; if the first message is 1, it means that the first channel learning model is applicable. For another example, if the first message is 0, it indicates that the first channel learning model is applicable; if the first message is 1, it indicates that the first channel learning model is not applicable.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the value of the first message.
  • the first message may be channel learning feedback signaling, and the first message may include channel information.
  • the first communication apparatus sends the first message to the second communication apparatus, and accordingly, the second communication apparatus may determine whether the first channel learning model is applicable according to the channel information in the first message.
  • the channel information may include at least one of the following: a rank value, a CQI value, and a CRI value.
  • the first message may include a rank value, and if the rank value is 0, it indicates that the first channel learning model is not applicable.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the rank value in the first message.
  • the number of bits in the rank field may be determined according to the maximum number of layers supported by the first communication device and the number of antenna ports. For example, the number of bits in the field of rank is log 2 (min(number of layers, number of antenna ports)) rounded up. For example, log 2 (R) is rounded up.
  • the rank value may be 1 to 4
  • the rank value may be indicated by 2 bits.
  • the rank value may be 1 to 4
  • the rank value may be indicated by 2 bits.
  • the rank value can be 0 to R, compared with the existing rank value, there are more cases where the rank value is 0, and the rank value is calculated according to the existing method of calculating the number of bits of the rank domain. There are not enough bits to indicate different rank values.
  • the number of bits in the field of rank may be log 2 (min (number of layers, number of antenna ports)+1) rounded up, for example, log 2 (R+1) rounded up.
  • the rank value may be 0 to 4, and the rank value may be indicated by 3 bits.
  • a rank value of 0 indicates that the first channel learning model is not applicable.
  • the first message may include a channel quality index (channel quality index, CQI) value, and if the CQI value is 0, it indicates that the first channel learning model is not applicable.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the CQI value in the first message.
  • the first message may include a channel state information reference signal resource index (channel state information reference signal resource index, CRI) value, and when the CRI value is 0, it indicates that the first channel learning model is not applicable.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the CRI value in the first message.
  • the CRI value fed back by the first communication device corresponds to the number of resources used for sending reference signals.
  • the number of bits in the CQI field may be determined according to the number of configured reference signal resources. For example, the number of bits of the field of the CRI is log 2 (C) rounded up. For example, if the number of configured reference signal resources is 2, the CRI value may be 1-2, and then 1 bit may be used to indicate the CRI value.
  • the CRI value can be 0 to C, compared with the existing CRI value, there are more cases where the CRI value is 0, and the CRI value is calculated according to the existing method for calculating the number of bits of the CRI field. There are not enough bits to indicate different CRI values.
  • the number of bits of the field of the CRI may be log 2 (C+1) rounded up. For example, if the number of configured reference signal resources is 2, the CRI value may be 0-2, and the CRI value may be indicated by 2 bits. For example, "00" indicates a CRI value of 0, "01” indicates a CRI value of 1, and "10" indicates a CRI value of 2. Wherein, a CRI value of 0 indicates that the first channel learning model is not applicable.
  • the first message may include the variation of the long-term statistical characteristics of the target channel.
  • the amount of change indicated by the first message is greater than or equal to the first preset threshold, it indicates that the first channel learning model is not applicable.
  • the first communication apparatus may report the variation of the long-term statistical characteristics of the target channel in the form of differential reporting.
  • the first communication device may report the difference value in the difference report in the first message; correspondingly, the second communication device may determine the offset between the reported change and the reference variable according to the difference value and the difference report value, further, the second communication device may determine the reported change amount according to the offset value and the reference variable; further, the second communication device may determine the first The channel learning model does not apply.
  • Table 6 shows an example of the difference report.
  • the difference value may be one or more lines in the following table, and the corresponding relationship in the following table may be predefined by the protocol, or it may be notified by the second communication device to the first communication device through signaling. Specifically, this application implements The example does not limit this.
  • the first communication device may report the absolute value of the change amount of the long-term statistical characteristic of the target channel in the first message; correspondingly, the second communication device may determine the reported change amount according to the reported absolute value; Further, the second communication apparatus may determine that the first channel learning model is not applicable under the condition that the amount of change is determined to be greater than or equal to the first preset threshold.
  • the first communication apparatus may report the variation of the long-term statistical characteristics of the target channel in the form of relative value reporting.
  • the first communication device may report the relative value in the report on the relative value in the first message; correspondingly, the second communication device may determine the difference between the reported change and the reference variable according to the relative value and the report on the relative value
  • the offset further, the second communication device may determine the reported change according to the offset and the reference variable; further, the second communication device may determine the change when it is determined that the change is greater than or equal to the first preset threshold
  • Table 7 shows an example of a report on relative values.
  • the relative value may be one or more lines in the following table, and the corresponding relationship in the following table may be predefined by the protocol, or may be notified by the second communication device to the first communication device through signaling.
  • this application implements The example does not limit this.
  • the first communication apparatus indicates that the first channel learning model is not applicable by sending the first message on the specific resource.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the resource for receiving the first message.
  • the first communication device sends the first message on the physical uplink shared channel (PUSCH)/physical downlink shared channel (PDSCH), it means that the first channel learning model is not applicable; the first If the communication device sends the first message on the physical uplink control channel (physical uplink control channel, PUCCH)/physical downlink control channel (physical downlink control channel, PDCCH), it means that the first channel learning model is applicable. For another example, if the first communication device sends the first message on PUSCH/PDSCH, it indicates that the first channel learning model is applicable; if the first communication device sends the first message on PUCCH/PDCCH, it indicates that the first channel learning model is not applicable.
  • PUSCH physical uplink shared channel
  • PDSCH physical downlink shared channel
  • the first communication device sends the first message on resource 1, it indicates that the first channel learning model is applicable; if the first communication device sends the first message on resource 2, it indicates that the first channel learning model is not applicable.
  • the size of the resources occupied by the first communication device for sending the first message is X resource units, indicating that the first channel learning model is applicable; the size of the resources occupied by the first communication device for sending the first message is Y resource units, indicating that The first channel learning model does not apply.
  • the resource unit may refer to a resource element (resource element, RE), a symbol (symbol), or a resource block (resource block, RB).
  • X is a positive integer, or X may also refer to a certain range, for example, X is X 1 to X 2 , or greater than X 1 , or less than X 2 , etc.
  • X 1 and X 2 are positive integers.
  • Y is a positive integer, or Y can also refer to a certain range, for example, Y is Y 1 to Y 2 , or greater than Y 1 , or less than Y 2 , etc. Y 1 and Y 2 are positive integers.
  • Table 8 shows an example of the correspondence between different resources used for sending the first message and whether the first channel learning model is applicable.
  • the corresponding relationship may be one or more lines in the following table, and the corresponding relationship in the following table may be predefined by the protocol, or may be notified by the second communication device to the first communication device through signaling.
  • this application implements The example does not limit this.
  • the first communication apparatus indicates that the first channel learning model is not applicable through the number of bits of the first message.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the number of bits of the first message.
  • Z is a positive integer, or Z can also refer to a certain range, for example, Z is Z 1 to Z 2 , or Z is greater than Z 1 , or Z is less than Z 2 , etc.
  • Z 1 and Z 2 are positive integers.
  • W 3 is a positive integer, or W can also refer to a certain range, for example, W is W 3 , or W is greater than W 3,1 , or W is less than W 3,2, etc., W 3,1 , W 3,2 is a positive integer.
  • the number of bits of the first message sent by the first communication device is 20 bits (bits), it indicates that the first channel learning model is not applicable; the number of bits of the first message sent by the first communication device is 10 bits, indicating that the first The channel learning model applies.
  • the number of bits of the first message sent by the first communication device is greater than 10 bits, it indicates that the first channel learning model is not applicable; the number of bits of the first message sent by the first communication device is less than 10 bits, indicating that the first channel learning model is not applicable. Be applicable.
  • Table 9 shows an example of the correspondence between the number of bits of the first message and whether the first channel learning model is applicable.
  • the corresponding relationship may be one or more lines in the following table, and the corresponding relationship in the following table may be predefined by the protocol, or may be notified by the second communication device to the first communication device through signaling. This is not limited.
  • W', W'', and W''' are similar to those of W 3
  • Z', Z'', and Z''' are similar to those of Z.
  • the embodiments of the present application will not repeat them.
  • the first communication apparatus may further determine the resource for sending the first message according to the number of bits of the first message.
  • the second communication apparatus may determine the resource of the first message according to the number of bits of the first message.
  • the first communication device determines that the resource identifier occupied by sending the first message is resource 1; for another example, the number of bits of the first message is W 3 , then the first communication device determines to send the first message.
  • the resource occupied by a message is identified as resource 1.
  • the first communication device determines that the resource identifier occupied by sending the first message is resource 1; the number of bits of the first message is 10 bits, then the first communication device determines that the resource occupied by sending the first message is identified as resource 1.
  • the resource is identified as resource 2.
  • the first communication device determines that the resource identifier occupied by sending the first message is resource 1; the number of bits of the first message is less than 10 bits, then the first communication device determines to send the first message.
  • the resource occupied by a message is identified as resource 2.
  • the first communication device determines that the size of the resources occupied by sending the first message is X resource units; the number of bits of the first message is W 3 , then the first communication device determines to send the first message.
  • the resource size occupied by a message is Y resource units.
  • Table 10 shows an example of the correspondence between the number of bits of the first message and the resource for transmitting the first message.
  • the corresponding relationship may be one or more lines in the following table, and the corresponding relationship in the following table may be predefined by the protocol, or may be notified by the second communication device to the first communication device through signaling. This is not limited.
  • the first communication apparatus indicates through the content of the first message that the first channel learning model is not applicable.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the content of the first message.
  • Table 11 shows an example of the correspondence between the content of the first message and whether the first channel learning model is applicable.
  • the corresponding relationship may be one or more lines in the following table, and the corresponding relationship in the following table may be predefined by the protocol, or may be notified by the second communication device to the first communication device through signaling. This is not limited.
  • the first message includes rank value, CQI and first channel information, it means that the first channel learning model is applicable; if the first message includes rank value and CQI, it means that the first channel learning model does not Applicable; if the first message includes CQI and first channel information, it indicates that the first channel learning model is applicable; if the first message includes CQI, it indicates that the first channel learning model is not applicable; if the first message includes the first channel learning model channel information, it means that the first channel learning model is applicable; if the first message includes a rank value, it means that the first channel learning model is not applicable; if the first message includes CQI and first channel information, it means that the first channel learning model is The model is applicable. If the first message includes a rank value, it means that the first channel learning model is not applicable.
  • the first communication apparatus may feed back the first channel information in the form of differential reporting.
  • the first channel learning model since the value range of the first channel information determined by the first communication device according to the first channel learning model and the target channel information is relatively stable, the currently determined first channel information can be fed back. The difference between the value of the channel information and the value of the first channel information determined previously, so as to reduce the feedback overhead.
  • the first message further includes one or more configuration parameters for updating the second channel learning model.
  • the second communication apparatus may update the second channel learning model according to the first message.
  • the one or more configuration parameters may include at least one of the following: model type, model structure, model algorithm, model weight vector, model weight matrix, model bias vector, model bias matrix, model activation function.
  • the first message may also include one or more configuration parameters for updating the first channel learning model.
  • one or more configuration parameters for updating the first channel learning model may also be sent through other messages, and the other messages are different from the first message, which is not specifically limited in this embodiment of the present application.
  • the types of models include machine learning algorithms, neural network models or auto-encoding models.
  • the configuration parameter may include a model algorithm for indicating which one of the machine learning algorithms the channel learning model is specifically.
  • the configuration parameters may include a model structure, which is used to indicate which type of neural network the channel learning model is, and the model structure may also include the dimension of the input layer and the dimension of the output layer.
  • the configuration parameters may include one or more of a transformation algorithm, a weight matrix, a weight vector, a bias vector, a bias matrix, and an activation function.
  • the configuration parameter may include a model structure for indicating the number of layers and/or the sequence of layers.
  • the configuration parameters can also include one or more of the following: parameters for the data input layer (preprocessing operation algorithm, dimension of input data, value range of input data), parameters for the convolution layer (size of input unit, Receptive field, stride, number of zero padding, depth, depth of output unit, weight matrix), parameters for excitation layer (activation function), parameters for pooling layer (pooling algorithm, spatial range, stride, The size of the input unit, the size of the output unit), the parameters for the fully connected layer (weight matrix, weight vector, bias matrix, bias vector).
  • the first communication device may determine a new first channel learning model and a new second channel learning model, that is, determine a new first channel learning model and a new second channel learning model for updating the first channel learning model and the second channel learning model. one or more configuration parameters of the model, and send the one or more configuration parameters for updating the second channel learning model to the second communication device for updating the previous second channel learning model.
  • the one or more configuration parameters of the new second channel learning model may be considered as one or more configuration parameters used to update the previous second channel learning model.
  • the first communication device may send one or more configuration parameters of the new second channel learning model that are different from the previous second channel learning model to the second communication device for updating the previous second channel learning model.
  • the second channel learning model may be considered as one or more configuration parameters used to update the previous second channel learning model.
  • the first communication device may determine a new second channel learning model, that is, determine one or more configuration parameters for updating the second channel learning model, and One or more configuration parameters of the new second channel learning model are sent to the second communication device for updating the previous second channel learning model.
  • This embodiment of the present application does not limit the manner in which the first communication apparatus determines one or more configuration parameters for updating the first channel learning model and/or the second channel learning model. That is, the embodiments of the present application do not limit the manner in which the first communication apparatus determines the first channel learning model and/or the second channel learning model.
  • the following embodiments provide the manner in which the first communication device determines the first channel learning model and/or the second channel learning model, and the manner in which the first communication device determines the channel learning model may be used as an independent embodiment, or may be related to other embodiments. In combination, specifically, this embodiment of the present application does not limit this. One or more of the following ways of determining the channel learning model may be used alone or in combination, and specifically, this embodiment of the present application does not limit this.
  • the first communication apparatus may determine the first channel learning model and/or the second channel learning model according to the first parameter.
  • the first parameter includes at least one of the following items: a cell identifier of a cell where the terminal device is located, a scene where the terminal device is located, a type of terminal device location, and a geographic location where the terminal device is located.
  • the first communication apparatus may determine the first parameter according to the cell where it is located, the scene where it is located, the type, or the geographic location where it is located.
  • the first communication device when the first communication device is a network device and the second communication device is a terminal device, the first communication device may receive information related to the first parameter sent by the second communication device, and determine the first parameter.
  • the scene in which the terminal device is located may be indoors, outdoors, suburbs, towns, external environments (eg, daytime, nighttime, sunny day, cloudy day, smooth traffic, traffic jam), and the like.
  • the first communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • One or more configuration parameters for the model may be used to determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • the type of the terminal device may be the number of antenna ports, processing capability, etc. of the terminal device.
  • the geographic location where the terminal device is located may be three-dimensional coordinates, two-dimensional coordinates, positioning data, and the like. As shown in FIG. 11 , if the geographic location where the terminal device is located is area 1, the first communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to area 1, and determine the first channel learning model and/or the second channel learning model corresponding to the area 1, and determine the first channel learning model and/or the second channel learning model for updating the first channel One or more configuration parameters of the channel learning model and/or the second channel learning model; if the geographic location where the terminal device is located is area 2, the first communication apparatus may determine the first channel learning model and/or the first channel learning model corresponding to area 2 A second channel learning model, and determining one or more configuration parameters for updating the first channel learning model and/or the second channel learning model.
  • one or more configuration parameters sent by the first communication apparatus to the second communication apparatus for updating the first channel learning model and/or the second channel learning model It can be terminal level, cell level or terminal group level.
  • the first communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to each second communication device in a unicast manner;
  • the first communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to a plurality of second communication devices in the cell in a broadcast manner;
  • the first communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to a plurality of second communication devices in the terminal group in a broadcast or multicast manner .
  • the first communication apparatus may determine the first channel learning model and the second channel learning model from the set of configuration parameters saved in the database #1 according to the first parameter.
  • the database #1 may be a database that stores relevant information or related data of the channel learning model, for example, the database #1 may store the configuration parameters of the channel learning model and the like.
  • the database #1 may be a local database of the first communication device, or may be a database stored at a high level.
  • the database stored in a high layer may be stored in a mobility management unit, a core network, a cloud, a central manager, an operator system, a first communication device group management system or data. database in the center.
  • the first communication device may communicate and interact with the upper layer, and determine the configuration parameters of the channel learning model from the database stored in the higher layer.
  • the first communication device may download and/or read the configuration parameters of the channel learning model stored in the database from the high-level database according to the first parameters, and then determine the first channel learning model and/or the second channel learning model.
  • the first communication apparatus may receive configuration parameters related to the channel learning model sent by the high-level network element, and then determine the first channel learning model and/or the second channel learning model.
  • the first communication device can download and/or read the configuration parameters of the channel learning model stored in the database from the local database, and then determine the first channel learning model. and/or the second channel learning model.
  • the configuration parameters of the channel learning model stored in the local database of the first communication device may be pre-configured, or may be determined by the first communication device side through training and learning and stored in the local database.
  • the first communication device may determine the channel learning model from the local database according to the first parameter before applying the channel learning model.
  • the first communication device may determine the first channel learning model and/or the second channel learning model according to the first parameter, and further determine the configuration parameters for updating the second channel learning model, that is, the channel learning model is not performed.
  • a new channel learning model can be determined after training, which reduces the processing complexity of the first communication device.
  • the first communication device may perform training on the first channel learning model and the second channel learning model to obtain the first channel learning model and the second channel learning model, and determine in one step for updating the first channel learning model and/or one or more configuration parameters of the second channel learning model.
  • the manner in which the first communication device trains the first channel learning model and the second channel learning model may be:
  • the above gradient information is sent to the first channel as one or more configuration parameters for updating the second channel learning model.
  • the first communication device continues to train the first channel learning model and the second channel learning model based on the above method until the value of the loss function is less than the preset threshold #10. Further, the multiple gradient information obtained in the multiple training processes is accumulated and sent to the second communication apparatus as one or more configuration parameters for updating and the second channel learning model.
  • the training parameters for training the channel learning model by the first communication apparatus may be pre-configured, or may be determined according to indication information from the second communication apparatus.
  • the first communication apparatus may determine each of the multiple configuration parameters by using the same method, or may use a different method to determine the multiple configuration parameters. each of the parameters.
  • the first communication apparatus may determine the configuration parameters for updating the channel learning model in a one-level or multi-level manner.
  • the first communication device may use one or more of the above methods to determine the configuration parameters for updating the channel learning model; taking the multi-stage mode as an example, the first communication device may first use the above method.
  • One or more of the above-mentioned methods determine a part of configuration parameters for updating the channel learning model, and then use one or more of the above methods to determine another part of the configuration parameters for updating the channel learning model.
  • the first communication device determines the structure of the channel learning model from the high-level database according to the first parameter, then determines the dimension, operation and/or function of the channel learning model from the local database according to the first parameter, and finally determines the dimension, operation and/or function of the channel learning model according to the data from the second parameter.
  • Information about the configuration parameters of the communication device determines the variables of the channel learning model.
  • the resource used for transmitting the first message the content of the first message or the form of sending the first message may be indicated by the first indication information; if the method 200 does not execute S230, Then the resource used for transmitting the first message, the content of the first message or the form of sending the first message may be pre-configured, so that the second communication apparatus can correctly receive the first message.
  • the method 200 may further include: the second communication device sends a second message, where the second message is used to indicate one or more configuration parameters for updating the first channel learning model; correspondingly, the first The communication device receives the second message, and updates the first channel learning model according to the second message.
  • the second communication device may determine a new first channel learning model and a new second channel learning model, that is, determine whether to update the first channel learning model.
  • One or more configuration parameters of a channel learning model and/or a second channel learning model and sending one or more configuration parameters of the new first channel learning model to the first communication device for updating the first channel Learning models.
  • the one or more configuration parameters of the new first channel learning model may be considered as one or more configuration parameters used to update the previous first channel learning model.
  • the second communication device may send to the first communication device one or more configuration parameters of the new first channel learning model that are different from the configuration parameters of the previous first channel learning model, for updating the previous first channel learning model.
  • the first channel learning model may be considered as one or more configuration parameters used to update the previous first channel learning model.
  • This embodiment of the present application does not limit the manner in which the second communication apparatus determines one or more configuration parameters for updating the first channel learning model and/or the second channel learning model. That is, the embodiments of the present application do not limit the manner in which the second communication apparatus determines the first channel learning model and/or the second channel learning model.
  • the following embodiments provide a method for the second communication device to determine the first channel learning model and/or the second channel learning model, and the method for the second communication device to determine the channel learning model may be used as an independent embodiment or may be related to other embodiments. In combination, specifically, this embodiment of the present application does not limit this.
  • One or more of the following methods for determining a channel learning model may be used alone or in combination, which is not specifically limited in this application.
  • the second communication apparatus may determine the first channel learning model and/or the second channel learning model according to the first parameter.
  • the first parameter includes at least one of the following: a cell identifier of a cell where the terminal device is located, a scene where the terminal device is located, a type of the terminal device, and a geographic location where the terminal device is located.
  • the second communication apparatus may determine the first parameter according to the cell where it is located, the scene where it is located, the type or the geographic location where it is located.
  • the second communication device when the second communication device is a network device and the first communication device is a terminal device, the second communication device may receive information related to the first parameter sent by the first communication device, and determine the first parameter.
  • the scene in which the terminal device is located may be indoors, outdoors, suburbs, towns, external environments (eg, daytime, nighttime, sunny day, cloudy day, smooth traffic, traffic jam), and the like.
  • the second communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • One or more configuration parameters for the model may be used to determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • the type of the terminal device may be the number of antenna ports, processing capability, etc. of the terminal device.
  • the geographic location where the terminal device is located may be three-dimensional coordinates, two-dimensional coordinates, positioning data, and the like. As shown in FIG. 11 , if the geographic location where the terminal device is located is area 1, the second communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to area 1, and determine the first channel learning model and/or the second channel learning model corresponding to the area 1, and determine the first channel learning model and/or the second channel learning model for updating the first channel One or more configuration parameters of the channel learning model and/or the second channel learning model; if the geographic location where the terminal device is located is area 2, the second communication apparatus may determine the first channel learning model and/or the first channel learning model corresponding to area 2 A second channel learning model, and determining one or more configuration parameters for updating the first channel learning model and/or the second channel learning model.
  • the one or more configuration parameters sent by the second communication apparatus to the first communication apparatus for updating the first channel learning model may be terminal-level, cell-level or Terminal group level.
  • the second communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to each first communication device in a unicast manner;
  • the second communication device sends one or more configuration parameters for updating the second channel learning model and/or the first channel learning model to multiple first communication devices in the cell in a broadcast manner;
  • the second communication device sends one or more configuration parameters for updating the second channel learning model and/or the first channel learning model to multiple first communication devices in the terminal group in a broadcast or multicast manner .
  • the second communication apparatus may determine the first channel learning model and/or the second channel learning model from the set of configuration parameters saved in the database #2 according to the first parameter.
  • the database #2 may be a database carrying relevant information or related data of the channel learning model, for example, the database #2 may store the configuration parameters of the channel learning model and the like.
  • the database #2 may be a local database of the second communication device, or may be a database stored at a higher level.
  • the database stored in a high layer may be stored in a mobility management unit, a core network, a cloud, a central manager, an operator system, a second communication device group management system, and data.
  • the second communication device may communicate and interact with the upper layer, and determine the configuration parameters of the channel learning model from the database of the higher layer.
  • the second communication device may download and/or read the channel learning model stored in the database from the high-level database according to the first parameter, and then determine a new first channel learning model and a new second channel learning model.
  • the second communication apparatus may receive configuration parameters about the channel learning model sent by the high-level network element, and then determine a new first channel learning model and a new second channel learning model.
  • the second communication device can download and/or read the channel learning model in the database from the local database, and then determine the new first channel learning model and the new channel learning model.
  • the second channel learning model can download and/or read the channel learning model in the database from the local database, and then determine the new first channel learning model and the new channel learning model.
  • the configuration parameter library of the channel learning model stored in the local database of the second communication device is pre-configured, or it may be determined by the second communication device side through training and learning and stored in the local database.
  • the second communication apparatus may determine the channel learning model from the local database according to the first parameter before applying the channel learning model.
  • the second communication device can determine the first channel learning model and/or the second channel learning model according to the first parameter, that is, a new channel learning model can be determined without performing channel learning model training, which reduces the need for second communication The processing complexity of the device.
  • the second communication apparatus may perform training on the first channel learning model and the second channel learning model to obtain the first channel learning model and the second channel learning model, and further determine to update the first channel learning model and/or one or more configuration parameters of the second channel learning model.
  • the manner in which the second communication apparatus trains the first channel learning model and the second channel learning model may be:
  • the above gradient information is sent to the first channel learning model as one or more configuration parameters for updating the first channel learning model.
  • the second communication device continues to train the first channel learning model and the second channel learning model based on the above method until the value of the loss function is less than the preset threshold. Further, the multiple gradient information obtained in multiple training processes is accumulated and sent to the second communication apparatus as one or more configuration parameters for updating the first channel learning model.
  • the first channel information may be sent by the first communication apparatus to the second communication apparatus.
  • the target channel information may be sent by the first communication apparatus to the second communication apparatus, or may be obtained by the second communication apparatus according to the reciprocity of the uplink and downlink channels.
  • the training parameters for training the channel learning model by the first communication apparatus may be pre-configured, or may be indicated by the first communication apparatus.
  • the second communication apparatus may determine each of the multiple configuration parameters by using the same method, or may use a different method to determine the multiple configuration parameters. each of the parameters.
  • the second communication apparatus may determine the configuration parameters for updating the channel learning model in a one-level or multi-level manner.
  • the second communication device may use one or more of the above methods to determine the configuration parameters for updating the channel learning model; taking the multi-stage method as an example, the second communication device may first use the above method.
  • One or more of the above-mentioned methods determine a part of configuration parameters for updating the channel learning model, and then use one or more of the above methods to determine another part of the configuration parameters for updating the channel learning model.
  • the second communication device determines the structure of the channel learning model from the high-level database according to the first parameter, determines the dimension, operation and/or function of the channel learning model from the local database according to the second parameter, and finally determines the dimension, operation and/or function of the channel learning model according to the second parameter.
  • Information about the configuration parameters of the communication device determines the variables of the channel learning model.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to the false alarm probability.
  • the second communication device receives the first message sent by J 1 first communication devices, and J 2 first communication devices feed back the first channel learning model Applicable (or J 2 first messages indicate that the first channel learning model is applicable), and J 3 first communication devices feedback that the first channel learning model is not applicable (or J 3 first messages indicate that the first channel learning model is not applicable) .
  • the false alarm probability can be the ratio of the number of first communication devices to which the first channel learning model is fed back (or the number of first messages indicating that the first channel learning model is applicable) to the number of first communication devices that perform channel learning model training (that is, J 2 /J); or it can be the ratio of the number of first communication devices to which the first channel learning model is applicable (or the number of first messages indicating that the first channel learning model is applicable) to the total number of received first messages (that is, J 2 /J 1 ); or it may be the number of first communication devices for which the feedback first channel learning model is not applicable (or the number of first messages indicating that the first channel learning model is not applicable) accounts for the first communication device that performs channel learning model training the ratio of the number (i.e., J 3 / J); or may be the number of the first communication device a first channel feedback learning model NA (or a first message indicating the first channel number is not applicable to track learning model) representing a first message received
  • the ratio of the total number ie J 3
  • the second communication device determines that the first channel learning model is applicable; if the false alarm probability is less than (or less than or equal to) the preset threshold #11, Then the second communication device determines that the first channel learning model is not applicable. Or, if the false alarm probability is greater than or equal to (or greater than) the preset threshold #12, the second communication device determines that the first channel learning model is not applicable; if the false alarm probability is less than (or less than or equal to) the preset threshold #12, Then the second communication device determines that the first channel learning model is applicable.
  • the preset threshold #11 is 70%, and if the ratio of the number of first messages indicating that the first channel learning model is applicable to the total number of first messages received by the second communication device is greater than 70%, the second communication device determines that the first A one-channel learning model applies.
  • the preset threshold #12 is 50%. If the ratio of the number of first messages indicating that the first channel learning model is not applicable to the total number of first messages received by the second communication device is greater than 50%, the second The communication device determines that the first channel learning model is not applicable.
  • the second communication device determines whether the channel learning model is applicable in the above manner, and can comprehensively consider the situation of the channel learning models of multiple first communication devices, so as to avoid making wrong decisions due to feedback from individual first communication devices, and can take into account multiple first communication devices.
  • the accuracy of determining whether the channel learning model is applicable is improved, and the communication performance is improved.
  • the second communication device may determine a new first channel learning model and a new second channel learning model, that is, determine a new first channel learning model and a new second channel learning model.
  • One or more configuration parameters of the model are learned, and the configuration parameters used to update the first channel learning model are sent to the first communication device.
  • the method 200 may further include: the first communication apparatus determines the first channel learning model according to the first parameter.
  • the first communication apparatus determines the first channel learning model
  • an embodiment of the present application provides a communication method that can be performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S210' to S230':
  • the first communication apparatus determines whether the first channel learning model is applicable. For a specific description of this step, reference may be made to the description in S210 above, which is not described in detail here for brevity.
  • the first communication apparatus sends a first message, where the first message is used to indicate whether the first channel learning model is applicable.
  • the specific description of this step can be described in the above S220, which is not described in detail here for the sake of brevity.
  • the first communication apparatus sends one or more configuration parameters for updating the second channel learning model.
  • This step can be described in the above S220, which is not described in detail here for the sake of brevity.
  • this step can be replaced by S230b'
  • the first communication apparatus receives one or more configuration parameters for updating the first channel learning model.
  • the first communication apparatus receives one or more configuration parameters for updating the first channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S210" to S230":
  • S210 that is, S220
  • the second communication device receives the first message, and the first message is used to indicate whether the first channel learning model is applicable.
  • the specific description of this step can refer to the description in S210 above. For brevity, no More details.
  • S220 the second communication device determines whether the first channel learning model is applicable according to the first message.
  • S220 the description in S220 above, which is not described in detail here for brevity.
  • the second communication device receives one or more configuration parameters for updating the second channel learning model.
  • This step can be described in the above S220, and for brevity, it will not be described in detail here.
  • this step can also be replaced by S230b".
  • the second communication device sends one or more configuration parameters for updating the first channel learning model.
  • the second communication device sends one or more configuration parameters for updating the first channel learning model.
  • the embodiment of the present application provides a communication method in which the first communication device determines that the channel learning model is not applicable and notifies the second communication device.
  • the method can report to the second communication device in time when the channel learning model is not applicable, thereby realizing the The update of the channel learning model can improve the accuracy and communication performance of the channel learning model.
  • FIG. 12 shows a communication method provided by another embodiment of the present application.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • the method 300 shown in FIG. 12 may include S310 to S320. Each step in the method 300 is described in detail below.
  • the first communication apparatus determines whether the first channel learning model is applicable.
  • the first channel learning model is used to determine the first channel information based on the target channel information, and the data volume of the first channel information is smaller than that of the target channel information, so it can also be said that the first channel learning model is used to compress the target channel information to Obtain first channel information.
  • the data amount of the channel information may refer to the dimension of the channel information.
  • the number of antenna ports of the transmitting end (for example, it may be the first communication device or the second communication device) is A 2
  • the number of antenna ports of the receiving end (for example, it may be the first communication device or the second communication device) is A 2 .
  • the target channel information between the transmitter and the receiver can be a 3 -dimensional matrix of A 2 *A
  • the data amount of the target channel information can be represented by A 2 *A 3 . If the elements in the matrix of the target channel information are complex numbers, and the real part and the imaginary part of each element are expressed separately, the data amount of the target channel information can also be expressed as A 2 *A 3 *2.
  • the data amount of the first channel information may be represented by B 2 .
  • the data amount of the channel information may also refer to the amount of information included in the channel information, and the like.
  • the target channel information may be regarded as the input of the first channel learning model, and the first channel information may be regarded as the output of the first channel learning model.
  • the data volume of the target channel information may be the input information dimension, and the data volume of the first channel information may be the output information dimension.
  • the first channel information is used to obtain the second channel information through the second channel learning model, and the data amount of the second channel information and the target channel information is the same or similar.
  • the second channel information may be used for data transmission, for example, the second communication apparatus may determine scheduling information for data transmission, etc., or determine precoding for data transmission, etc., according to the second channel information.
  • the first channel learning model corresponds to the second channel learning model. Therefore, determining whether the first channel learning model is applicable by the first communication device can be understood as the first communication device determining the first channel learning model and the second channel learning model. Whether the learning model is applicable. That is, when the first communication device determines that the first channel learning model is not applicable, it may determine that the second channel learning model is also not applicable; when the first communication device determines that the first channel learning model is applicable, it may determine that the second channel learning model is applicable. Channel learning models are also applicable.
  • the target channel information may be downlink channel information; for another example, when the first communication apparatus is a network device, the target channel information may be uplink channel information.
  • the target channel information may be uplink channel information, or the target channel information may be uplink channel information and downlink channel information, and the first communication apparatus may be based on the part of the uplink and downlink channels. For reciprocity, determine whether the first channel learning model and/or the second channel learning model is applicable according to the uplink channel information and the downlink channel information, or determine a new first channel learning model and/or the second channel learning model.
  • the target channel information may be downlink channel information, or the target channel information may be uplink channel information and downlink channel information. Based on the partial reciprocity of the uplink and downlink channels, according to The uplink channel information and the downlink channel information determine whether the first channel learning model and/or the second channel learning model is applicable, or determine a new first channel learning model and/or the second channel learning model.
  • the first communication device may periodically determine whether the first channel learning model is applicable. For example, the first communication device starts a timer after determining whether the first channel learning model is applicable for the i-th time. The communication device determines whether the first channel learning model is applicable for the i+1th time.
  • the overhead of interactive signaling between the first communication device and the second communication device can be reduced, and the first communication device can periodically determine whether the first channel learning model is applicable, which can avoid the situation that the first channel learning model is not applicable resulting in a decrease in communication performance.
  • This embodiment of the present application does not limit the method for the first communication apparatus to determine whether the first channel learning model is applicable.
  • the first communication apparatus may use one or more of the following implementation manners to determine whether the first channel learning model is applicable.
  • the first communication device and/or the second communication device may adjust the first channel learning model in time to improve the accuracy and applicability of the channel learning model, thereby improving the communication performance.
  • the following provides a way for the first communication device to determine whether the first channel learning model and/or the second channel learning model is applicable.
  • the method for the first communication device to determine whether the channel learning model is applicable can be used as an independent embodiment, and can also be implemented with other In combination with examples, specifically, the embodiments of the present application do not limit this.
  • One or more of the following ways of determining whether the channel learning model is applicable may be used alone or in combination, which is not specifically limited in this embodiment of the present application.
  • channel learning model mentioned in the embodiments of this application may also refer to whether the channel learning model matches, whether the channel learning model is accurate, whether the channel learning model is outdated, or whether the channel learning model is wrong, etc.
  • the first communication apparatus may determine whether the channel learning model is applicable according to the long-term statistical characteristics of the target channel. For example, the first communication apparatus may determine whether the first channel learning model is applicable according to whether the change amount of the long-term statistical characteristic of the target channel is greater than or equal to the first preset threshold. Specifically, reference may be made to the description in S210 above, which is not repeated here for brevity.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the received first scheduling information.
  • the first scheduling information is sent by the second communication apparatus according to the second channel information. Specifically, reference may be made to the description in S210 above, which is not repeated here for brevity.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to the data transmission performance. Specifically, reference may be made to the description in S210 above, which is not repeated here for brevity.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the scene in which it is located changes. Specifically, reference may be made to the description in S210 above, which is not repeated here for brevity.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the performance index of the first channel learning model is smaller than the second preset threshold. Specifically, reference may be made to the description in S210 above, which is not repeated here for brevity.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to an error between the target channel information and the second channel information. Specifically, reference may be made to the description in S210 above, which is not repeated here for brevity.
  • the first communication apparatus may determine whether the first channel learning model is applicable according to whether the first indication information is received.
  • the first indication information is used to instruct the first communication apparatus to perform channel learning model training.
  • the first communication apparatus may determine that the first channel learning model is not applicable in the case of receiving the first indication information.
  • the first indication information is sent by the second communication apparatus, and the second communication apparatus may send the first indication information when it is determined that the second channel learning model is not applicable.
  • the first communication device can determine whether the first channel learning model is applicable according to whether the first indication information is received, that is, if the channel learning model training is not performed, determining whether the first channel learning model is applicable can reduce the Processing complexity of the first terminal device.
  • the first indication information it can be determined whether the first channel learning model is applicable under the condition that the two communication devices have unified understanding and recognition, so this method is simple and fast.
  • the first communication device or the second communication device determines that the first channel learning model is not applicable, because the first channel learning model and the second channel learning model are corresponding, it can be inferred that, Either the first communication device or the second communication device determines that the second channel learning model is not applicable.
  • the first communication device or the second communication device determines that the second channel learning model is not applicable, because the first channel learning model and the second channel learning model are corresponding, it can be inferred that, Either the first communication device or the second communication device determines that the first channel learning model is not applicable.
  • the first communication device or the second communication device determines that the first channel learning model or the second channel learning model is not applicable, because the first channel learning model and the second channel learning model are corresponding , so it can be inferred that the first communication device or the second communication device determines that the first channel learning model and the second channel learning model are not applicable.
  • This embodiment of the present application does not limit the method for the second communication apparatus to determine whether the second channel learning model is applicable.
  • the second communication apparatus may use one or more of the following implementations to determine whether the second channel learning model is applicable.
  • the first communication device and/or the second communication device may adjust the second channel learning model in time to improve the accuracy and applicability of the channel learning model, thereby improving the communication performance.
  • the following provides a way for the second communication device to determine whether the first channel learning model and/or the second channel learning model is applicable.
  • the method for the second communication device to determine whether the channel learning model is applicable can be used as an independent embodiment, and can also be implemented with other In combination with examples, specifically, the embodiments of the present application do not limit this.
  • One or more of the following ways of determining whether the channel learning model is applicable may be used alone or in combination, which is not specifically limited in this embodiment of the present application.
  • channel learning model mentioned in the embodiments of this application may also refer to whether the channel learning model matches, whether the channel learning model is accurate, whether the channel learning model is outdated, or whether the channel learning model is wrong, etc.
  • the second communication apparatus may determine whether the channel learning model is applicable according to the long-term statistical characteristics of the target channel. For example, when the second communication device determines that the long-term statistical characteristics of the target channel change greatly, it indicates that the channel characteristics or the channel environment between the first communication device and the second communication device have changed greatly, so the second communication device can determine The channel learning model does not apply.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to whether the variation of the long-term statistical characteristics of the target channel is greater than or equal to a third preset threshold.
  • the third preset threshold may be predefined by the protocol or may be preconfigured, which is not limited in this embodiment of the present application.
  • the long-term statistical characteristics of the target channel may include at least one of the following: rank value, large-scale characteristics, channel covariance matrix, channel correlation matrix, coherence time, coherence bandwidth, and the like.
  • the large-scale characteristics of the channel can be one or more of the following: delay spread, Doppler spread, Doppler shift, average channel gain and average delay, receive angle of arrival, angle of arrival spread, transmit departure angle, departure Angular spread, spatial reception parameters, and spatial correlation.
  • the second communication device may determine the long-term statistical characteristics of the target channel according to the signal from the first communication device, and further determine whether the second channel learning model is applicable according to the long-term statistical characteristics of the target channel.
  • the signal from the second communication device may be a reference signal or a data signal, the reference signal may be DMRS, CSI-RS, PTRS, TRS, SBB, SRS, etc., and the data signal may be transmitted on PDSCH, PDCCH, PUSCH or PUCCH signal of.
  • the network device may determine the long-term statistical characteristics of the target channel according to the reference signal and/or the data signal sent by the terminal device.
  • the reference signal may be at least one of SRS, DMRS, and PTRS
  • the data signal may be at least one of a signal transmitted on PUSCH and a signal transmitted on PUCCH.
  • the target channel may refer to a downlink channel, and the network device may infer the long-term statistical characteristics of the downlink channel according to the long-term statistical characteristics of the uplink channel.
  • the second communication apparatus may determine the variation of the long-term statistical characteristics of the target channel according to the channel state information fed back by the first communication apparatus.
  • the channel state information (channel state information, CSI) reported by the first communication device may include at least one of the following: a rank value, a channel covariance matrix, a correlation matrix, and the like.
  • the second communication device may determine the long-term statistical characteristics of the target channel according to the information reported by the first communication device.
  • the second communication device may determine that the second channel learning model is not applicable; when the long-term statistical characteristics of the target channel are less than the third preset threshold, it indicates that the channel characteristics or the channel environment between the first communication device and the second communication device are relatively stable, Thus the second communication device may determine that the second channel learning model is applicable.
  • the second communication apparatus may determine that the second channel learning model is not applicable when it is determined that the rank value changes greatly.
  • the second communication apparatus may determine that the second channel learning model is not applicable when it is determined that the variation of the rank value is greater than or equal to the preset threshold #12 (an example of the third preset threshold).
  • the preset threshold #12 may be R 1 , where R 1 is a positive integer. For example, R 1 is 2, that is, when the change of the rank value is greater than or equal to 2, the second communication apparatus may determine that the second channel learning model is not applicable.
  • the path of the target channel increases, and the rank value changes.
  • the second channel learning model may no longer be applicable.
  • the larger the rank value is the more complex the channel learning model used may be, for example, the higher the number of layers of the channel learning model may be.
  • the long-term statistical characteristic (rank value) of the target channel also changes, and the second communication device may determine that the second channel learning model may not be applicable.
  • the second communication apparatus may determine that the second channel learning model is not applicable when it is determined that the Doppler frequency shift changes greatly. For example, the second communication apparatus may determine that the second channel learning model is not applicable when the variation of the Doppler frequency shift is greater than the preset threshold #13 (an example of the third preset threshold).
  • the preset threshold #13 may be F 2 , where F 2 is a real number. For example, F 2 is 2, that is, when the variation of the Doppler frequency shift is greater than or equal to 2, the second communication apparatus may determine that the second channel learning model is not applicable.
  • the Doppler frequency shift can reflect the moving speed of the first communication device.
  • the second channel learning model may no longer be applicable. For example, if the variation of the movement speed of the first communication device is greater than the preset threshold #14, the second communication device determines that the second channel learning model is not applicable.
  • the second communication device can determine whether the channel learning model is applicable according to the long-term statistical characteristics of the target channel, that is, without performing channel learning model training, determining whether the second channel learning model is applicable can reduce the second communication
  • the processing complexity of the device and at the same time, it can determine whether the channel learning model is applicable without the assistance of the first signal device, and the signaling interaction is reduced, so this method is simple and fast.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to the first scheduling information.
  • the first scheduling information may be determined by the second communication apparatus according to the second channel information.
  • the first scheduling information may include at least one of the following items: MCS indication, TBS indication, rank indication, antenna port indication, and the like.
  • the first scheduling information may be downlink control information (downlink control information, DCI) of the physical layer, or may be scheduling information in higher-layer signaling.
  • DCI downlink control information
  • the second communication apparatus may determine whether the second channel learning model is applicable according to the variation of the first scheduling information determined multiple times. For example, the second communication apparatus may determine whether the second channel learning model is applicable according to the difference between the first scheduling information determined in the previous B' times and the currently determined first scheduling information. The second communication apparatus may determine that the second channel learning model is not applicable in the case that the variation of the first scheduling information determined multiple times is greater than or equal to the preset difference threshold.
  • B' is an integer.
  • the scheduling information of the data can reflect the quality of the channel, so the second communication apparatus can determine whether the second channel learning model is applicable according to the difference of the scheduling information multiple times. For example, when the difference between the first scheduling information indicated by the second communication device to the first communication device B' times before and the first scheduling information currently instructed is large, it indicates that the channel of the first communication device has changed greatly, that is, the first Two-channel learning model is not applicable. Therefore, the second communication apparatus may determine whether the second channel learning model is applicable according to at least one of the following manners.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to whether the difference between the first scheduling information determined B' times before and the currently determined first scheduling information is greater than or equal to a preset difference threshold.
  • the preset difference threshold may be predefined by a protocol, may be indicated by the first communication apparatus to the second communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the second communication device determines that the second channel learning model is not applicable when the difference between the first scheduling information determined B' times before and the currently determined first scheduling information is greater than or equal to a preset difference threshold;
  • the communication apparatus determines that the second channel learning model is applicable when the difference value between the first scheduling information determined in the previous B' times and the currently determined first scheduling information is smaller than a preset difference threshold.
  • the preset difference threshold may be at least one of the difference threshold of modulation order, the difference threshold of code rate, the difference threshold of MCS index (index) value, the difference threshold of rank value, the difference threshold of TBS, and the difference threshold of antenna port. .
  • the difference threshold of the modulation order can be N 3 orders, where N 3 is a real number, for example, N 3 is 1, 2, 3, 4, 1/2, 3/ 2, 5/2 etc.
  • the second communication apparatus determines that the second channel learning model is inapplicable when the difference value between the order of the modulation scheme determined in the previous B' times and the order of the modulation scheme currently determined is greater than or equal to the difference threshold of the modulation order.
  • the first communication device determines that the first channel learning model is not applicable.
  • the code rate difference threshold may be M 1 , where M 1 is a real number, for example, M 1 is 200/1024, 250/1024, 300/1024, 500/1024, and so on.
  • the second communication apparatus determines that the second channel learning model is not applicable in the case that the difference value between the code rate determined in the previous B' times and the currently determined code rate is greater than or equal to the difference threshold of the code rate.
  • the second communication device may determine that the second channel learning model is not applicable.
  • the difference threshold of the MCS index value may be P 1 , and P 1 is an integer, for example, P 1 is 1, 2, 3, 4, 5, 6, 8, 10 Wait.
  • the second communication apparatus determines that the second channel learning model is not applicable when the difference value between the MCS index value determined at the B previous times and the currently determined MCS index value is greater than or equal to the difference threshold of the MCS index value.
  • the second communication device can determine the second channel learning model Not applicable.
  • the difference threshold of TBS may be Q, where Q is an integer, for example, Q is 32, 64, 128, 256, 612, 1024 and so on.
  • the second communication apparatus determines that the second channel learning model is not applicable when the difference value between the TBS determined in the previous B' times and the currently determined TBS is greater than or equal to the difference threshold of the TBS.
  • the second communication device may secondly determine that the channel learning model is not applicable.
  • the second communication device may further determine whether the second channel learning model is applicable according to the difference value between the first scheduling information determined B' times before and the multiple items in the currently determined first scheduling information.
  • the second communication device may have a difference value between the code rate determined in the previous B' times and the currently determined code rate greater than or equal to the difference threshold of the code rate, and the difference value between the TBS determined in the previous B' times and the TBS currently determined When the value is greater than or equal to the TBS difference threshold, it is determined that the second channel learning model is not applicable.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to whether the difference value of the first scheduling information is greater than or equal to the preset difference threshold.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to whether the similarity of the first scheduling information is less than a preset similarity threshold.
  • the preset similarity threshold value may be predefined by a protocol, may be indicated by the first communication apparatus to the second communication apparatus, or may be preconfigured, which is not limited in this embodiment of the present application.
  • the second communication apparatus may further determine whether the second channel learning model is applicable according to the difference between the CQI information fed back by the first communication apparatus and the first scheduling information. Specifically, the second communication apparatus may determine that the second channel learning model is not applicable when the difference between the CQI information fed back by the first communication apparatus and the first scheduling information is greater than or equal to a preset difference threshold.
  • the second communication device may determine whether the second channel learning model is applicable according to the first scheduling information, the CQI information fed back by the second communication device, and the second mapping relationship, and the second mapping relationship is used to indicate whether the scheduling information and the channel learning model are applicable corresponding relationship.
  • Table 12 shows an example of the second mapping relationship.
  • the second mapping relationship may be one or more lines in the following table.
  • the second mapping relationship may be predefined by the protocol, or may be notified by the second communication device to the first communication device through signaling. Specifically, this application implements The example does not limit this.
  • the second communication device determines that the second channel learning model is applicable according to Table 12;
  • the code rate indicated by the CQI information fed back by a communication device is 2/3, and the code rate indicated by the first scheduling information is 3/4, then the first communication device can determine according to Table 12 that the second channel learning model is not applicable.
  • the second communication device can determine whether the channel learning model is applicable according to the first scheduling information, that is, without performing channel learning model training, determining whether the second channel learning model is applicable can reduce the complexity of the second communication device. Handling complexity. At the same time, according to determining whether the first channel learning model is applicable, the communication performance based on the first channel learning model can also be guaranteed. Furthermore, in this implementation, the assistance of the first communication device is not required, so signaling interaction can be reduced.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to the data transmission performance. Specifically, reference may be made to the description in S210 about whether the first communication apparatus determines whether the first channel learning model is applicable according to the data transmission performance, which is not described in detail here for brevity.
  • the data transmission performance may refer to the data transmission performance of a certain first communication device that communicates with the second communication device, and whether the second channel learning model corresponding to the first communication device is applicable may be determined according to the data transmission performance.
  • the second channel learning model corresponding to the first communication device that is, the second channel learning model corresponding to the first channel learning model deployed on the side of the first communication device.
  • the data transmission performance may also refer to the data transmission performance of the cell of the second communication apparatus, such as cell edge throughput, cell center throughput, and total cell throughput.
  • the second communication apparatus may determine whether the second channel learning model corresponding to the cell is applicable.
  • the second channel learning model corresponding to the cell that is, the second channel learning model corresponding to the first communication device in the cell.
  • the second communication device can determine whether the second channel learning model is applicable according to the data transmission performance, that is, determining whether the second channel learning model is applicable without performing the channel learning model training, which can reduce the speed of the second communication device. processing complexity.
  • the way to determine whether the second channel learning model is suitable according to the data transmission performance is to measure the accuracy of the second channel learning model based on the final communication performance. Therefore, the communication performance based on the second channel learning model can also be guaranteed, that is, the way to help improve communication performance.
  • the second communication apparatus may determine whether the first channel learning model is applicable according to whether the scene in which the first communication apparatus is located changes.
  • the scene may be at least one of the following: indoor stillness, outdoor stillness, low-speed motion, high-speed motion, suburban, town, macro station, micro station, vehicle scene, V2X scene, scene defined by 3GPP protocol, etc.
  • the channel of the first communication device and the second communication device is a direct path; and in some scenarios, the channel of the first communication device and the second communication device is a non-direct path.
  • there are few reflectors between the first communication device and the second communication device, and correspondingly, the channel between the first communication device and the second communication device is simple; There are many reflectors between the communication device and the second communication device, and correspondingly, the channel between the first communication device and the second communication device is complicated.
  • the second channel learning model may not be applicable. For example, if the first communication device moves from indoor to outdoor, the second communication device may determine that the second channel learning model is not applicable; for another example, if the first communication device moves from a macro station to a micro station, the second communication device may determine that The second channel learning model is not applicable.
  • the first communication device may send location information to the second communication device, and the second communication device determines the scene where the first communication device is located according to the received location information, and then determines the first communication device according to the scene where the first communication device is located. Whether the two-channel learning model is applicable.
  • the first communication apparatus may send scene information to the second communication apparatus, and the second communication apparatus determines whether the second channel learning model is applicable according to the received scene where the first communication apparatus is located.
  • the second communication device may determine whether the second channel learning model is applicable according to the scene in which the first communication device is located, that is, in the case of not performing channel learning model training, to determine whether the second channel learning model is applicable.
  • the processing complexity of the second communication device is reduced, and the method is simple and fast.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to the scene in which it is located.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to the performance index of the second channel learning model.
  • the second communication apparatus may determine whether the channel learning model is applicable according to the difference or similarity between the target channel information (or the first channel information) and the second channel information. For example, the second communication apparatus may compress and decompress the target channel information according to the second channel learning model to obtain the second channel information. For example, the second communication apparatus may decompress the first channel information according to the second channel learning model to obtain the second channel information. Whether the second channel learning model is applicable is determined by comparing the characteristics of the target channel information and the second channel information, or by comparing the characteristics of the first channel information and the second channel information, and then determining the error of the second channel learning model. Specifically, reference may be made to the description in S210 about whether the first communication device determines whether the first channel learning model is applicable according to the performance index of the first channel learning model, which is not described in detail here for brevity.
  • the target channel information may be sent by the first communication apparatus to the second communication apparatus, or may be obtained by the second communication apparatus according to the reciprocity of the uplink and downlink channels.
  • the first channel information may be sent by the first communication apparatus to the second communication apparatus.
  • the second communication device may determine whether the second channel learning model is applicable according to the performance index of the channel learning model, that is, by performing channel learning model training to determine whether the second channel learning model is applicable, this method may not depend on the first With the assistance of the communication device, the second communication device can determine whether the second channel learning model is applicable, so this method is simple and fast
  • the second communication apparatus may determine whether the second channel learning model is applicable according to an error between the target channel information and the second channel information. Specifically, reference may be made to S210 about whether the first communication device may determine whether the first channel learning model is applicable according to the error between the target channel information and the second channel information, which is not described in detail here for brevity.
  • the target channel information may be sent by the first communication apparatus to the second communication apparatus, or may be obtained by the second communication apparatus according to the reciprocity of the uplink and downlink channels.
  • the second communication device may determine whether the second channel learning model is applicable according to the error between the target channel information and the second channel information, that is, determine whether the second channel learning model is applicable by performing channel learning model training. With the assistance of the first communication device, the second communication device can determine whether the channel learning model is applicable, so this method is simple and fast. Determining whether the second channel learning model is applicable under the condition of considering the characteristics of the target channel information and the second channel information at the same time can ensure that the first communication device and the second communication device obtain the same or similar characteristics of the channel information, which is helpful for The performance of data transmission is improved when the channel information is used for subsequent data transmission.
  • the second communication apparatus may determine whether the second channel learning model is applicable according to the performance index of the channel learning model sent by the first communication apparatus.
  • the manner in which the first communication apparatus determines the performance index of the channel learning model may refer to the description in S210, which is not described in detail here for brevity.
  • the second communication device can determine whether the second channel learning model is applicable according to the performance index of the channel learning model sent by the first communication device, that is, the second communication device can determine the second channel without training the channel learning model. Learning whether the model is applicable reduces the processing complexity of the second communication device, so this method is simple and fast.
  • S320 The first communication apparatus sends a first message, where the first message is used to indicate one or more configuration parameters for updating the second channel learning model.
  • the second communication apparatus receives the first message, determines that the second channel learning model is not applicable according to the first message, and updates the second channel learning model.
  • the first message may also include one or more configuration parameters for updating the first channel learning model.
  • one or more configuration parameters for updating the first channel learning model may also be sent through other messages, and the other messages are different from the first message, which is not specifically limited in this embodiment of the present application.
  • the first communication device may determine a new first channel learning model and a new second channel learning model, and configure one or more configurations of the new second channel learning model.
  • the parameters are sent to the second communication device for updating the second channel learning model.
  • the first communication device may send one or more configuration parameters of the new second channel learning model that are different from the previous second channel learning model to the second communication device for updating the previous second channel learning model.
  • the second channel learning model may be determined a new first channel learning model and a new second channel learning model, and configure one or more configurations of the new second channel learning model.
  • the parameters are sent to the second communication device for updating the second channel learning model.
  • the first communication device may send one or more configuration parameters of the new second channel learning model that are different from the previous second channel learning model to the second communication device for updating the previous second channel learning model.
  • the second channel learning model may be used to determine a new first channel learning model and a new second channel learning model, and configure one or more configurations of the new second channel learning model.
  • the parameters are sent to
  • the first communication device may determine a new second channel learning model, and send one or more configuration parameters of the new second channel learning model to the first communication device.
  • the first message includes one or more configuration parameters for updating the second channel learning model.
  • the second communication apparatus may update the second channel learning model according to the first message.
  • the one or more configuration parameters may include at least one of the following: model type, model structure, model algorithm, model weight vector, model weight matrix, model bias vector, model bias matrix, model activation function.
  • the types of models include machine learning algorithms, neural network models or auto-encoding models.
  • the configuration parameter may include a model algorithm for indicating which one of the machine learning algorithms the channel learning model is specifically.
  • the configuration parameters may include a model structure, which is used to indicate which type of neural network the channel learning model is, and the model structure may also include the dimension of the input layer and the dimension of the output layer.
  • the configuration parameters may include one or more of a transformation algorithm, a weight matrix, a weight vector, a bias vector, a bias matrix, and an activation function.
  • the configuration parameter may include a model structure for indicating the number of layers and/or the sequence of layers.
  • the configuration parameters can also include one or more of the following: parameters for the data input layer (preprocessing operation algorithm, dimension of input data, value range of input data), parameters for the convolution layer (size of input unit, Receptive field, stride, number of zero padding, depth, depth of output unit, weight matrix), parameters for excitation layer (activation function), parameters for pooling layer (pooling algorithm, spatial range, stride, The size of the input unit, the size of the output unit), the parameters for the fully connected layer (weight matrix, weight vector, bias matrix, bias vector).
  • This embodiment of the present application does not limit the manner in which the first communication apparatus determines one or more configuration parameters for updating the first channel learning model and the second channel learning model. That is, the embodiments of the present application do not limit the manner in which the first communication apparatus determines the first channel learning model and/or the second channel learning model.
  • the following embodiments provide the manner in which the first communication device determines the first channel learning model and/or the second channel learning model, and the manner in which the first communication device determines the channel learning model may be used as an independent embodiment, or may be related to other embodiments. In combination, specifically, this application does not limit this.
  • One or more of the following manners for determining the channel learning model may be used alone or in combination, which is not specifically limited in this application.
  • the first communication apparatus may determine the first channel learning model and the second channel learning model according to the first parameter.
  • the first parameter includes at least one of the following: a cell identifier of a cell where the terminal device is located, a scene where the terminal device is located, a type of the terminal device, and a geographic location where the terminal device is located.
  • the first communication apparatus may determine the first parameter according to the cell where it is located, the scene where it is located, the type, or the geographic location where it is located.
  • the first communication device when the first communication device is a network device and the second communication device is a terminal device, the first communication device may receive information related to the first parameter sent by the second communication device, and determine the first parameter.
  • the scene in which the terminal device is located may be indoors, outdoors, suburbs, towns, external environments (eg, daytime, nighttime, sunny day, cloudy day, smooth traffic, traffic jam), and the like.
  • the first communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • One or more configuration parameters for the model may be used to determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • the type of the terminal device may be the number of antenna ports, processing capability, etc. of the terminal device.
  • the geographic location where the terminal device is located may be three-dimensional coordinates, two-dimensional coordinates, positioning data, and the like. As shown in FIG. 11 , if the geographic location where the terminal device is located is area 1, the first communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to area 1, and determine the first channel learning model and/or the second channel learning model corresponding to the area 1, and determine the first channel learning model and/or the second channel learning model for updating the first channel One or more configuration parameters of the channel learning model and/or the second channel learning model; if the geographic location where the terminal device is located is area 2, the first communication apparatus may determine the first channel learning model and/or the first channel learning model corresponding to area 2 A second channel learning model, and determining one or more configuration parameters for updating the first channel learning model and/or the second channel learning model.
  • one or more configuration parameters sent by the first communication apparatus to the second communication apparatus for updating the first channel learning model and/or the second channel learning model It can be terminal level, cell level or terminal group level.
  • the first communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to each second communication device in a unicast manner;
  • the first communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to a plurality of second communication devices in the cell in a broadcast manner;
  • the first communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to a plurality of second communication devices in the terminal group in a broadcast or multicast manner .
  • the first communication apparatus may determine the first channel learning model and the second channel learning model from the set of configuration parameters saved in the database #1 according to the first parameter.
  • the database #1 may be a database that stores relevant information or related data of the channel learning model, for example, the database #1 may store the configuration parameters of the channel learning model and the like.
  • the database #1 may be a local database of the first communication device, or may be a database stored at a high level.
  • the database stored in a high layer can be stored in a mobility management unit, a core network, a cloud, a central manager, an operator system, a first communication device group management system, and a data center. in the database.
  • the first communication device may communicate and interact with the upper layer, and determine the configuration parameters of the channel learning model from the database of the higher layer.
  • the first communication device may download and/or read the configuration parameters of the channel learning model stored in the database from the high-level database, and then determine the first channel learning model and/or the second channel learning model.
  • the first communication apparatus may receive configuration parameters related to the channel learning model sent by the high-level network element, and further determine the first channel learning model and/or the second channel learning model.
  • the first communication device can download and/or read the configuration parameters of the channel learning model stored in the database from the local database, and then determine the first channel learning model. and/or the second channel learning model.
  • the configuration parameters of the channel learning model stored in the local database of the first communication device may be pre-configured, or may be determined by the first communication device side through training and learning and stored in the local database.
  • the first communication device may determine the channel learning model from the local database according to the first parameter before applying the channel learning model.
  • the first communication device may determine the first channel learning model and/or the second channel learning model according to the first parameter, and further determine the configuration parameters for updating the second channel learning model, that is, the channel learning model is not performed.
  • a new channel learning model can be determined after training, which reduces the processing complexity of the first communication device.
  • the first communication device may perform training on the first channel learning model and the second channel learning model to obtain the first channel learning model and the second channel learning model, and further determine to update the first channel learning model and/or one or more configuration parameters of the second channel learning model.
  • the manner in which the first communication device trains the first channel learning model and the second channel learning model may be:
  • the above gradient information is sent to the first channel as one or more configuration parameters for updating the second channel learning model.
  • the first communication device continues to train the first channel learning model and the second channel learning model based on the above method until the value of the loss function is less than the preset threshold. Further, the multiple gradient information obtained in multiple training processes is accumulated and sent to the second communication apparatus as one or more configuration parameters for updating the second channel learning model.
  • the training parameters for training the channel learning model by the first communication apparatus may be pre-configured, or may be determined according to indication information from the second communication apparatus.
  • the first communication apparatus may determine each of the multiple configuration parameters by using the same method, or may use a different method to determine the multiple configuration parameters. each of the parameters.
  • the first communication apparatus may determine the configuration parameters for updating the channel learning model in a one-level or multi-level manner.
  • the first communication device may use one or more of the above methods to determine the configuration parameters for updating the channel learning model; taking the multi-stage mode as an example, the first communication device may first use the above method.
  • One or more of the above-mentioned methods determine a part of configuration parameters for updating the channel learning model, and then use one or more of the above methods to determine another part of the configuration parameters for updating the channel learning model.
  • the first communication device determines the structure of the channel learning model from the high-level database according to the first parameter, then determines the dimension, operation and/or function of the channel learning model from the local database according to the first parameter, and finally determines the dimension, operation and/or function of the channel learning model according to the data from the second parameter.
  • Information about the configuration parameters of the communication device determines the variables of the channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S310' to S320':
  • the first communication apparatus determines whether the first channel learning model is applicable. For a specific description of this step, reference may be made to the description in S310 above, which is not described in detail here for brevity.
  • the first communication apparatus sends a first message, where the first message is used to indicate one or more configuration parameters for updating the second channel learning model.
  • the first message is used to indicate one or more configuration parameters for updating the second channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S310" to S320":
  • the second communication device receives the first message, where the first message is used to indicate one or more configuration parameters used to update the second channel learning model.
  • the first message is used to indicate one or more configuration parameters used to update the second channel learning model.
  • S320 that is, S320
  • the second communication device determines that the second channel learning model is not applicable according to the first message, and updates the second channel learning model.
  • the specific description of this step can refer to the description in S320 above. For brevity, this will not be detailed here.
  • the embodiment of the present application provides a communication method in which the first communication device determines that the channel learning model is not applicable and notifies the second communication device.
  • the method can report to the second communication device in time when the channel learning model is not applicable, thereby realizing the The update of the channel learning model can improve the accuracy and communication performance of the channel learning model.
  • FIG. 13 shows a communication method provided by another embodiment of the present application.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • the method 400 shown in FIG. 13 may include S410 to S420. Each step in the method 400 is described in detail below.
  • the second communication apparatus determines whether the second channel learning model is applicable.
  • the second channel learning model is used to determine the second channel information according to the first channel information, and the data amount of the second channel information and the target channel information is the same or similar.
  • the second channel information may be used for data transmission, for example, the second communication apparatus may determine scheduling information for data transmission, etc., or determine precoding for data transmission, etc., according to the second channel information.
  • the first channel information may be determined according to the first channel learning model and the target channel information, and the data volume of the first channel information is smaller than the data volume of the target channel information, so it can also be said that the first channel learning model is used for the target channel information.
  • the information is compressed to obtain first channel information.
  • the description about the data amount of the channel information may be the same as the description in S210 above, which is not described in detail here for brevity.
  • the second communication device determining whether the second channel learning model is applicable can be understood as the second communication device determining the second channel learning model and the first channel Whether the learning model is applicable. That is, when the second communication device determines that the second channel learning model is not applicable, it may determine that the first channel learning model is also not applicable; when the second communication device determines that the second channel learning model is applicable, it may determine that the first channel learning model is not applicable. Channel learning models are also applicable.
  • the second communication device may periodically determine whether the second channel learning model is applicable. For example, the second communication device starts a timer after determining whether the second channel learning model is applicable for the i-th time. The communication device determines whether the second channel learning model is applicable for the i+1th time.
  • the overhead of interactive signaling between the first communication device and the second communication device can be reduced, and the second communication device can periodically determine whether the second channel learning model is applicable, which can avoid the situation that the second channel learning model is not applicable resulting in a decrease in communication performance.
  • This embodiment of the present application does not limit the method for the second communication apparatus to determine whether the second channel learning model is applicable. Specifically, for the method for the second communication apparatus to determine whether the second channel learning model is applicable, reference may be made to the description in S310 above, which is not described in detail here for brevity.
  • the second communication apparatus sends a second message, where the second message is used to indicate one or more configuration parameters for updating the first channel learning model.
  • the first communication apparatus receives the second message, determines that the first channel learning model is not applicable according to the second message, and updates the first channel learning model.
  • the second message may also include one or more configuration parameters for updating the second channel learning model.
  • one or more configuration parameters for updating the second channel learning model may also be sent through other messages, and the other messages are different from the second message, which is not specifically limited in this embodiment of the present application.
  • the second communication device may determine a new first channel learning model and a new second channel learning model, and configure one or more configurations of the new first channel learning model The parameters are sent to the first communication device for updating the first channel learning model.
  • the second communication device may send one or more configuration parameters of the new first channel learning model that are different from the previous first channel learning model to the first communication device, so as to be used for updating the previous one.
  • the first channel learning model may be determined a new first channel learning model and a new second channel learning model, and configure one or more configurations of the new first channel learning model The parameters are sent to the first communication device for updating the first channel learning model.
  • the second communication device may send one or more configuration parameters of the new first channel learning model that are different from the previous first channel learning model to the first communication device, so as to be used for updating the previous one.
  • the first channel learning model may be used for updating the previous one.
  • the second communication device may determine a new first channel learning model, and send one or more configuration parameters of the new first channel learning model to the first channel learning model.
  • two communication devices for updating the first channel learning model may determine a new first channel learning model, and send one or more configuration parameters of the new first channel learning model to the first channel learning model.
  • the second message includes one or more configuration parameters for updating the first channel learning model.
  • the first communication apparatus may update the first channel learning model according to the second message.
  • the one or more configuration parameters may include at least one of the following: model type, model structure, model algorithm, model weight vector, model weight matrix, model bias vector, model bias matrix, model activation function.
  • the types of models include machine learning algorithms, neural network models or auto-encoding models.
  • the configuration parameter may include a model algorithm for indicating which one of the machine learning algorithms the channel learning model is specifically.
  • the configuration parameters may include a model structure, which is used to indicate which type of neural network the channel learning model is, and the model structure may also include the dimension of the input layer and the dimension of the output layer.
  • the configuration parameters may include one or more of a transformation algorithm, a weight matrix, a weight vector, a bias vector, a bias matrix, and an activation function.
  • the configuration parameter may include a model structure for indicating the number of layers and/or the sequence of layers.
  • the configuration parameters can also include one or more of the following: parameters for the data input layer (preprocessing operation algorithm, dimension of input data, value range of input data), parameters for the convolution layer (size of input unit, Receptive field, stride, number of zero padding, depth, depth of output unit, weight matrix), parameters for excitation layer (activation function), parameters for pooling layer (pooling algorithm, spatial range, stride, The size of the input unit, the size of the output unit), the parameters for the fully connected layer (weight matrix, weight vector, bias matrix, bias vector).
  • This embodiment of the present application does not limit the manner in which the second communication apparatus determines the first channel learning model and the second channel learning model.
  • the following embodiments provide a manner for the second communication apparatus to determine the first channel learning model and/or the second channel learning model, and the manner for the second communication apparatus to determine the channel learning model may be used as an independent embodiment, or may be related to other embodiments. In combination, specifically, this embodiment of the present application does not limit this. One or more of the following ways to determine the channel learning model may be used alone or in combination, which is not specifically limited in this embodiment of the present application.
  • the second communication apparatus may determine a new first channel learning model and a new second channel learning model according to the first parameter.
  • the second communication apparatus may determine the first channel learning model and the second channel learning model from the configuration parameter set saved in the database #2 according to the first parameter.
  • the first parameter includes at least one of the following: a cell identifier of a cell where the terminal device is located, a scene where the terminal device is located, a type of the terminal device, and a geographic location where the terminal device is located.
  • the second communication apparatus may determine the first parameter according to the cell where it is located, the scene where it is located, the type or the geographic location where it is located.
  • the second communication device when the second communication device is a network device and the first communication device is a terminal device, the second communication device may receive information related to the first parameter sent by the first communication device, and determine the first parameter.
  • the scene in which the terminal device is located may be indoors, outdoors, suburbs, towns, external environments (eg, daytime, nighttime, sunny day, cloudy day, smooth traffic, traffic jam), and the like.
  • the second communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • One or more configuration parameters for the model may be used to determine the first channel learning model and/or the second channel learning model corresponding to the indoor scene, and further determine the first channel learning model and/or the second channel learning model for updating the first channel learning model and/or the second channel learning model.
  • the type of the terminal device may be the number of antenna ports, processing capability, etc. of the terminal device.
  • the geographic location where the terminal device is located may be three-dimensional coordinates, two-dimensional coordinates, positioning data, and the like. As shown in FIG. 11 , if the geographical location where the terminal device is located is area 1, the second communication apparatus may determine the first channel learning model and the second channel learning model corresponding to area 1, and determine the first channel learning model for updating the first channel learning model One or more configuration parameters of the model and/or the second channel learning model; if the geographic location where the terminal device is located is area 2, the second communication apparatus may determine the first channel learning model and/or the second channel learning model corresponding to area 2 a channel learning model, and determining one or more configuration parameters for updating the first channel learning model and/or the second channel learning model.
  • one or more configuration parameters sent by the second communication apparatus to the first communication apparatus for updating the first channel learning model and/or the second channel learning model It can be terminal level, cell level or terminal group level.
  • the second communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to each first communication device in a unicast manner;
  • the second communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to a plurality of first communication devices in the cell in a broadcast manner;
  • the second communication device sends one or more configuration parameters for updating the first channel learning model and/or the second channel learning model to a plurality of first communication devices in the terminal group in a broadcast or multicast manner .
  • the database #2 may be a database carrying relevant information or related data of the channel learning model, for example, the database #2 may store the configuration parameters of the channel learning model and the like.
  • the database #2 may be a local database of the second communication device, or may be a database stored at a higher level.
  • the database stored in a high layer may be stored in a mobility management unit, a core network, a cloud, a central manager, an operator system, a second communication device group management system, and data.
  • the second communication device may communicate and interact with the upper layer, and determine the configuration parameters of the channel learning model from the database of the higher layer.
  • the second communication device may download and/or read the channel learning model stored in the database from the high-level database according to the first parameter, and then determine a new first channel learning model and a new second channel learning model.
  • the second communication apparatus may receive configuration parameters about the channel learning model sent by the high-level network element, and then determine a new first channel learning model and a new second channel learning model.
  • the second communication device can download and/or read the channel learning model in the database from the local database, and then determine the first channel learning model and the second channel learning model. Model.
  • the configuration parameter library of the channel learning model stored in the local database of the second communication device is pre-configured, or it may be determined by the second communication device through training and learning and stored in the local database.
  • the second communication device may determine the channel learning model from the local database according to the first parameter before applying the channel learning model.
  • the second communication device may determine the first channel learning model and/or the second channel learning model according to the first parameter, and further determine the configuration parameters for updating the first channel learning model, that is, the channel learning model is not performed.
  • a new channel learning model can be determined after training, which reduces the processing complexity of the second communication device.
  • the second communication apparatus may perform training on the first channel learning model and the second channel learning model to obtain the first channel learning model and the second channel learning model, and further determine to update the first channel learning model one or more configuration parameters.
  • the manner in which the second communication apparatus trains the first channel learning model and the second channel learning model may be:
  • the above gradient information is sent to the first channel learning model as one or more configuration parameters for updating the first channel learning model.
  • the second communication device continues to train the first channel learning model and the second channel learning model based on the above method until the value of the loss function is less than the preset threshold. Further, the multiple gradient information obtained in multiple training processes is accumulated and sent to the second communication apparatus as one or more configuration parameters for updating the first channel learning model.
  • the first channel information may be sent by the first communication apparatus to the second communication apparatus.
  • the target channel information may be sent by the first communication apparatus to the second communication apparatus, or may be obtained by the second communication apparatus according to the reciprocity of the uplink and downlink channels.
  • the training parameters for training the channel learning model by the first communication apparatus may be pre-configured, or may be indicated by the first communication apparatus.
  • the second communication apparatus may determine each of the multiple configuration parameters by using the same method, or may use a different method to determine the multiple configuration parameters. each of the parameters.
  • the second communication apparatus may determine the configuration parameters for updating the channel learning model in a one-level or multi-level manner.
  • the second communication device may use one or more of the above methods to determine the configuration parameters for updating the channel learning model; taking the multi-stage method as an example, the second communication device may first use the above method.
  • One or more of the above-mentioned methods determine a part of configuration parameters for updating the channel learning model, and then use one or more of the above methods to determine another part of the configuration parameters for updating the channel learning model.
  • the second communication device determines the structure of the channel learning model from the high-level database according to the first parameter, then determines the dimension, operation and/or function of the channel learning model from the local database according to the second parameter, and finally Information about the configuration parameters of the communication device determines the variables of the channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S410' to S420':
  • the first communication apparatus receives a second message, where the second message is used to indicate one or more configuration parameters for updating the first channel learning model.
  • the second message is used to indicate one or more configuration parameters for updating the first channel learning model.
  • the first communication apparatus determines that the first channel learning model is not applicable according to the second message, and updates the first channel learning model.
  • S420' that is, S420
  • the first communication apparatus determines that the first channel learning model is not applicable according to the second message, and updates the first channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S410" to S420":
  • S410 the second communication apparatus determines whether the second channel learning model is applicable. For a specific description of this step, reference may be made to the description in S410 above, which is not described in detail here for brevity.
  • S420 that is, S420
  • the second communication device sends a second message, where the second information is used to indicate one or more configuration parameters used to update the first channel learning model.
  • the second information is used to indicate one or more configuration parameters used to update the first channel learning model.
  • an embodiment of the present application provides a communication method.
  • the following embodiments may be used as independent embodiments, and may also be combined with other embodiments in this application, which are not specifically limited in this application.
  • an embodiment of the present application provides a communication method performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S410* to S420*:
  • the first communication apparatus receives first indication information, where the first indication information is used to instruct to perform channel learning model training.
  • first indication information is used to instruct to perform channel learning model training.
  • the first communication apparatus sends a first message, where the first message is used to indicate one or more configuration parameters for updating the channel learning model.
  • the first message is used to indicate one or more configuration parameters for updating the channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S410** to S430**:
  • S410** (ie, S410), the second communication apparatus determines whether the second channel learning model is applicable.
  • S410 the description in S410 above, and for brevity, it will not be described in detail here.
  • the second communication apparatus sends first indication information, where the first indication information is used to instruct to perform channel learning model training.
  • first indication information is used to instruct to perform channel learning model training.
  • the second communication apparatus receives a first message, where the first message is used to indicate one or more configuration parameters for updating the second channel learning model.
  • the first message is used to indicate one or more configuration parameters for updating the second channel learning model.
  • the embodiment of the present application provides a communication method in which the second communication device determines that the channel learning model is not applicable and notifies the first communication device. This method can realize the timely reporting to the first communication device when the channel learning model is not applicable, thereby realizing the The update of the channel learning model can improve the accuracy and communication performance of the channel learning model.
  • FIG. 14 shows a schematic flowchart of a communication method provided by an embodiment of the present application.
  • the embodiments of the present application provide a schematic flowchart of a method for determining channel information in communication.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • the first communication device is a terminal device and the second communication device is a network device as an example for description.
  • the method 500 may include S510 to S550, and each step will be described in detail below.
  • the terminal device and the network device determine the first channel learning model and the second channel learning model.
  • this step includes the terminal device determining the first channel learning model, or the terminal device determining the first channel learning model and the second channel learning model.
  • this step includes the network device determining the second channel learning model, or the network device determining the first channel learning model and the second channel learning model.
  • the first channel learning model may be deployed on the terminal device side, and is used to determine the first channel information according to the measured downlink channel information, and the data volume of the first channel information is smaller than the measured data volume of the downlink channel information. Deployment may also mean that in practical applications, the first channel learning model may be stored on the terminal device side, and the terminal device side may perform information processing or calculation based on the first channel learning model.
  • the second channel learning model may be deployed on the network device side to determine the second channel information according to the first channel information, and the data amount of the second channel information is the same as or similar to the measured data amount of the downlink channel information.
  • the second channel information may be used for downlink transmission.
  • Deployment may also mean that in practical applications, the second channel learning model may be stored on the network device side, and the network device side may perform information processing or calculation based on the second channel learning model.
  • Determining the channel learning model may include determining configuration parameters in the channel learning model.
  • the configuration parameters in the channel learning model include one or more of the structure of the model, the dimension of the model, operations in the model, functions in the model, variables in the model, parameters in the model, and the like.
  • the structure of the model may include one or more of the structure of the model being a neural network, principal component analysis, auto-encoding, and the like.
  • the structure of the model may also include features of the input information of the model, features of the output information of the model, and the like.
  • the features of the input information of the model may refer to the features of the input information after operations are performed on high-dimensional channel information (eg, measured downlink channel information).
  • the operation performed on the high-dimensional channel information may be one or more of de-averaging, normalization, discrete Fourier transform, time-delay-angle domain transform, or separation of real and imaginary parts.
  • the input information may be the information of the combination of the real part and the imaginary part of the high-dimensional channel information, the independent real part or the independent imaginary part of the high-dimensional channel information, or the delay-angle domain information and the like.
  • Operating on high-dimensional channel information can also be referred to as data preprocessing.
  • the characteristics of the input information to determine the model are normalization, time domain-angle domain transformation, and independent real and imaginary parts, etc.
  • the features of the output information of the model may refer to information features after encoding or dimension reduction.
  • the information after encoding or dimensionality reduction may refer to the information after normalization, the information after discrete Fourier transform, the information after time-delay-angle domain transform, the information of the combination of real and imaginary parts, or the information of real and imaginary parts. Information after the partial and imaginary parts are separated.
  • the dimension of the model can be one or more of the number of layers in the model, the dimension of each layer, the dimension of the input layer/hidden layer/output layer, and so on.
  • the dimension of the model is determined to be an N-layer convolutional neural network, the dimension of the input layer is N 1 , and the dimension of the output layer is N 2 .
  • N, N 1 , and N 2 are positive integers.
  • the operations in the model can be one or more of linear operations, nonlinear operations, and so on. For example, determine the operation of the model as a linear operation.
  • the operation can also be measured or characterized in terms of complexity.
  • the functions in the model can be mathematical operations, logical operations, etc., such as one or more of addition, subtraction, multiplication and division, weighted summation, weighted summation plus bias, activation function, and the like.
  • the functions that determine the model are the weighted sum plus bias and the Relu activation function.
  • the activation function can be determined separately for each layer in the model, or the same activation function can be used for one or more layers.
  • the variables in the model may refer to information about parameters involved in the model, for example, may be one or more of the number of parameters, the value range of the parameters, the values of the parameters, and the types of the parameters. Parameters can be constants or variables. For example, determine the weight matrix of the model as W, the bias matrix as b, the variable value of the activation function ⁇ , etc.
  • the first channel learning model may include one or more models, and correspondingly, the second channel learning model may also include one or more models.
  • the first channel learning model may include one or more models
  • the second channel learning model may also include one or more models.
  • the real part and imaginary part of the high-dimensional channel information can respectively correspond to a convolutional neural network, that is, the input information corresponding to the convolutional neural network 1 is the high-dimensional channel information.
  • the real part, the encoded output information is the real part of the low-dimensional channel information (for example, the first channel information)
  • the input information corresponding to the convolutional neural network 2 is the imaginary part of the high-dimensional channel information
  • the encoded output information is the low-dimensional channel information. imaginary part.
  • the configuration parameters of the two neural networks may be the same or different.
  • the real part and the imaginary part of the low-dimensional channel information may correspond to a convolutional neural network respectively, that is, the input information corresponding to the convolutional neural network 3 is the low-dimensional channel information.
  • the decoding output information is the real part of the high-dimensional channel information (for example, the second channel information)
  • the input information corresponding to the convolutional neural network 4 is the imaginary part of the low-dimensional channel information
  • the decoding output information is the high-dimensional channel information.
  • the imaginary part of the channel information When the real part and the imaginary part of the low-dimensional channel information correspond to two neural networks, the configuration parameters of the two neural networks may be the same or different.
  • Determining the channel learning model may be determining configuration parameters of one or more channel learning models.
  • the network device determines the first channel learning model and the second channel learning model, and sends the configuration parameters of the first channel learning model to the terminal device.
  • the network device determines the first channel learning model and the second channel learning model, and sends the configuration parameters of the first channel learning model to the terminal device.
  • the network device determines the first channel learning model and the second channel learning model, reference may be made to the description in S420 above about the second communication apparatus determining the new first channel learning model and the new second channel learning model, For the sake of brevity, details are not described here.
  • the terminal device determines the first channel learning model and the second channel learning model, and sends configuration parameters of the second channel learning model to the network device.
  • the terminal device determines the first channel learning model and the second channel learning model, and sends configuration parameters of the second channel learning model to the network device.
  • the method 500 may further include: the terminal device receives configuration information #1 sent by the network device, where the configuration information #1 is used to indicate sending the configuration of the first channel learning model and/or the second channel learning model One or more of the parameter resources, code rate, modulation method, number of bits, feedback order, and feedback configuration parameter content.
  • the terminal device can determine, according to the configuration information #1, the resources, code rate, modulation method, number of bits or feedback order for sending the configuration parameters of the first channel learning model and/or the second channel learning model.
  • the network device may send the configuration information #1 through high layer signaling, or may send the configuration information #1 through physical layer signaling.
  • the high layer signaling may refer to RRC signaling or MAC signaling
  • the physical layer signaling refers to DCI signaling.
  • the method 500 may further include: the terminal device receives configuration information #2 sent by the network device, where the configuration information #2 is used to indicate receiving the configuration of the first channel learning model and/or the second channel learning model One or more of the parameter resources, code rate, modulation method, number of bits, feedback order, and feedback configuration parameter content.
  • the terminal device can determine, according to the configuration information #2, the resource, code rate, modulation method, number of bits or feedback sequence for receiving the configuration parameters of the first channel learning model and/or the second channel learning model.
  • the network device may send the configuration information #2 through high layer signaling, or may send the configuration information #2 through physical layer signaling.
  • the high layer signaling may refer to RRC signaling or MAC signaling
  • the physical layer signaling refers to DCI signaling.
  • the terminal device determines a first channel learning model
  • the network device determines a second channel learning model. It should be understood that when the terminal device and the network device each determine the channel learning model based on the same rule, the first channel learning model determined by the terminal device and the second channel learning model determined by the network device are corresponding. For example, the first channel learning model determined by the terminal device according to the scene in which the terminal device is located corresponds to the second channel learning model determined by the network device according to the scene in which the terminal device is located.
  • the first channel learning model may be an encoding module in the corresponding auto-encoding model
  • the second channel learning model may be a decoding module in the corresponding auto-encoding model.
  • the first channel learning model and the second channel learning model are corresponding.
  • the network device sends a reference signal.
  • the terminal device receives the reference signal.
  • S520 may be performed before S510, or S520 may be performed after S510, which is not limited in this application.
  • the reference signal may be DMRS, CSI-RS, TRS, SSB, etc., which is not limited in this application.
  • the terminal device may receive CSI-RS configuration information, and may receive CSI-RS periodically, semi-persistently, or aperiodically according to the configuration information.
  • the CSI-RS configuration information may refer to CSI measurement configuration information, and the CSI configuration information may include at least one of CSI resource configuration information (CSI-ResourceConfig) and CSI reporting configuration information (CSI-ReportConfig).
  • the terminal device receives the CSI-RS configuration information sent by the network device, and may further include the configuration parameters of the channel learning model that the terminal device receives and sent by the network device.
  • the configuration parameters of the channel learning model may be included in the CSI-RS configuration information.
  • the configuration parameters of the channel learning model may be included in the CSI reporting configuration information.
  • the terminal device may determine the channel learning model based on the configuration parameters of the channel learning model sent by the network device.
  • the network device may send the configuration parameters of the channel learning model through high layer signaling and/or physical layer signaling.
  • the sequence of S520 and S510 is not limited in this application.
  • the channel learning model may be determined first and then the reference signal is sent, or the reference signal may be sent first and then the channel learning model is determined.
  • the channel learning model may be determined first and then the reference signal is received, or the reference signal may be received first and then the channel learning model may be determined.
  • the terminal device determines the first channel information based on the target channel information and the first channel learning model.
  • the target channel information may be downlink channel information obtained by the terminal device based on the received reference signal.
  • the target channel information reference may be made to the description in the present application. For brevity, the embodiment of the present application will not repeat them.
  • the terminal device uses the measured downlink channel information as input data of the first channel learning model, and encodes or compresses the downlink channel information according to the first channel learning model to obtain the first channel information.
  • the terminal device can encode the real part of the downlink channel information to obtain the real part of the low-dimensional channel information, and then encode the imaginary part of the downlink channel information to obtain the low-dimensional channel information The imaginary part of the information.
  • the first channel information may include a real part of the low-dimensional channel information and an imaginary part of the low-dimensional channel information.
  • the terminal device After the terminal device encodes the downlink channel information of a 64*1-dimensional complex number according to the first channel learning model, it can obtain 2-dimensional real part information and 2-dimensional imaginary part information, then the terminal device can convert the 2-dimensional real part information The information and the 2-dimensional imaginary part information are quantized as the first channel information.
  • the terminal device may jointly encode the real part and the imaginary part of the downlink channel information to obtain low-dimensional channel information.
  • the first channel information may include low-dimensional real number information.
  • the terminal device encodes downlink channel information of a 64*1-dimensional complex number according to the first channel learning model
  • 4-dimensional real number information can be obtained, and then the terminal device can quantize the 4-dimensional real number information as the first channel information .
  • the terminal device sends the first channel information.
  • the network device receives the first channel information.
  • the method 500 may further include: the terminal device receives configuration information #3 sent by the network device, where the configuration information #3 is used to indicate the resource, code rate, modulation method, and number of bits for sending the first channel information , one or more of the feedback sequence, the content of the feedback configuration parameter, the content of the first channel information, and the like.
  • the terminal device can determine, according to the configuration information #2, the resources, code rate, modulation method, number of bits, feedback order, feedback configuration parameter content or content of the first channel information for sending the first channel information. That is, the method 500 includes: the network device sends the configuration information #3, and the terminal device receives the configuration information #3.
  • the network device may send the configuration information #3 through high layer signaling, or may send the configuration information #3 through physical layer signaling.
  • the high layer signaling may refer to RRC signaling or MAC signaling
  • the physical layer signaling refers to DCI signaling.
  • the terminal device may send the configuration parameters of the second channel learning model together with the first channel information, or may send the configuration parameters of the second channel learning model and the first channel information together.
  • a channel letter is sent separately.
  • the terminal device sends the configuration parameters of the second channel learning model on channel #1 and the first channel information on channel #2; or the terminal device sends the configuration parameters of the second channel learning model and the first channel information on channel #3 .
  • channel #1 to channel #3 may be PUCCH, PUSCH, or physical feedback channel (physical feedback channel, PFCH), or may be other uplink channels.
  • the terminal device may combine the configuration parameters of the first channel learning model and the configuration parameters of the second channel learning model with the first channel information.
  • the configuration parameters of the first channel learning model, the configuration parameters of the second channel learning model and the first channel information can also be sent separately.
  • the terminal device sends the configuration parameters of the first channel learning model and the configuration parameters of the second channel learning model on channel #1, and sends the first channel information on channel #2; or, the terminal device sends the first channel learning model on channel #3 Configuration parameters of the model, configuration parameters of the second channel learning model, and first channel information.
  • channel #1 to channel #3 may be PUCCH, PUSCH or PFCH, or may be other uplink channels.
  • the method 500 may further include: the terminal device receives configuration information #4 sent by the network device, where the configuration information #4 is used to indicate the resources and codes for sending the first channel information and the configuration parameters of the channel learning model.
  • the configuration information #4 is used to indicate the resources and codes for sending the first channel information and the configuration parameters of the channel learning model.
  • the terminal device can determine the resources, code rate, modulation mode, number of bits, feedback order or content of the first channel information for sending the first channel information and the configuration parameters of the channel learning model according to the configuration information #4. That is, the method 500 includes: the network device sends the configuration information #4, and the terminal device receives the configuration information #4.
  • the network device may send the configuration information #4 through high layer signaling, or may send the configuration information #4 through physical layer signaling.
  • the high layer signaling may refer to RRC signaling or MAC signaling
  • the physical layer signaling refers to DCI signaling.
  • the method 500 may further include: the terminal device receives configuration information #5 sent by the network device, where the configuration information #5 is used to indicate the resources, code rate, modulation mode, One or more of the number of bits, feedback order, feedback configuration parameter content, etc.
  • the terminal device can determine the resources, code rate, modulation mode, number of bits, feedback order, and feedback configuration parameter content of the configuration parameters of the transmission channel learning model. That is, the method 500 includes: the network device sends the configuration information #5, and the terminal device receives the configuration information #5.
  • the network device may send the configuration information #5 through high layer signaling, or may send the configuration information #5 through physical layer signaling.
  • the high layer signaling may refer to RRC signaling or MAC signaling
  • the physical layer signaling refers to DCI signaling.
  • the network device determines the second channel information according to the first channel information and the second channel learning model.
  • the data volume of the second channel information is the same as or similar to the data volume of the downlink channel information of the measured downlink channel information, that is, the network device decodes or decodes the first channel information according to the second channel learning model. Decompression obtains the second channel information.
  • the network device can decode or decompress the real part information of the low-dimensional channel information in the first channel information to obtain high-dimensional real part information,
  • the imaginary part information of the dimensional channel information is decoded or decompressed to obtain high-dimensional imaginary part information.
  • the second channel information may include high-dimensional real part information and high-dimensional imaginary part information. That is, the second channel information may be high-dimensional complex channel information.
  • the network device decodes or decompresses the 2-dimensional real part information, it can obtain 64*1-dimensional high-dimensional real part information, and after decoding or decompresses the 2-dimensional imaginary part information, it can obtain 64*1-dimensional high-dimensional information imaginary part information, the network device may use the 64*1-dimensional real part information and the 64*1-dimensional imaginary part information as the second channel information.
  • the network device can jointly decode or decompress the real part and imaginary part of the low-dimensional channel information to obtain high-dimensional real number information. That is, the second channel information may include high-dimensional real number information.
  • the network device decodes or decompresses the low-dimensional channel information of 4-dimensional real numbers, it can obtain 64*2-dimensional real number information.
  • the network device may use the 64*2-dimensional real number information as the second channel information.
  • the network device can perform precoding according to the second channel information, and perform data transmission according to the determined precoding, which can improve the accuracy of the precoding and reduce inter-cell and/or multiple terminal devices. signal interference between them, thereby improving communication performance.
  • the network device determines the second channel information, it can determine data scheduling information according to the second channel information, and perform data transmission according to the determined scheduling information, which can improve the accuracy of data transmission and thus improve communication performance.
  • an embodiment of the present application provides a communication method that can be performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S510' to S540':
  • S510' the first communication apparatus determines a channel learning model.
  • S510 the first communication apparatus determines a channel learning model.
  • S520' (ie, S520), the first communication apparatus receives the reference signal.
  • S520 the first communication apparatus receives the reference signal.
  • the sequence of S510' and S520' may also be to execute S520' first, and then execute S510', which is not limited in this application.
  • the first communication apparatus determines the first channel information based on the target channel information and the first channel learning model.
  • S530 the first communication apparatus determines the first channel information based on the target channel information and the first channel learning model.
  • S540' that is, S540
  • the first communication apparatus sends the first channel information.
  • S540 the description in S540 above, which is not described in detail here for brevity.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S510" to S540":
  • S510 the second communication apparatus determines the channel learning model.
  • S510 the second communication apparatus determines the channel learning model.
  • S520 (ie S520), the second communication device sends a reference signal.
  • S520 the second communication device sends a reference signal.
  • the sequence of S510" and S520" can also be S520' is executed first, and then S510' is executed, which is not limited in this application.
  • S530' ie, S540
  • the second communication apparatus receives the first channel information.
  • S540 the description in S540 above, which is not described in detail here for brevity.
  • the second communication apparatus determines the second channel information according to the first channel information and the second channel learning model.
  • S540' ie, S550
  • the second communication apparatus determines the second channel information according to the first channel information and the second channel learning model.
  • the terminal device determines the first channel information according to the first channel learning model, and sends the first channel information to the network device, which can reduce the overhead of channel information feedback and improve the communication performance.
  • the network device can determine the second channel information according to the received first channel information and the second channel learning model, and further, determine the precoding and scheduling information of data transmission according to the second channel information, which can improve the accuracy of channel acquisition, That is, to improve the accuracy of precoding, it is beneficial to reduce the interference between cells or multiple terminals, improve the accuracy of data transmission, and improve the communication performance.
  • the following embodiments of the present application provide a communication method.
  • This embodiment provides a training and/or feedback and/or update method for determining a channel learning model according to the capability of a terminal device.
  • the following methods can be used as independent embodiments, and can also be combined with other embodiments in this application, which are not specifically limited in this application.
  • the channel learning model may not be suitable for the current channel conditions as time changes. At this time, the network device and/or the terminal device The channel learning model needs to be trained and/or fed back and/or updated, etc.
  • the capability of the terminal device may refer to at least one of the following terminal types: active learning terminal, network device indicating terminal, passive receiving terminal, and the like.
  • the active learning terminal refers to a terminal device that can actively perform channel learning model training and send channel learning feedback signaling to the network device.
  • a network device indicating terminal refers to a terminal device that can perform channel learning model training and send channel learning feedback signaling to the network device according to the channel learning training instruction sent by the network device.
  • a passive receiving terminal refers to a terminal device that is not capable of channel learning model training and can determine the channel learning model by receiving signaling of configuration parameters of the channel learning model sent by the network device.
  • the channel learning feedback signaling may include the result of channel learning model training, for example, whether the first channel learning model is applicable, and/or, for updating the configuration parameters of the first channel learning model, and/or for updating Configuration parameters of the second channel learning model.
  • the first message in S210 in this application which is not described in detail here for brevity.
  • the capability of the terminal device may refer to at least one of the following types of terminals: the terminal device supports channel learning model training, the terminal device does not support channel learning model training, the terminal device enables channel learning model training, the terminal device does not enable channel learning model training Channel learning model training, etc.
  • the terminal device can send information about the capabilities of the terminal device to the network device, that is, the terminal device can report the capabilities of the terminal device to the network device. For example, the terminal device can report whether it supports channel learning model training, or report the terminal type.
  • the method of determining the channel learning model according to the capabilities of the terminal equipment can be more suitable for the situation that there are multiple terminal equipments in the actual communication system, and different methods are used to determine and/or train and/or determine and/or train and/or use different terminal equipments according to the capabilities of different terminal equipments.
  • Feedback and/or update of the channel learning model can improve the flexibility of the communication system, take into account terminal devices with different capabilities, and improve communication performance.
  • an active learning terminal can communicate with a network device according to at least one step in the method of FIG. 16
  • an indicating terminal can communicate with a network device according to at least one step in the method of FIG. 15
  • a passive reception The terminal-like terminal may communicate with the network device according to at least one step in the method of FIG. 17 .
  • a terminal device that does not support (or does not enable) channel learning model training may communicate with a network device according to at least one step in the method of FIG. 15 or FIG. 16
  • a terminal device that supports (or enables) channel learning model training The terminal device may communicate with the network device according to at least one step in the method of FIG. 17 .
  • the first communication apparatus and/or the second communication apparatus may first determine the capability of the terminal device.
  • the specific determination method can be as described above. That is, it may include the steps of the first communication apparatus sending information about the capability of the terminal equipment, and the second communication apparatus receiving the information about the capability of the terminal equipment.
  • FIG. 15 shows a schematic flowchart of a communication method provided by an embodiment of the present application.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • the first communication device is a terminal device and the second communication device is a network device as an example for description.
  • the method 600 may include at least one of S610 to S660, and each step will be described in detail below.
  • the network device sends channel learning training signaling (an example of the first indication information).
  • the terminal device receives the channel learning training signaling.
  • the channel learning training signaling is used to instruct the terminal device to perform channel learning model training, that is, to instruct the terminal device to determine whether the first channel learning model is applicable, and/or to instruct the terminal device to determine the channel learning model.
  • the channel learning training signaling may further include configuration parameters of the first channel learning model.
  • the first channel learning model is used to determine the first channel information according to the measured downlink channel information, and the data amount of the first channel information is smaller than the measured data amount of the downlink channel information.
  • the channel learning training signaling may further include configuration parameters of the second channel learning model.
  • the second channel learning model is used to determine the second channel information according to the first channel information.
  • the channel learning training signaling may further include one or more of the time of the channel learning model training, the configuration information of the reference signal for the channel learning model training, and the information related to the feedback content of the channel learning training.
  • the configuration parameters of the first channel learning model, the configuration parameters of the second channel learning model, the time of the channel learning model training, the configuration information of the reference signal of the channel learning model training, and the information related to the feedback content of the channel learning model training can be sent to the terminal device through different signaling.
  • the channel learning training signaling may be sent through higher layer signaling, such as MAC layer signaling, RRC layer signaling, or may be sent through physical layer signaling, such as DCI signaling.
  • higher layer signaling such as MAC layer signaling, RRC layer signaling
  • RRC layer signaling or may be sent through physical layer signaling, such as DCI signaling.
  • the terminal device performs channel learning model training according to the channel learning training signaling.
  • the terminal device can perform channel learning model training according to the received channel learning training signaling, and determine whether the first channel learning model is applicable or whether it needs to be updated. For the method for the terminal device to determine whether the first channel learning model is applicable, reference may be made to the description in S210 above.
  • the terminal device may further determine a new first channel learning model and a new second channel learning model.
  • the terminal device may further determine a new first channel learning model and a new second channel learning model.
  • the terminal device sends the channel learning feedback signaling (an example of the first message).
  • the network device receives the channel learning feedback signaling.
  • the channel learning feedback signaling may include the result of channel learning model training, for example, whether the first channel learning model is applicable, and/or, for updating the configuration parameters of the first channel learning model, and/or for updating the second channel learning model. Configuration parameters for the channel learning model.
  • the terminal device may send a request for sending the channel learning feedback signaling to the network device.
  • the sending request is used to instruct the terminal device to request the terminal device to send channel learning feedback signaling.
  • the sending request may be similar to a scheduling request (SR) in the prior art.
  • the network device may send scheduling information, where the scheduling information instructs the terminal device to send channel learning feedback signaling.
  • the scheduling information may include related configuration information of the channel learning feedback signaling, such as one or more of time-frequency resources, feedback content, feedback code rate, feedback bit number, and the like.
  • the terminal device and the network device are preconfigured with several groups of second channel learning models, and are preconfigured with model identifiers corresponding to the second channel learning model, the terminal device is feeding back the information used to update the second channel learning model.
  • the terminal device and the network device are preconfigured with 4 groups of channel learning models, and the 4 groups of channel learning models correspond to model ID 1, model ID 2, model ID 3, and model ID 4 respectively.
  • the terminal device carries the model identifier 3 in the channel learning feedback signaling and sends it to the network device.
  • the network device can determine the second channel learning model applicable to the terminal device according to the model identifier 3 .
  • the terminal device and the network device are preconfigured with several groups of first channel learning models, and are preconfigured with model identifiers corresponding to the first channel learning models, the terminal device is feeding back the information used to update the first channel learning model.
  • the terminal device and the network device are preconfigured with 4 groups of channel learning models, and the 4 groups of channel learning models correspond to model ID 1, model ID 2, model ID 3, and model ID 4 respectively.
  • the terminal device carries the model identifier 3 in the channel learning feedback signaling and sends it to the network device.
  • the network device can determine the first channel learning model applicable to the terminal device according to the model identifier 3 .
  • the terminal device and the network device are preconfigured with several groups of channel learning models (including the first channel learning model and the second channel learning model), and the model identifiers corresponding to the channel learning models are preconfigured, the terminal device is When the feedback is used to update the configuration parameters of the first channel learning model and the second channel learning model, only the identifier of the determined channel learning model may be fed back.
  • the terminal device and the network device are preconfigured with 4 groups of channel learning models, and the 4 groups of channel learning models correspond to model ID 1, model ID 2, model ID 3, and model ID 4 respectively.
  • the terminal device carries the model identifier 3 in the channel learning feedback signaling and sends it to the network device.
  • the network device can determine the channel learning model applicable to the terminal device according to the model identifier 3 .
  • the channel learning learning model is determined by feeding back the channel learning model identification, which can reduce the number of feedback bits, reduce the feedback overhead, and improve the communication performance.
  • the method 600 may further include: the terminal device receives configuration information #6 sent by the network device, where the configuration information #6 is used to indicate resources, code rates, modulation methods, bits for sending channel learning feedback signaling One or more of the number, feedback order, feedback content, etc.
  • the terminal device can determine the resources, code rate, modulation mode, number of bits, feedback order or feedback content for sending the channel learning feedback signaling.
  • the network device may send the configuration information #6 through high layer signaling, or may send the configuration information #6 through physical layer signaling.
  • the high layer signaling may refer to RRC signaling or MAC signaling
  • the physical layer signaling refers to DCI signaling.
  • the channel learning feedback signaling may also be sent together with the first channel information, or may be sent separately from the first channel information.
  • the terminal device may send the channel learning feedback signaling and the first channel information through one signaling, or the terminal device may send the channel learning feedback signaling and the first channel information through multiple signalings.
  • the specific sending manner may be predefined, or may be indicated by a network device, which is not specifically limited here.
  • the terminal device may send and/or store the configuration parameters of the channel learning model to a database or other network elements.
  • a database For the specific description of the database, reference may be made to the description of the database in S220 above, which is not described in detail here for brevity.
  • the network device determines a channel learning model.
  • the network device may determine whether the current channel learning model is applicable, and/or determine the configuration parameters of the channel learning model.
  • the network device may determine whether the channel learning model is applicable by referring to the description in S410 above about the second communication apparatus determining whether the channel learning model is applicable, which is not described in detail here for brevity.
  • the network device may determine the channel learning model with reference to the description of the second communication apparatus determining the channel learning model in S420 above, which is not described in detail here for brevity.
  • the network device According to the channel learning feedback signaling, it can be determined that the current channel learning model is not applicable, and the current first channel learning model and/or the second channel learning model can be updated to determine a new second channel learning model.
  • the network device may determine that the current channel learning model is not applicable according to the channel learning feedback signaling. Further, the method 600 may further include S650 and S660.
  • S650 and S660 may also be used as independent embodiments, or may be combined with other embodiments, which are not specifically limited in this application.
  • This embodiment may provide a method for determining a channel learning model.
  • the network device sends configuration information of the first channel learning model.
  • the terminal device receives the configuration information of the first channel learning model.
  • the network device may determine a new first channel learning model and a new second channel learning model, and send the configuration parameters of the new first channel learning model to the terminal device to update the current channel learning model.
  • the first channel learning model That is, the network device may send the configuration parameter for updating the first channel learning model to the terminal device.
  • the network device may send/or store the configuration parameters of the new first channel learning model into database #2, and for the description of database #2, reference may be made to the description in S220 above.
  • the network device may send a channel learning message to the terminal device.
  • Configuration information for the model when the terminal device accesses the cell where the network device is located, or when the terminal device establishes an RRC connection with the network device, or when the terminal device performs cell handover, the network device may send a channel learning message to the terminal device. Configuration information for the model.
  • the terminal device determines a first channel learning model.
  • the terminal device updates the current first channel learning model according to the received configuration parameters for updating the first channel learning model to determine a new first channel learning model.
  • the terminal device may send/or store the configuration parameters of the new first channel learning model into database #1, and for the description of database #1, reference may be made to the description in S220 above.
  • the terminal device when the terminal device accesses the cell where the network device is located, or when the terminal device establishes an RRC connection with the network device, or when the terminal device performs cell handover, the terminal device can receive the channel sent by the network device.
  • the configuration information of the learning model the terminal device can determine the channel learning model according to the configuration parameters of the channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S610' to S640':
  • S610' that is, S610
  • the first communication apparatus receives the channel learning training signaling.
  • S610 the description in S610 above, which is not described in detail here for brevity.
  • S620' that is, S610
  • the first communication apparatus performs channel learning model training according to the channel learning training signaling.
  • S620 For a specific description of this step, reference may be made to the description in S620 above, which is not described in detail here for brevity.
  • S630' that is, S630
  • the first communication apparatus sends the channel learning feedback signaling.
  • S630 the description in S630 above, which is not described in detail here for brevity.
  • S640' (ie, S650), the first communication apparatus receives the configuration information of the channel learning model.
  • S650 For a specific description of this step, reference may be made to the description in S650 above, which is not described in detail here for brevity.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S610" to S640":
  • S610 (ie, S610), the second communication device sends the channel learning training signaling.
  • S610 the second communication device sends the channel learning training signaling.
  • S620 (ie, S630)
  • the second communication apparatus receives the channel learning feedback signaling.
  • S630 the description in S630 above, which is not described in detail here for brevity.
  • S630 (ie, S640)
  • the second communication device determines the channel learning model.
  • S640 For a specific description of this step, reference may be made to the description in S640 above, which is not described in detail here for brevity.
  • S640 (ie, S650)
  • the second communication apparatus sends the configuration information of the channel learning model.
  • S650 For a specific description of this step, reference may be made to the description in S650 above, which is not described in detail here for brevity.
  • the terminal device performs channel learning model training according to the channel learning training signaling of the network device, which may be applicable to the scenario where the network device determines that the channel learning model is not applicable.
  • the network device determines that the channel learning model is not applicable, it informs the terminal device in time to train the channel learning model, and determines the applicable channel learning model, so that the accuracy of the channel learning model can be improved, and the timely updating of the channel learning model can improve the communication performance.
  • FIG. 16 shows a schematic flowchart of a communication method provided by an embodiment of the present application.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • the first communication device is a terminal device and the second communication device is a network device as an example for description.
  • the method 700 may include at least one of S710 to S770, and each step will be described in detail below.
  • the terminal device sends a channel learning training request signaling (an example of a first request message) to the network device.
  • the network device receives the channel learning training request signaling sent by the terminal device.
  • the channel learning training request signaling is used to instruct the terminal device to request for channel learning model training, and/or to instruct the terminal device to request to feed back the channel learning training signaling.
  • the channel learning training request signaling may be sent in the SR resource.
  • the network device sends channel learning training signaling (an example of the first indication information).
  • the terminal device receives the channel learning training signaling.
  • the terminal device performs channel learning model training according to the channel learning training signaling.
  • the terminal device sends channel learning feedback signaling (an example of the first message).
  • the network device receives the channel learning feedback signaling.
  • the network device determines a channel learning model.
  • the network device sends configuration information of the first channel learning model.
  • the terminal device receives the configuration information of the first channel learning model.
  • the terminal device determines a first channel learning model.
  • FIG. 16 only takes S710 before S730 as an example, which should not limit the embodiments of the present application.
  • S710 may be performed between S730 and S740.
  • S710 may be performed after S730.
  • S710 may be executed before S740.
  • S720 to S770 may be the same as S610 to S660 in the method 600, and for the sake of brevity, the embodiments of the present application will not describe them in detail.
  • the method 700 may not perform S710 and S720.
  • the configuration information related to sending the channel learning feedback signaling may include: time-frequency resources occupied by the channel learning feedback information, content of the channel learning feedback signaling, feedback code rate of the channel learning feedback signaling, and channel learning feedback signaling The number of feedback bits.
  • an embodiment of the present application provides a method for communication by a first communication device. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S710' to S750':
  • S710' that is, S710
  • the first communication apparatus sends channel learning training request signaling.
  • the specific description of this step can be described in the above S710, which is not described in detail here for brevity.
  • S720' that is, S720
  • the first communication apparatus receives the channel learning training signaling.
  • S720' that is, S720
  • the first communication apparatus receives the channel learning training signaling.
  • the first communication apparatus performs channel learning model training according to the channel learning training signaling.
  • the description in S620 above which is not described in detail here for brevity.
  • S740' that is, S740
  • the first communication apparatus sends the channel learning feedback signaling.
  • S630 the description of S630 above, which is not described in detail here for brevity.
  • the first communication apparatus receives the configuration information of the channel learning model.
  • S750' that is, S760
  • the first communication apparatus receives the configuration information of the channel learning model.
  • the specific description of this step can be described in the above S650, which is not described in detail here for the sake of brevity.
  • an embodiment of the present application provides a method for communication by a second communication device. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S710" to S750":
  • S710 (ie, S710), the second communication apparatus receives the channel learning training request signaling.
  • S710 the description in S710 above, which is not described in detail here for brevity.
  • S720 (ie, S720), the second communication device sends the channel learning training signaling.
  • S720 the second communication device sends the channel learning training signaling.
  • S730 (ie, S740), the second communication apparatus receives the channel learning feedback signaling.
  • S630 the description in S630 above, which is not described in detail here for brevity.
  • the second communication device determines the channel learning model.
  • S740 ie, S750
  • the second communication device determines the channel learning model.
  • S750 (ie, S760)
  • the second communication device sends the configuration information of the channel learning model.
  • S650 description in S650 above, which is not described in detail here for brevity.
  • the terminal device may send a channel learning training request signaling to the network device to request channel learning model training, which may be suitable for a scenario where the terminal device determines that the channel learning model is not applicable.
  • the terminal device determines that the channel learning model is not applicable, it can notify the network device in time to request an instruction for channel learning model training, so as to determine the applicable channel learning model, improve the accuracy of the channel learning model, and update the channel learning model in time. , which can improve communication performance.
  • FIG. 17 shows a schematic flowchart of a communication method provided by an embodiment of the present application.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • the first communication device is a terminal device and the second communication device is a network device as an example for description.
  • the method 800 may include at least one of S810 to S830, and each step will be described in detail below.
  • the network device determines a channel learning model.
  • the network device may determine the channel learning model with reference to the description of the second communication apparatus determining the channel learning model in S410 above, which is not described in detail here for brevity.
  • the network device sends configuration information of the channel learning model.
  • the terminal device receives the configuration information of the channel learning model.
  • the configuration information of the channel learning model may be information including configuration parameters of the channel learning model.
  • the configuration information of the channel learning model sent by the network device reference may be made to the second message sent by the second communication device in S420 above, where the second message is used to indicate one or more configuration parameters used to update the first channel learning model.
  • the description for the sake of brevity, will not be described in detail here.
  • the terminal device determines the channel learning model according to the configuration information of the channel learning model.
  • an embodiment of the present application provides a communication method that can be performed by a first communication apparatus. That is, the following method is described from the perspective of the first communication device, and the method may include at least one of S810' to S820':
  • the first communication apparatus receives the configuration information of the channel learning model.
  • S810' ie, S820
  • the first communication apparatus receives the configuration information of the channel learning model.
  • the first communication apparatus determines the channel learning model according to the configuration information of the channel learning model. For a specific description of this step, reference may be made to the description in S660 above, which is not described in detail here for brevity.
  • an embodiment of the present application provides a communication method that can be performed by a second communication apparatus. That is, the following method is described from the perspective of the second communication device, and the method may include at least one of S810" to S820":
  • the terminal device when the terminal device accesses the cell where the network device is located, or when the terminal device establishes an RRC connection with the network device, or when the terminal device performs cell handover, the terminal device can receive the channel sent by the network device.
  • the configuration information of the learning model the terminal device can determine the channel learning model according to the configuration parameters of the channel learning model.
  • the terminal device determines the channel learning model according to the configuration information of the channel learning model sent by the network device, which may be suitable for scenarios where the terminal device does not have the ability to train the channel learning model, or may be suitable for example to reduce the processing complexity of the terminal device In the scenario, that is, the terminal device can determine the applicable channel learning model without training the channel learning model, which can improve the accuracy of the channel learning model, and update the channel learning model in time, which can improve the communication performance.
  • the following embodiments provide a method of communication.
  • This embodiment provides a method for determining channel learning feedback signaling according to the presence or absence of channel learning training signaling.
  • the embodiments of the present application may be used as independent embodiments, or may be combined with other embodiments in the present application, which are not specifically limited in the present application.
  • One or more of the following methods can be used alone or in combination, which is not specifically limited in this application.
  • the first communication apparatus may determine at least one item of the following information according to whether the channel learning training signaling is received:
  • the channel learning feedback signaling includes information on whether the channel learning model is applicable
  • the second communication apparatus may determine at least one of the following information according to whether to send channel learning training signaling:
  • the channel learning feedback signaling includes information on whether the channel learning model is applicable
  • the following description will be given by taking the first communication device being a terminal device and the second communication device being a network device as an example.
  • whether the channel learning feedback signaling includes information on whether the channel learning model is applicable may be determined according to whether the network device sends channel learning training signaling. That is, the network device determines whether the channel learning feedback signaling includes information on whether the channel learning model is applicable according to whether to send the channel learning training signaling.
  • whether the channel learning feedback signaling includes information on whether the channel learning model is applicable may be determined according to whether the terminal device receives the channel learning training signaling. That is, the terminal device determines whether the channel learning feedback signaling includes information on whether the channel learning model is applicable according to whether to send the channel learning training signaling.
  • Scenario 1 The network device sends channel learning training signaling to the terminal device.
  • the terminal device When the terminal device receives the channel learning training signaling sent by the network device, the terminal device can send the channel learning feedback signaling to the network device based on the target channel information and the channel learning model.
  • the channel learning feedback signaling may include information on whether the channel learning model is applicable, or the channel learning feedback signaling does not include information on whether the channel learning model is applicable.
  • the network device instructs the terminal device to perform channel learning model training to determine whether the channel learning model is applicable, so the network device sends channel learning training signaling.
  • the network device is not sure whether the channel learning model is applicable, so the channel learning feedback signaling sent by the terminal device may include information on whether the channel learning model is applicable.
  • the network device determines that the channel learning model is not applicable, so the network device sends the channel learning training signaling to determine a more suitable channel learning model.
  • the channel learning feedback signaling sent by the terminal device may not include information on whether the channel learning model is applicable.
  • whether the channel learning feedback signaling includes information on whether the channel learning model is applicable may be predefined by the protocol, or may be notified by the network device to the terminal device through signaling, which is not specifically limited in this application.
  • Scenario 2 The network device does not send channel learning training signaling to the terminal device.
  • the network device determines that the channel learning feedback includes information on whether the channel learning model is applicable.
  • the network device determines that the channel learning feedback does not include information on whether the channel learning model is applicable.
  • the terminal device determines that the channel learning feedback includes information on whether the channel learning model is applicable.
  • the terminal device determines that the channel learning feedback does not include information on whether the channel learning model is applicable.
  • the network device may determine whether the channel learning model is applicable in at least one of the following ways:
  • the information in the channel learning feedback signaling indicates whether the channel learning model is applicable.
  • 1-bit information indicates whether the channel learning model is applicable.
  • Whether the channel learning model is applicable can be fed back 1-bit information in the channel learning feedback signaling, where the 1-bit information is used to indicate whether the channel learning model is applicable. For example, if the 1-bit information is "0", it means that the channel learning model is not applicable, and if the 1-bit information is "1", it means that the channel learning model is applicable. The reverse is also possible.
  • the feedback value in the channel learning feedback signaling indicates whether the channel learning model is applicable.
  • the feedback value in the channel learning feedback signaling may be at least one of the following: a rank value, a CQI value, and a CRI value.
  • the rank value fed back by the terminal device may range from 1 to R, where R is a positive integer.
  • R is 8.
  • the number of bits in the rank field can be determined according to the maximum number of layers supported by the terminal device and the number of antenna ports.
  • the number of bits in the rank field is log 2 (min(number of layers, number of antenna ports)) rounded up, for example, log 2 (R) is rounded up.
  • the terminal device supports a maximum of 4 layers
  • the rank value can be 1 to 4
  • the rank value can be indicated by 2 bits.
  • the rank value can be 1 to 4
  • the rank value can be indicated by 2 bits.
  • the network device and/or the terminal device may determine that the number of bits in the rank field is log 2 (min(number of layers, number of antenna ports)) is rounded up.
  • the terminal device and the network device can determine that the rank value can be 0 to R, where R is positive Integer.
  • R is 8.
  • the number of bits in the field of rank is log 2 (min (number of layers, number of antenna ports)+1) rounded up, for example, log 2 (R+1) rounded up.
  • the rank value is 0, indicating that the channel learning model is not applicable.
  • the network device and/or the terminal device may determine that the number of bits in the rank field is log 2 (min(number of layers, number of antenna ports)+1) rounded up.
  • the network device may determine that the first channel learning model is not applicable according to the rank value of 0.
  • a rank value of 0 indicates that the channel learning model is not applicable.
  • the terminal device when the terminal device receives the channel learning training signaling sent by the network device, when the terminal device determines that the first channel learning model is not applicable, the terminal device may send information with a rank value of 0, and the rank value of 0 indicates the channel learning model Not applicable.
  • the network device may determine that the first channel learning model is not applicable according to the CQI value of 0.
  • a CQI value of 0 indicates that the channel learning model is not applicable.
  • the terminal device when the terminal device receives the channel learning training signaling sent by the network device, when the terminal device determines that the first channel learning model is not applicable, the terminal device may send information with a CQI value of 0, and the CQI value of 0 indicates the channel learning model Not applicable.
  • the CRI value fed back by the terminal device corresponds to the measured CSI-RS resources.
  • C is a positive integer.
  • the CRI value may be 1-C.
  • the number of bits in the CRI field may be determined according to the number of configured CSI-RS resources, for example, the number of bits in the CRI field is log 2 (C) rounded up.
  • the number of CSI-RS resources configured and measured by the terminal device is 2, and the CRI needs only 1 bit indication.
  • bit 0 represents the first configured CSI-RS resource
  • bit 2 represents the second configured CSI-RS resource.
  • the network device and/or the terminal device may determine that the number of bits of the CRI field is log 2 (C) is rounded up, where C is the number of CSI-RS resources.
  • the CRI value can be 0 to C, where C is a positive integer.
  • C is 2.
  • the number of bits in the CRI field is log 2 (C+1) rounded up.
  • the CRI value is 0, indicating that the channel learning model is not applicable.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Electromagnetism (AREA)
  • Computing Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Power Engineering (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请提供了一种通信的方法及通信装置，该方法可以包括：第一通信装置确定第一信道学习模型是否适用，该第一信道学习模型用于基于目标信道信息确定第一信道信息，该第一信道信息的数据量小于所述目标信道信息的数据量；在确定该第一信道学习模型不适用的情况下，该第一通信装置发送第一消息，该第一消息用于指示该第一信道学习模型不适用。根据本申请实施例提供的方法，第一通信装置在不需要第二通信装置的协助的情况下，就可以判断第一信道学习模型的适用性，因此可以减少信令的交互和对信道学习模型的适用性进行判断的复杂性。

Description

通信的方法及通信装置
本申请要求于2020年07月13日提交中国国家知识产权局、申请号为202010667447.X、申请名称为“通信的方法及通信装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及通信领域，并且更具体地，涉及通信的方法及通信装置。
背景技术
在大规模多输入多输出(massive multiple-input multiple-output，Massive MIMO)技术中，网络设备可通过预编码将有限的功率分配给能够有效传输的数据流，同时减小多个终端设备之间的干扰以及同一终端设备的多个信号流之间的干扰，有利于提高信号质量，实现空分复用，提高频谱利用率。终端设备可以基于下行信道测量来确定预编码矩阵，并通过反馈，使得网络设备可以基于终端设备反馈的预编码矩阵确定数据传输的预编码，进而提升信号传输性能。
目前，已知一种基于人工智能(artificial intelligence，AI)的信息传输的方法，终端设备和网络设备通过联合训练，得到第一信道学习模型和第二信道学习模型，其中，第一信道学习模型设置于终端设备侧、第二信道学习模型设置于网络设备侧，包括该第一信道学习模型和第二信道学习模型的通信系统称为AI的通信系统。具体地，该基于AI的信息传输的方法包括：终端设备获得待反馈的信息，终端设备对该待反馈的信息至少通过第一信道学习模型处理，得到需要通过空口反馈的信息，终端设备再通过反馈链路将该需要通过空口反馈的信息反馈给网络设备，网络设备收到终端设备反馈的信息，网络设备对该反馈的信息至少通过第二信道学习模型的处理，以获得终端设备侧待反馈的信息。然而，在这种基于AI的信息传输的方式中，离线训练得到的第一信道学习模型和第二信道学习模型被直接应用在信息在线传输的过程中，如果终端设备发生位移或者终端设备所处通信环境发生变化，可能会导致第一信道学习模型和第二信道学习模型不适用，从而影响基于第一信道学习模型和第二信道学习模型的AI的信息传输的方式进行信息传输的性能。
发明内容
本申请提供一种通信的方法，可以实现对正在使用的信道学习模型的适用性进行判断。
第一方面，提供了一种通信的方法，该方法可以包括：第一通信装置确定第一信道学习模型是否适用，该第一信道学习模型用于基于目标信道信息确定第一信道信息，该第一信道信息的数据量小于该目标信道信息的数据量；在确定该第一信道学习模型不适用的情况下，该第一通信装置发送第一消息，该第一消息用于指示该第一信道学习模型不适用。
其中，第一信道信息用于通过第二信道学习模型确定第二信道信息，第二信道信息与目标信道信息相同或相近。
基于上述方案，第一通信装置在不需要第二通信装置的协助的情况下，就可以判断第一信道学习模型的适用性，因此可以减少信令的交互和对信道学习模型的适用性进行判断的复杂性。
结合第一方面，在第一方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：第一通信装置在目标信道的长期统计特性的变化量大于或等于第一预设阈值的情况下，确定该第一信道学习模型不适用；或者，该第一通信装置在目标信道的长期统计特性的变化量小于该第一预设阈值的情况下，确定该第一信道学习模型适用。
基于上述技术方案，第一通信装置根据目标信道的长期统计特性确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第一方面，在第一方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据接收的第一调度信息确定该第一信道学习模型是否适用。
第一调度信息是第二通信装置根据第二信道信息发送的，第二信道信息是根据第一信道信息和第二信道学习模型确定的。
基于上述技术方案，第一通信装置根据第一调度信息确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第一方面，在第一方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据数据传输性能确定该第一信道学习模型是否适用。
其中，数据的传输性能可以包括第一数据的传输性能和/或第二数据的传输性能。第一数据是第一通信装置根据目标信道信息发送的，第二数据是第二通信装置根据第二信道信息发送的。
基于上述技术方案，第一通信装置根据数据传输性能确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第一方面，在第一方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置在所处的场景发生变化的情况下确定该第一信道学习模型不适用；或者，该第一通信装置在所处的场景没有发生变化的情况下，确定该第一信道学习模型适用；其中，该场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、车到其他设备场景、第三代合作伙伴项目(3rd generation partnership project，3GPP)协议中定义的场景。
基于上述技术方案，第一通信装置根据所处的场景是否发生变化确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。此外，根据所处的场景是否发生变化确定第一信道学习模型是否适用更易于实现。
结合第一方面，在第一方面的某些实现方式中，该第一通信装置确定第一信道学习模 型是否适用，包括：该第一通信装置在该第一信道学习模型的性能指标小于第二预设阈值的情况下，确定该第一信道学习模型不适用，该性能指标包括连续性和/或真实性；或者，该第一通信装置在该第一信道学习模型的性能指标大于或等于第二预设阈值的情况下，确定该第一信道学习模型适用。
基于上述技术方案，第一通信装置根据第一信道学习模型的性能指标确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第一方面，在第一方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据该目标信道信息与该第二信道信息的误差确定该第一信道学习模型是否适用，该第二信道信息是根据第一信道信息以及第二信道学习模型确定的，该第二信道学习模型与该第一信道学习模型对应。
基于上述技术方案，第一通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用，即通过进行信道学习模型训练确定第一信道学习模型是否适用，该方式可以不依赖于第二通信装置的辅助，第一通信装置即可确定信道学习模型是否适用，因此该方式简单快捷。在同时考虑目标信道信息和第二信道信息的特征的情况下确定第一信道学习模型是否适用，可以保证第一通信装置和第二通信装置获得相同或相近的信道信息的特征，有助于后续利用该信道信息进行数据传输时提高数据传输的性能。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：该第一通信装置接收第一指示信息，该第一指示信息用于指示该第一通信装置进行信道学习模型训练。
基于上述技术方案，不具备主动学习能力的第一通信装置在接收到第一指示信息的情况下，可以对第一信道学习模型的适用性进行判断。
基于上述技术方案，第二通信装置可以通过信令指示第一通信装置确定第一信道学习模型是否适用，以便于第二通信装置在发现第一信道学习模型不适用的情况下，及时通知第一通信装置进行验证，从而可以降低确定第一信道学习模型是否适用的时延。此外，在第二通信装置及时发现第一信道学习不适用的情况下，可以对第一信道学习模型进行及时更新，避免第一信道学习模型不适用的情况下导致的通信性能的下降。
可选地，该第一指示信息还用于指示以下一项或多项：
用于传输该第一消息的资源、该第一消息的内容、发送该第一消息的形式、信道学习模型的训练参数，该训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：该第一通信装置发送第一请求消息，该第一请求消息用于请求以下一项或多项：进行信道学习模型训练、发送该第一消息、该第一指示信息。
基于上述技术方案，第一通信装置可以通过信令请求确定第一信道学习模型是否适用，以便于在第一通信装置发现第一信道学习模型不适用的情况下，及时向第二通信装置请求进行验证，以降低确定第一信道学习模型是否适用的时延。此外在第一通信装置及时发现第一信道学习模型不适用的情况下，可以对第一信道学习模型及时更新，从而避免第一信道学习模型不适用的情况下导致的通信性能的下降。
结合第一方面，在第一方面的某些实现方式中，该第一消息还用于指示用于更新第二 信道学习模型的一个或多个配置参数，该第二信道学习模型用于根据该第一信道信息确定第二信道信息。
基于上述方案，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置，从而使得第二通信装置可以对第二信道学习模型进行更新。进一步地，第一通信装置和第二通信装置基于更新后的第一信道学习模型和第二信道学习模型进行信息传输的性能比较好。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：该第一通信装置接收第二消息，该第二消息用于指示用于更新该第一信道学习模型的一个或多个配置参数。
基于上述方案，第二通信装置在确定第一信道学习模型不适用的情况下，可以确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第一信道学习模型的一个或多个配置参数发送给第一通信装置，从而使得第一通信装置可以对第一信道学习模型进行更新。进一步地，第一通信装置和第二通信装置基于更新后的第一信道学习模型和第二信道学习模型进行信息传输的性能比较好。
结合第一方面，在第一方面的某些实现方式中，该方法还包括：该第一通信装置根据第一参数确定用于更新该第二信道学习模型的一个或多个配置参数；其中，该第一参数包括如下至少一项：该第一通信装置所在的小区的小区标识、该第一通信装置所在的场景、该第一通信装置的类型、该第一通信装置所在的地理位置。
基于上述技术方案，第一通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型，并进一步确定用于更新第二信道学习模型的配置参数，即不进行信道学习模型训练即可确定新的信道学习模型，降低了第一通信装置的处理复杂度。
第二方面，提供了一种通信的方法，该方法可以包括：第二通信装置接收第一消息；该第二通信装置根据该第一消息确定第一信道学习模型不适用，该第一信道学习模型用于基于目标信道信息确定第一信道信息，该第一信道信息的维度小于该目标信道信息的维度；该第二通信装置发送第一指示信息，该第一指示信息用于指示进行信道学习模型训练。
其中，第一信道信息用于通过第二信道学习模型确定第二信道信息，第二信道信息与目标信道信息相同或相近。
基于上述方案，第一通信装置在不需要第二通信装置的协助的情况下，就可以判断第一信道学习模型的适用性，因此可以减少信令的交互和对信道学习模型的适用性进行判断的复杂性。此外，不具备主动学习能力的第一通信装置在接收到第一指示信息的情况下，可以对第一信道学习模型的适用性进行判断。
可选地，该第一指示信息还用于指示以下一项或多项：
用于传输该第一消息的资源、该第一消息的内容、该第一消息的形式、信道学习模型的训练参数，该训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
结合第二方面，在第二方面的某些实现方式中，该方法还包括：该第二通信装置接收第一请求信令，该第一请求信令用于请求以下一项或多项：进行信道学习模型训练、发送该第一消息、该第一指示信息。
基于上述技术方案，第一通信装置可以通过信令请求确定第一信道学习模型是否适 用，以便于在第一通信装置发现第一信道学习模型不适用的情况下，及时向第二通信装置请求进行验证，以降低确定第一信道学习模型是否适用的时延。此外在第二通信装置收到请求信令时可以及时发现第一信道学习模型和/或第二信道信息不适用，并且可以让第一通信装置对信道学习模型及时进行训练以及更新，从而避免第一信道学习模型和/或第二信道信息不适用的情况下导致的通信性能的下降。
结合第二方面，在第二方面的某些实现方式中，该第一消息还用于指示用于更新第二信道学习模型的一个或多个配置参数，该第二信道学习模型用于根据该第一信道信息确定第二信道信息。
基于上述方案，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置，从而使得第二通信装置可以对第二信道学习模型进行更新。进一步地，第一通信装置和第二通信装置基于更新后的第一信道学习模型和第二信道学习模型进行信息传输的性能比较好。
结合第二方面，在第二方面的某些实现方式中，该方法还包括：该第二通信装置发送第二消息，该第二消息用于指示用于更新该第一信道学习模型的一个或多个配置参数。
基于上述方案，第二通信装置在确定第一信道学习模型不适用的情况下，可以确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第一信道学习模型的一个或多个配置参数发送给第一通信装置，从而使得第一通信装置可以对第一信道学习模型进行更新。进一步地，第一通信装置和第二通信装置基于更新后的第一信道学习模型和第二信道学习模型进行信息传输的性能比较好。
结合第二方面，在第二方面的某些实现方式中，该方法还包括：该第二通信装置根据第一参数确定用于更新该第一信道学习模型的一个或多个配置参数；其中，该第一参数包括如下至少一项：该第一通信装置所在的小区的小区标识、该第一通信装置所在的场景、该第一通信装置的类型、该第一通信装置所在的地理位置。
基于上述技术方案，第二通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型，并进一步确定用于更新第一信道学习模型的配置参数，即不进行信道学习模型训练即可确定新的信道学习模型，降低了第二通信装置的处理复杂度。
第三方面，提供了一种通信的方法，该方法可以包括：第一通信装置确定第一信道学习模型是否适用，该第一信道学习模型用于基于目标信道信息确定第一信道信息，该第一信道信息的数据量小于该目标信道信息的数据量；在确定该第一信道学习模型不适用的情况下，该第一通信装置发送第一消息，该第一消息用于指示用于更新第二信道学习模型的配置参数，该第二信道学习模型用于基于该第一信道信息确定第二信道信息。
其中，第一信道信息用于通过第二信道学习模型确定第二信道信息，第二信道信息与目标信道信息相同或相近。
基于上述方案，第一通信装置在不需要第二通信装置的协助的情况下，就可以判断第一信道学习模型的适用性，因此可以减少信令的交互和对信道学习模型的适用性进行判断的复杂性。上述方案可以适用于在第一通信装置发现第一信道学习模型不适用的情况下，及时向第二通信装置发送信息，以降低确定信道学习模型是否适用的时延。此外在第二通信装置收到信息时可以及时发现第一信道学习模型和/或第二信道信息不适用，并且可以 让第一通信装置对信道学习模型及时进行训练以及更新，从而避免第一信道学习模型和/或第二信道信息不适用的情况下导致的通信性能的下降。
此外，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置，从而使得第二通信装置可以对第二信道学习模型进行更新。进一步地，第一通信装置和第二通信装置基于更新后的第一信道学习模型和第二信道学习模型进行信息传输的性能比较好。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置在目标信道的长期统计特性的变化量大于或等于第一预设阈值的情况下，确定该第一信道学习模型不适用；或者，该第一通信装置在目标信道的长期统计特性的变化量小于该第一预设阈值的情况下，确定该第一信道学习模型适用。
基于上述技术方案，第一通信装置根据目标信道的长期统计特性确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据接收的第一调度信息确定该第一信道学习模型是否适用。
第一调度信息是第二通信装置根据第二信道信息发送的，第二信道信息是根据第一信道信息和第二信道学习模型确定的。
基于上述技术方案，第一通信装置根据第一调度信息确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据数据传输性能确定该第一信道学习模型是否适用。
其中，数据的传输性能可以包括第一数据的传输性能和/或第二数据的传输性能。第一数据是第一通信装置根据目标信道信息发送的，第二数据是第二通信装置根据第二信道信息发送的。
基于上述技术方案，第一通信装置根据数据传输性能确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置在所处的场景发生变化的情况下确定该第一信道学习模型不适用；或者，该第一通信装置在所处的场景没有发生变化的情况下，确定该第一信道学习模型适用；其中，该场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、车到其他设备场景、3GPP协议中定义的场景。
基于上述技术方案，第一通信装置根据所处的场景是否发生变化确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。此外，根据所处的场景是否发生变化确定第一信道学习模型 是否适用更易于实现。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置在该第一信道学习模型的性能指标小于第二预设阈值的情况下，确定该第一信道学习模型不适用，该性能指标包括连续性和/或真实性；或者，该第一通信装置在该第一信道学习模型的性能指标大于或等于第二预设阈值的情况下，确定该第一信道学习模型适用。
基于上述技术方案，第一通信装置根据第一信道学习模型的性能指标确定第一信道学习模型的适用性，不需要对第一信道信息进行恢复，因此不需要进行大量的计算，因此可以减小第一通信装置的处理负担。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据该目标信道信息与该第二信道信息的误差确定该第一信道学习模型是否适用，该第二信道信息是根据第一信道信息以及第二信道学习模型确定的，该第二信道学习模型与该第一信道学习模型对应。
基于上述技术方案，第一通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用，即通过进行信道学习模型训练确定第一信道学习模型是否适用，该方式可以不依赖于第二通信装置的辅助，第一通信装置即可确定信道学习模型是否适用，因此该方式简单快捷。在同时考虑目标信道信息和第二信道信息的特征的情况下确定第一信道学习模型是否适用，可以保证第一通信装置和第二通信装置获得相同或相近的信道信息的特征，有助于后续利用该信道信息进行数据传输时提高数据传输的性能。
结合第三方面，在第三方面的某些实现方式中，该第一通信装置确定第一信道学习模型是否适用，包括：该第一通信装置根据是否接收到第一指示信息确定该第一信道学习模型是否适用，该第一指示信息用于指示该第一通信装置进行信道学习模型训练。
可选地，该第一指示信息还用于指示信道学习模型的训练参数，该训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
基于上述技术方案，第二通信装置可以通过信令指示第一通信装置确定第一信道学习模型是否适用，以便于第二通信装置在发现第一信道学习模型不适用的情况下，及时通知第一通信装置进行验证，从而可以降低确定第一信道学习模型是否适用的时延，同时降低第一通信装置确定信道学习模型是否适用的复杂度。此外，在第二通信装置及时发现第一信道学习不适用的情况下，可以对第一信道学习模型进行及时更新，避免第一信道学习模型不适用的情况下导致的通信性能的下降。
第四方面，提供了一种通信的方法，该方法可以包括：第二通信装置确定第二信道学习模型是否适用，该第二信道学习模型用于根据第一信道信息确定第二信道信息，该第一信道信息是根据第一信道学习模型和目标信道信息确定的，该第一信道信息的数据量小于该目标信道信息的数据量；在确定该第二信道学习模型不适用的情况下，该第二通信装置发送第一指示信息，该第一指示信息用于指示进行信道学习模型训练；该第二通信装置接收第一消息，该第一消息用于指示用于更新该第二信道学习模型的一个或多个配置参数。
其中，第一信道信息用于通过第二信道学习模型确定第二信道信息，第二信道信息与目标信道信息相同或相近。
基于上述方案，第二通信装置在不需要第一通信装置的协助的情况下，就可以判断第 二信道学习模型的适用性，因此可以减少信令的交互和对信道学习模型的适用性进行判断的复杂性。上述方案可以适用于在第二通信装置发现第二信道学习模型不适用的情况下，及时向第一通信装置发送信息，以降低确定信道学习模型是否适用的时延。此外在第一通信装置收到信息时可以及时发现第二信道学习模型和/或第一信道信息不适用，并且第一通信装置可以对信道学习模型及时进行训练以及更新，从而避免第二信道学习模型和/或第一信道信息不适用的情况下导致的通信性能的下降。
此外，第二通信装置在确定第二信道学习模型不适用的情况下，可以向第一通信装置发送第一指示信息，并接收第一通信装置发送的第一消息；进一步地，根据第一消息对第二信道学习模型进行更新。进一步地，第一通信装置和第二通信装置基于更新后的第一信道学习模型和第二信道学习模型进行信息传输的性能比较好。
可选地，该第一指示信息还用于指示信道学习模型的训练参数，该训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
结合第四方面，在第四方面的某些实现方式中，该第二通信装置确定第二信道学习模型是否适用，包括：该第二通信装置在目标信道的长期统计特性的变化量大于或等于第三预设阈值的情况下，确定该第二信道学习模型不适用；或者，该第二通信装置在目标信道的长期统计特性的变化量小于该第三预设阈值的情况下，确定该第二信道学习模型适用。
基于上述技术方案，第二通信装置根据目标信道的长期统计特性确定第二信道学习模型的适用性，不需要第一通信装置反馈目标信道信息，因此可以减小第一通信装置的反馈开销，且不需要进行大量的计算，可以减小第二通信装置的处理负担。
结合第四方面，在第四方面的某些实现方式中，该第二通信装置确定第二信道学习模型是否适用，包括：该第二通信装置根据第一调度信息确定该第二信道学习模型是否适用。
第一调度信息是第二通信装置根据第二信道信息发送的，第二信道信息是根据第一信道信息和第二信道学习模型确定的。
基于上述技术方案，第二通信装置根据第一调度信息确定第二信道学习模型的适用性，不需要第一通信装置反馈目标信道信息，因此可以减小第一通信装置的反馈开销，且不需要进行大量的计算，可以减小第二通信装置的处理负担。
结合第四方面，在第四方面的某些实现方式中，该第二通信装置确定第二信道学习模型是否适用，包括：该第二通信装置根据数据传输性能确定该第二信道学习模型是否适用。
其中，数据的传输性能可以包括第一数据的传输性能和/或第二数据的传输性能。第一数据是第一通信装置根据目标信道信息发送的，第二数据是第二通信装置根据第二信道信息发送的。
基于上述技术方案，第二通信装置根据数据传输性能确定第二信道学习模型的适用性，不需要第一通信装置反馈目标信道信息，因此可以减小第一通信装置的反馈开销，且不需要进行大量的计算，可以减小第二通信装置的处理负担。
结合第四方面，在第四方面的某些实现方式中，该方法还包括：该第二通信装置接收第三消息，该第三消息用于指示第一通信装置所处的场景，该场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、车到其他设备场景、3GPP协议中定义的场景；该第二通信装置确定第二信道学习模型是否适用，包括：该第二通信装置在该场景发生变化的情况下确定该第二信道学习模型不适用；或者， 该第二通信装置在该场景没有发生变化的情况下，确定该第二信道学习模型适用。
基于上述技术方案，第二通信装置根据第一通信装置所处的场景确定第二信道学习模型的适用性，不需要第一通信装置反馈目标信道信息，因此可以减小第一通信装置的反馈开销，且不需要进行大量的计算，可以减小第二通信装置的处理负担。此外，根据第一通信装置所处的场景是否发生变化确定第一信道学习模型是否适用更易于实现。
结合第四方面，在第四方面的某些实现方式中，该第二通信装置确定第二信道学习模型是否适用，包括：该第二通信装置根据该目标信道信息与该第二信道信息的误差确定该第二信道学习模型是否适用。
基于上述技术方案，第二通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用，即通过进行信道学习模型训练确定第二信道学习模型是否适用，该方式可以不依赖于第一通信装置的辅助，第二通信装置即可确定信道学习模型是否适用，因此该方式简单快捷。在同时考虑目标信道信息和第二信道信息的特征的情况下确定第二信道学习模型是否适用，可以保证第一通信装置和第二通信装置获得相同或相近的信道信息的特征，有助于后续利用该信道信息进行数据传输时提高数据传输的性能。
第五方面，提供了一种通信装置，包括用于执行第一方面以及第一方面中任一种可能实现方式中的方法的各个模块或单元。
第六方面，提供了一种通信装置，包括用于执行第二方面以及第二方面中任一种可能实现方式中的方法的各个模块或单元。
第七方面，提供了一种通信装置，包括收发单元和处理单元：该处理单元用于确定第一信道学习模型是否适用，该第一信道学习模型用于基于目标信道信息确定第一信道信息，该第一信道信息的数据量小于该目标信道信息的数据量；在确定该第一信道学习模型不适用的情况下，该收发单元用于发送第一消息，该第一消息用于指示用于更新第二信道学习模型的配置参数，该第二信道学习模型用于基于该第一信道信息确定第二信道信息。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：根据信道的长期统计特征的变化量是否大于第一预设阈值确定该第一信道学习模型是否适用。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：在目标信道的长期统计特性的变化量大于或等于第一预设阈值的情况下，确定该第一信道学习模型不适用；或者，在目标信道的长期统计特性的变化量小于该第一预设阈值的情况下，确定该第一信道学习模型适用。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：根据数据传输性能确定该第一信道学习模型是否适用。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：在所处的场景发生变化的情况下确定该第一信道学习模型不适用；或者，在所处的场景没有发生变化的情况下，确定该第一信道学习模型适用，其中，该场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、车到其他设备场景、3GPP协议中定义的场景。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：在该第一信道学习模型的性能指标小于第二预设阈值的情况下，确定该第一信道学习模型不适用，该性能指标包括连续性和/或真实性；或者，在该第一信道学习模型的性能指标大于或等于第 二预设阈值的情况下，确定该第一信道学习模型适用。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：根据该目标信道信息与该第二信道信息的误差确定该第一信道学习模型是否适用，该第二信道信息是根据第一信道信息以及第二信道学习模型确定的，该第二信道学习模型与该第一信道学习模型对应。
结合第七方面，在第七方面的某些实现方式中，该处理单元具体用于：根据是否接收到第一指示信息确定该第一信道学习模型是否适用，该第一指示信息用于指示该第一通信装置进行信道学习模型训练。
第八方面，提供了一种通信装置，包括收发单元和处理单元：该处理单元用于确定第二信道学习模型是否适用，该第二信道学习模型用于根据第一信道信息确定第二信道信息，该第一信道信息是根据第一信道学习模型和目标信道信息确定的，该第一信道信息的数据量小于该目标信道信息的数据量；在确定该第二信道学习模型不适用的情况下，该收发单元用于发送第一指示信息，该第一指示信息用于指示进行信道学习模型训练；该收发单元还用于接收第一消息，该第一消息用于指示用于更新该第二信道学习模型的一个或多个配置参数。
结合第八方面，在第八方面的某些实现方式中，该处理单元具体用于：在目标信道的长期统计特性的变化量大于或等于第三预设阈值的情况下，确定该第二信道学习模型不适用；或者，在目标信道的长期统计特性的变化量小于该第三预设阈值的情况下，确定该第二信道学习模型适用。
结合第八方面，在第八方面的某些实现方式中，该处理单元具体用于：根据第一调度信息确定该第二信道学习模型是否适用。
结合第八方面，在第八方面的某些实现方式中，该处理单元具体用于：根据数据传输性能确定该第二信道学习模型是否适用。
结合第八方面，在第八方面的某些实现方式中，该收发单元还用于接收第三消息，该第三消息用于指示第一通信装置所处的场景，该场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇；该处理单元具体用于：在该场景发生变化的情况下确定该第二信道学习模型不适用；或者，在该场景没有发生变化的情况下，确定该第二信道学习模型适用。
结合第八方面，在第八方面的某些实现方式中，该处理单元具体用于：根据该目标信道信息与该第二信道信息的误差确定该第二信道学习模型是否适用。
第九方面，提供了一种通信装置，包括处理器。该处理器与存储器耦合，可用于执行存储器中的指令，以实现上述第一方面及第三方面中任一种可能实现方式中的方法。可选地，该通信装置还包括存储器。可选地，该通信装置还包括通信接口，处理器与通信接口耦合。
在一种实现方式中，该通信装置为第一通信装置。当该通信装置为第一通信装置时，该通信接口可以是收发器，或，输入/输出接口。
在另一种实现方式中，该通信装置为配置于第一通信装置中的芯片。当该通信装置为配置于第一通信装置中的芯片时，该通信接口可以是输入/输出接口。
可选地，该收发器可以为收发电路。可选地，该输入/输出接口可以为输入/输出电路。
第十方面，提供了一种通信装置，包括处理器。该处理器与存储器耦合，可用于执行存储器中的指令，以实现上述第二方面及第四方面中任一种可能实现方式中的方法。可选地，该通信装置还包括存储器。可选地，该通信装置还包括通信接口，处理器与通信接口耦合。
在一种实现方式中，该通信装置为第二通信装置。当该通信装置为第二通信装置时，该通信接口可以是收发器，或，输入/输出接口。
在另一种实现方式中，该通信装置为配置于第二通信装置中的芯片。当该通信装置为配置于第二通信装置中的芯片时，该通信接口可以是输入/输出接口。
可选地，该收发器可以为收发电路。可选地，该输入/输出接口可以为输入/输出电路。
第十一方面，提供了一种处理器，包括：输入电路、输出电路和处理电路。所述处理电路用于通过所述输入电路接收信号，并通过所述输出电路发射信号，使得所述处理器执行第一方面至第四方面中任一种可能实现方式中的方法。
在具体实现过程中，上述处理器可以为一个或多个芯片，输入电路可以为输入管脚，输出电路可以为输出管脚，处理电路可以为晶体管、门电路、触发器和各种逻辑电路等。输入电路所接收的输入的信号可以是由例如但不限于接收器接收并输入的，输出电路所输出的信号可以是例如但不限于输出给发射器并由发射器发射的，且输入电路和输出电路可以是同一电路，该电路在不同的时刻分别用作输入电路和输出电路。本申请实施例对处理器及各种电路的具体实现方式不做限定。
第十二方面，提供了一种处理装置，包括处理器和存储器。该处理器用于读取存储器中存储的指令，并可通过接收器接收信号，通过发射器发射信号，以执行第一方面至第四方面中任一种可能实现方式中的方法。
可选地，所述处理器为一个或多个，所述存储器为一个或多个。
可选地，所述存储器可以与所述处理器集成在一起，或者所述存储器与处理器分离设置。
在具体实现过程中，存储器可以为非瞬时性(non-transitory)存储器，例如只读存储器(read only memory，ROM)，其可以与处理器集成在同一块芯片上，也可以分别设置在不同的芯片上，本申请实施例对存储器的类型以及存储器与处理器的设置方式不做限定。
应理解，相关的数据交互过程例如发送指示信息可以为从处理器输出指示信息的过程，接收能力信息可以为处理器接收输入能力信息的过程。具体地，处理器输出的数据可以输出给发射器，处理器接收的输入数据可以来自接收器。其中，发射器和接收器可以统称为收发器。
上述第十二方面中的处理装置可以是一个或多个芯片。该处理装置中的处理器可以通过硬件来实现也可以通过软件来实现。当通过硬件实现时，该处理器可以是逻辑电路、集成电路等；当通过软件来实现时，该处理器可以是一个通用处理器，通过读取存储器中存储的软件代码来实现，该存储器可以集成在处理器中，可以位于该处理器之外，独立存在。
第十三方面，提供了一种计算机程序产品，所述计算机程序产品包括：计算机程序(也可以称为代码，或指令)，当所述计算机程序被运行时，使得计算机执行上述第一方面至第四方面中任一种可能实现方式中的方法。
第十四方面，提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序(也可以称为代码，或指令)当其在计算机上运行时，使得上述第一方面至第四方面中任一种可能实现方式中的方法被执行。
第十五方面，提供了一种通信系统，包括前述的第一通信装置和第二通信装置。
附图说明
图1示出了适用于本申请实施例的通信系统的示意图。
图2示出了神经网络的结构示意图。
图3示出了对称神经网络的结构示意图。
图4示出了神经网络进行运算的示意图。
图5示出了一维卷积神经网络进行运算的示意图。
图6示出了二维卷积神经网络进行运算的示意图。
图7示出了卷积神经网络的激励层进行运算的示意图。
图8示出了卷积神经的池化层进行运算的示意图。
图9示出了卷积神经网络的结构示意图。
图10示出了本申请实施例提供的通信方法的示意性流程图。
图11示出了终端设备所处的位置示意图。
图12至17示出了本申请实施例提供的通信方法的示意性流程图。
图18示出了本申请实施例提供的通信装置的示意图。
图19示出了本申请另一实施例提供的通信装置的示意性框图。
图20示出了本申请实施例提供的一种芯片系统的示意图。
具体实施方式
下面将结合附图，对本申请中的技术方案进行描述。
本申请实施例的技术方案可以应用于各种通信系统，例如：长期演进(Long Term Evolution，LTE)系统、LTE频分双工(frequency division duplex，FDD)系统、LTE时分双工(time division duplex，TDD)、通用移动通信系统(universal mobile telecommunication system，UMTS)、全球互联微波接入(worldwide interoperability for microwave access，WiMAX)通信系统、第五代(5th Generation，5G)移动通信系统或新无线接入技术(new radio access Technology，NR)、第六代(6G)移动通信系统、或者未来演进的通信系统。其中，5G移动通信系统可以包括非独立组网(non-standalone，NSA)和/或独立组网(standalone，SA)。
本申请实施例的技术方案还可以应用于卫星通信系统、高空平台(high altitude platform station,HAPS)通信等非地面网络(non-terrestrial network,NTN)系统，以及与卫星通信系统融合的各种移动通信系统。
本申请提供的技术方案还可以应用于机器类通信(machine type communication，MTC)、机器间通信长期演进技术(Long Term Evolution-machine，LTE-M)、设备到设备(device to device，D2D)网络、机器到机器(machine to machine，M2M)网络、物联网(internet of things，IoT)网络或者其他网络。其中，IoT网络例如可以包括车联网。其 中，车联网系统中的通信方式统称为车到其他设备(vehicle to X，V2X，X可以代表任何事物)，例如，该V2X可以包括：车辆到车辆(vehicle to vehicle，V2V)通信，车辆与基础设施(vehicle to infrastructure，V2I)通信、车辆与行人之间的通信(vehicle to pedestrian，V2P)或车辆与网络(vehicle to network，V2N)通信等。
本申请实施例中，网络设备可以是任意一种具有无线收发功能的设备。该设备包括但不限于：演进型节点B(evolved Node B，eNB)、无线网络控制器(radio network controller，RNC)、节点B(Node B，NB)、基站控制器(base station controller，BSC)、基站收发台(base transceiver station，BTS)、家庭基站(例如，home evolved NodeB，或home Node B，HNB)、基带单元(baseband unit，BBU)，无线保真(wireless fidelity，WiFi)系统中的接入点(access point，AP)、无线中继节点、无线回传节点、传输点(transmission point，TP)或者发送接收点(transmission and reception point，TRP)等，还可以为5G，如，NR，系统中的gNB，或，传输点(TRP或TP)，5G系统中的基站的一个或一组(包括多个天线面板)天线面板，或者，还可以为构成gNB或传输点的网络节点，如基带单元(BBU)，或，分布式单元(distributed unit，DU)等，或者未来的通信系统中的基站等。
在一些部署中，gNB可以包括集中式单元(centralized unit，CU)和DU。gNB还可以包括有源天线单元(active antenna unit，AAU)。CU实现gNB的部分功能，DU实现gNB的部分功能，例如，CU负责处理非实时协议和服务，实现无线资源控制(radio resource control，RRC)，分组数据汇聚层协议(packet data convergence protocol，PDCP)层的功能。DU负责处理物理层协议和实时服务，实现无线链路控制(radio link control，RLC)层、介质接入控制(medium access control，MAC)层和物理(physical，PHY)层的功能。AAU实现部分物理层处理功能、射频处理及有源天线的相关功能。由于RRC层的信息最终会变成PHY层的信息，或者，由PHY层的信息转变而来，因而，在这种架构下，高层信令，如RRC层信令，也可以认为是由DU发送的，或者，由DU和AAU发送的。可以理解的是，网络设备可以为包括CU节点、DU节点、AAU节点中一项或多项的设备。此外，可以将CU划分为接入网(radio access network，RAN)中的网络设备，也可以将CU划分为核心网(core network，CN)中的网络设备，本申请对此不做限定。
网络设备为小区提供服务，终端设备通过网络设备分配的传输资源(例如，频域资源，或者说，频谱资源)与小区进行通信，该小区可以属于宏基站(例如，宏eNB或宏gNB等)，也可以属于小小区(small cell)对应的基站，这里的小小区可以包括：城市小区(metro cell)、微小区(micro cell)、微微小区(pico cell)、毫微微小区(femto cell)等，这些小小区具有覆盖范围小、发射功率低的特点，适用于提供高速率的数据传输服务。
在本申请实施例中，终端设备也可以称为用户设备(user equipment，UE)、接入终端、用户单元、用户站、移动站、移动台、远方站、远程终端、移动设备、用户终端、终端、无线通信设备、用户代理或用户装置。
终端设备可以是一种向用户提供语音/数据连通性的设备，例如，具有无线连接功能的手持式设备、车载设备等。目前，一些终端的举例可以为：手机(mobile phone)、无人机、平板电脑(pad)、带无线收发功能的电脑(如笔记本电脑、掌上电脑等)、移动互联网设备(mobile internet device，MID)、虚拟现实(virtual reality，VR)设备、增强现实(augmented reality，AR)设备、工业控制(industrial control)中的无线终端、无人 驾驶(self driving)中的无线终端、远程医疗(remote medical)中的无线终端、智能电网(smart grid)中的无线终端、运输安全(transportation safety)中的无线终端、智慧城市(smart city)中的无线终端、智慧家庭(smart home)中的无线终端、蜂窝电话、无绳电话、会话启动协议(session initiation protocol，SIP)电话、无线本地环路(wireless local loop，WLL)站、个人数字助理(personal digital assistant，PDA)、具有无线通信功能的手持设备、计算设备或连接到无线调制解调器的其它处理设备、车载设备、可穿戴设备，5G网络中的终端设备或者未来演进的公用陆地移动通信网络(public land mobile network，PLMN)中的终端设备等。
其中，可穿戴设备也可以称为穿戴式智能设备，是应用穿戴式技术对日常穿戴进行智能化设计、开发出可以穿戴的设备的总称，如眼镜、手套、手表、服饰及鞋等。可穿戴设备即直接穿在身上，或是整合到用户的衣服或配件的一种便携式设备。可穿戴设备不仅仅是一种硬件设备，更是通过软件支持以及数据交互、云端交互来实现强大的功能。广义穿戴式智能设备包括功能全、尺寸大、可不依赖智能手机实现完整或者部分的功能，例如：智能手表或智能眼镜等，以及只专注于某一类应用功能，需要和其它设备如智能手机配合使用，如各类进行体征监测的智能手环、智能首饰等。
此外，终端设备还可以是物联网(internet of things，IoT)系统中的终端设备。IoT是未来信息技术发展的重要组成部分，其主要技术特点是将物品通过通信技术与网络连接，从而实现人机互连，物物互连的智能化网络。IoT技术可以通过例如窄带(narrow band，NB)技术，做到海量连接，深度覆盖，终端省电。
此外，终端设备还可以包括智能打印机、火车探测器、加油站等传感器，主要功能包括收集数据(部分终端设备)、接收网络设备的控制信息与下行数据，并发送电磁波，向网络设备传输上行数据。
为便于理解本申请实施例，首先结合图1详细说明适用于本申请实施例提供的方法的通信系统。图1示出了适用于本申请实施例提供的方法的通信系统100的示意图。如图所示，该通信系统100可以包括至少一个网络设备，如图1中所示的5G系统中的网络设备101；该通信系统100还可以包括至少一个终端设备，如图1中所示的终端设备102至107。其中，该终端设备102至107可以是移动的或固定的。网络设备101和终端设备102至107中的一个或多个均可以通过无线链路通信。每个网络设备可以为特定的地理区域提供通信覆盖，并且可以与位于该覆盖区域内的终端设备通信。例如，网络设备可以向终端设备发送配置信息，终端设备可以基于该配置信息向网络设备发送上行数据；又例如，网络设备可以向终端设备发送下行数据。因此，图1中的网络设备101和终端设备102至107构成一个通信系统。
可选地，终端设备之间可以直接通信，例如旁链路(sidelink)通信。例如可以利用D2D或者V2X技术等实现终端设备之间的直接通信。如图中所示，终端设备105与106之间、终端设备105与107之间，可以利用D2D或V2X技术直接通信。终端设备106和终端设备107可以单独或同时与终端设备105通信。
终端设备105至107也可以分别与网络设备101通信。例如可以直接与网络设备101通信，如图中的终端设备105和106可以直接与网络设备101通信；也可以间接地与网络设备101通信，如图中的终端设备107经由终端设备106与网络设备101通信。
应理解，图1示例性地示出了一个网络设备和多个终端设备，以及各通信设备之间的通信链路。可选地，该通信系统100可以包括多个网络设备，并且每个网络设备的覆盖范围内可以包括其它数量的终端设备，例如更多或更少的终端设备。可选的，该通信系统100可以是包括多个网络设备，该多个网络设备可以进行多点协作传输，例如多个网络设备可以协同或协作与一个终端设备进行通信。本申请对此不做限定。
上述各个通信设备，如图1中的网络设备101和终端设备102至107，可以配置多个天线。该多个天线可以包括至少一个用于发送信号的发射天线和至少一个用于接收信号的接收天线。其中，发射天线和接收天线可以相同也可以不同。例如相同的情况下，一个天线即可用于发射也可用于接收；不同的情况下，发射天线和接收天线是不同的天线。另外，各通信设备还附加地包括发射机链和接收机链，本领域普通技术人员可以理解，它们均可包括与信号发送和接收相关的多个部件(例如处理器、调制器、复用器、解调器、解复用器或天线等)。因此，网络设备与终端设备之间可通过多天线技术通信。
可选地，该无线通信系统100还可以包括网络控制器、移动管理实体等其他网络实体，本申请实施例不限于此。
在无线通信系统中，通常使用MIMO技术增加系统容量，即在发送端和接收端同时使用多根天线。理论上，多天线的使用结合空分复用，能成倍增加系统容量，但是实际上由于多天线的使用，也带来了干扰增强的问题，因此往往需要对信号进行一定的处理以抑制干扰带来的影响。这种通过信号处理进行干扰抑制的方法可以在接收端实现，也可以在发送端实现。在发送端实现时，可以对待发送信号进行预处理，再经过MIMO信道发送，这种发送方式就是预编码。
为了识别MIMO信道矩阵H有用的通道，需要把多个通道转化成类似于单输入单输出(single input single output，SISO)系统的一对一模式，实现发送信号S1对应接收信号R1，发送信号S2对应接收信号R2，……，也就是将多个MIMO交叉通道转换成多个平行的一对一信道。这个过程可以通过对H进行奇异值分解(singular value decomposition，SVD)实现，即H＝U∑V T，其中U和V为正交矩阵，∑为对角矩阵，其非零元素(即对角线上的元素)即为信道矩阵H的奇异值，这些奇异值通常可以按照由大到小的顺序排列，上标“T”表示转置操作。如r＝H*s+n，可以写成r＝U∑V T*s+n，其中r为接收信号，s为发送信号，n为信道噪声。在待发送数据为x的情况下，可以使s＝Vx。在接收端使用∑ -1U T对接收到的信号进行解码，则可以得到无干扰的多个一对一信道。在发送端的s＝Vx即为预编码操作，V为预编码矩阵。
由上可知，要得到与MIMO信道匹配的预编码矩阵，需要已知MIMO信道，因此需要对MIMO信道进行估计。
目前，对于TDD系统，由于上下行信道具有互易性，因此网络设备可以根据测量得到的上行信道信息获得下行信道信息，从而进行预编码矩阵的计算和下行传输。对于FDD系统，网络设备根据终端设备反馈的信道状态信息进行预编码矩阵的计算和下行传输。
为了减小反馈开销，目前提出一种基于AI的信息传输的方法，终端设备和网络设备通过联合训练，得到第一信道学习模型和第二信道学习模型，其中，第一信道学习模型设置于终端设备侧、第二信道学习模型设置于网络设备侧，包括该第一信道学习模型和第二信道学习模型的通信系统称为AI的通信系统。具体地，该基于AI的信息传输的方法包括：终端设备获得待反馈的信息，终端设备对该待反馈的信息至少通过第一信道学习模型处理 得到需要通过空口反馈的信息，终端设备再通过反馈链路将该需要通过空口反馈的信息反馈给网络设备，网络设备收到终端设备反馈的信息，网络设备对该反馈的信息至少通过第二信道学习模型的处理，以获得终端设备侧待反馈的信息。然而，在这种基于AI的信息传输的方式，离线训练得到的第一信道学习模型和第二信道学习模型直接应用在信息在线传输的过程中，如果终端设备发生位移或者终端设备所处通信环境发生变化，可能会导致第一信道学习模型和第二信道学习模型不适用，从而影响基于第一信道学习模型和第二信道学习模型的AI的信息传输的方式进行信息传输的性能。
有鉴于此，本申请实施例提供一种通信的方法，以实现对正在使用的信道学习模型的适用性的判断。
为了便于理解本申请实施例，下面首先对信道学习模型进行介绍和说明。
信道学习模型：
本申请实施例中提及的信道学习模型可以是用于进行信道获取的模型或者算法，或者可以是用于确定信道信息的模型或者算法，或者可以是与信道相关的模型或者算法，或者是在通信系统中应用到的模型或者算法等，本申请实施例对此不做限定。
例如，信道学习模型可以是机器学习的算法等，例如可以是以下机器学习算法中的一种或多种：决策树算法、朴素贝叶斯算法、支持向量机算法、随机森林算法、人工神经网络算法、推进(boosting)与装袋(bagging)算法、期望最大化(expectation maximum，EM)算法、深度学习。上述各个算法的定义和实现可以参考现有技术。
又例如，信道学习模型可以是一个神经网络(neural network，NN)模型，例如可以是以下神经网络中的至少一种：卷积神经网络、全连接神经网络、深度神经网络、前馈型神经网络、反馈型神经网络、径向基神经网络、霍普菲尔网络、马尔科夫链、玻尔兹曼机、受限玻尔兹曼机、自编码机、稀疏自编码机、变分自编码机、去噪自编码机、深度信念网络、解卷积网络、深度卷积逆向图网络、生成式对抗网络、循环神经网络、长短期记忆、神经图灵机、深度残差网络、回声状态网络、极限学习机、支持向量机。
又例如，信道学习模型也可以是：一个主成分分析算法、矩阵特征值分解算法、矩阵特征向量分解算法或者矩阵奇异值分解算法。
又例如，信道学习模型可以是一个自动编码(auto-encoder，AE)模型等。例如信道学习模型可以是实现信道降维和/或信道恢复(或者信道重构)的模型或者算法等。
下面以信道学习模型是神经网络模型为例，对信道学习模型进行进一步介绍和说明。
神经网络模型：
信道学习模型为神经网络模型时，神经网络主要由：输入层、隐藏层和输出层构成。图2示出了一个最基本的神经网络。当隐藏层只有一层时，该网络可以称为两层神经网络，由于输入层未做任何变换，可以不看做单独的一层。实际中，网络输入层的每个神经元代表了一个特征，输入层的维数也可以称为输入维数。输出层个数可以代表分类标签的个数，输出层的维数也可以称为输出维数。而隐藏层层数以及隐藏层神经元可以设定为正整数。
图3示出了一种对称型的神经网络。其中，该神经网络也可以称为自动编码。该神经网络中包括编码(从M维到D维)，即f：c n(高维数据)→z n(低维数据)；解码(从D维到M维)，即f -1：z n(低维数据)→
Figure PCTCN2021100637-appb-000001
(高维数据)；其中，M可以大于D。
在训练神经网络的过程中可以以平均近似误差为损失函数对神经网络进行训练，具体 的训练算法可以有很多，例如反向传播算法，梯度下降算法等，本申请对此不做限定。其中，平均训练误差可以表示为：
Figure PCTCN2021100637-appb-000002
如图3所示，c n为M维的信道向量，z n为D维的信道向量。因为编码的输出维数小于编码的输入维数，所以可以实现降维的思想，并且期望z n＝f(c n)可以表征输入c n的主要关键特征。即在终端设备侧利用该神经网络的编码方程f将高维信道信息进行降维，然后反馈给网络设备。相应地，网络设备可以利用该神经网络的解码方程f -1，将低维数据恢复成高维数据。例如，信道矩阵可以是A*B*S维的复数。其中，A为网络设备的天线端口数，B为终端设备的天线端口数，S为子载波个数。则M可以为A*B*S(实部和虚部独立输入，或者复数输入)或者A*B*S*2(实部和虚部联合输入)。假设A为64，B为1，S为1，则M为64或者128。D可以取值为正整数，例如2，4，5，6，8，16，32等。图3所示的神经网络为4层神经网络结构，每层的神经元的数目可以逐渐减小。例如，若M为64，D为4，则可以实现将64维的高维信道信息降维至4维信道信息。
可选的，编码和解码的方程可以对称，也可以不对称，即两者可以采用相同的结构或者不同的结构。
具体的，每个隐藏层神经元/输出层神经元的值(激活值)，都是由上一层神经元，经过运算(例如加权求和、加权求和加偏置等)与非线性变换而得到的。其中非线性变换函数(又被称为激活函数)可以是：Sigmoid函数、双曲正切(hyperbolic tangent，Tanh)函数、线性整流(rectified linear unit，Relu)函数等，Sigmoid函数、Tanh函数、Relu函数可以包括同类型函数的变种。例如图3中是以激活函数为Ranh函数和Relu函数进行举例。
(1)Sigmoid函数：是一个在生物学中常见的S型函数，也称为S型生长曲线。在信息科学中，由于其单增以及反函数单增等性质，Sigmoid函数常被用作神经网络的阈值函数。将变量映射到0,1之间，Sigmoid函数可以用如下公式表示：
Figure PCTCN2021100637-appb-000003
(2)Tanh函数：是双曲函数中的一种。在数学中，Tanh函数是由基本双曲函数、双曲正弦函数和双曲余弦函数推导而来。Tanh函数可以用如下公式表示：
Figure PCTCN2021100637-appb-000004
(3)Relu函数：用于计算隐藏层神经元的输出。Relu函数可以用如下公式表示：
f(x)＝max(0,x)         (4)
Relu激活函数的变种，可以是带泄露线性整流函数(leakly Relu，LRelu)，带泄露随机线性整流函数(random leakly Relu，RRelu)或者参数线性整流函数(parameter Relu，PRelu)等。
LRelu函数可以用如下公式表示：
Figure PCTCN2021100637-appb-000005
其中常数λ∈(0,1)。
PRelu函数可以用如下公式表示：：
Figure PCTCN2021100637-appb-000006
其中，λ为一个可通过反向传播算法(backpropagation)学习的变量。
RRelu函数可以用如下公式表示：
Figure PCTCN2021100637-appb-000007
其中λ□U(l,u)，且l,u∈(0,1)，是随机确定的。
假设N层神经网络下，从输入层到隐藏层，再从隐藏层到输出层，经过N次的变换，每次变换可以包括加权求和加偏置与非线性变换运算。例如，以从X变换为H的运算为例，则变换的过程可以表示为H＝g(X*W+b)。其中W为权值矩阵或者权值向量，简称权值，b为偏置向量或者偏置矩阵，简称偏置，g()为激活函数。假设X是1*x的矩阵，隐藏层的神经元数为h，则W可以是x*h维的矩阵，b可以是1*h维的矩阵。
如下图4所示，图中示出的输入层的神经元数为2，输入数据X＝(x 1,x 2)是一个1*2维的矩阵；隐藏层的神经元数为50，则从隐藏层输出的数据H＝(h 1,h 2,…,h 50)是一个1*50维的矩阵；输出层的神经元数为4，则从输出层输出的数据Y＝(y 1,y 2,y 3,y 4)是一个1*4维的矩阵。
从输入层到隐藏层的变换过程可以表示为：H＝g 1(X*W 1+b 1)，即：
h 1＝g 1(x 1*w 1,1,1+x 2*w 1,1,2+b 1,1)，
h 2＝g 1(x 1*w 1,2,1+x 2*w 1,2,2+b 1,2)，
……
h 50＝g 1(x 1*w 1,50,1+x 2*w 1,50,2+b 1,50)。
其中，
Figure PCTCN2021100637-appb-000008
b 1＝[b 1,1b 1,2…b 1,50]，函数g 1可以是激活函数，例如可以是Sigmoid函数、Tanh函数或Relu函数。
进一步地，从隐藏层到输出层的变换过程可以表示为：Y＝g 2(H*W 2+b 2)，即：
y 1＝g 2(h 1*w 2,1,1+h 2*w 2,1,2+…+h 50*w 2,1,50+b 2,1)，
y 2＝g 2(h 1*w 2,2,1+h 2*w 2,2,2+…+h 50*w 2,2,50+b 2,2)，
y 3＝g 2(h 1*w 2,3,1+h 2*w 2,3,2+…+h 50*w 2,3,50+b 2,3)，
y 4＝g 2(h 1*w 2,4,1+h 2*w 2,4,2+…+h 50*w 2,4,50+b 2,4)。
其中，
Figure PCTCN2021100637-appb-000009
b 2＝[b 2,1b 2,2b 2,3b 2,4]，函数g 2可以是激活函数，例如可以是Sigmoid函数、Tanh函数或Relu函数。
该步骤涉及的信道学习模型的配置参数可以是变换算法，权值向量，权值矩阵，偏置向量，偏置矩阵，激活函数，输入层维数，输出层维数，隐藏层层数，隐藏层神经元数等中的一项或多项。
下面以信道学习模型是卷积神经网络模型为例，对信道学习模型进行进一步介绍和说明。
卷积神经网络模型：
卷积神经网络中有三个基本的概念：局部感受域(local receptive fields)、共享权值(shared weights)、池化(pooling)。神经网络首先是在图像中被应用，因此下面以图像来进行举例。对应的，将神经网络应用在信道学习中是类似的，只是输入信号从图像变成信道数据。
(1)局部感受域：对于一般的深度神经网络，往往会把图像的每一个像素点连接到全连接层的每一个神经元中，而卷积神经网络则只是把每一个隐藏层神经元与图像的某个局部区域连接，从而减少参数训练的数量。例如，一张1024×720的图像，若使用9×9的感受域，则只需要81个权值参数。对于一般的视觉也是如此，当观看一张图像时，更多的时候关注的是局部。
(2)共享权值：在卷积神经网络的卷积层中，每个神经元对应的权值是相同的，即可以认为每个神经元只关注一个特性。神经元可以是滤波器，例如可以是边缘检测专用的索贝儿(Sobel)滤波器，因此可以认为卷积层的每个滤波器都会有其所关注一个数据特征，例如均值、方差、幅度、相位、垂直边缘、水平边缘、颜色、纹理等。卷积层中所有的神经元加起来之后可以认为是整个数据的特征提取器集合，因此卷积即对一组固定的权值和不同窗口内的数据做内积。
权重共享的好处：一方面，重复单元能够对特征进行识别，而不考虑它在可视域中的位置。另一方面，权值共享使得能更有效的进行特征抽取，因为极大的减少了需要学习的自由变量的个数。通过控制模型的规模，卷积网络对信道问题可以具有很好的泛化能力。
(3)池化：由于待处理的图像往往都比较大，而在实际过程中，没有必要对原图进行分析，能够有效获得图像的特征才是最主要的，因此可以采用类似于图像压缩的思想，对图像进行卷积之后，通过一个下采样过程，来调整图像的大小。
卷积神经网络可以包括数据输入层(input layer)，卷积计算层(convolution layer)，Relu激励层(Relu layer)，池化层(pooling layer)，全连接层(fully connected layer，FC layer)。
(1)数据输入层：该层要做的处理主要是对原始数据进行预处理，其中可以包括如下一项或多项：
去均值：把输入数据各个维度都中心化为0，其目的就是把样本的中心拉回到坐标系原点上。
归一化：把输入数据的幅度归一化到同样的范围，即减少各维度数据取值范围的差异而带来的干扰，例如，假设输入数据有两个维度的特征(例如为特征A 1和特征B 1)，特征A 1的数据的范围是0到10，而特征B 1的数据的范围是0到10000，如果直接使用这两个特征的数据是有问题的，比较好的做法就将是两个特征的数据的幅度做归一化处理，即可以将特征A 1和特征B 1的数据都变为0到1的范围。
主成分分析(principal component analysis，PCA)和白化：PCA用于对输入数据进行降维；白化用于对输入数据的各个特征轴上的幅度归一化。
该步骤涉及的配置参数可以是预处理操作算法，输入数据的维数，输入数据的取值范 围等中的一项或多项。
(2)卷积计算层：这一层就是卷积神经网络最重要的一个层次，也是“卷积神经网络”的名字来源。
在卷积层中有两个关键操作：1)局部关联：即将卷积层中的每个神经元看做一个滤波器(filter)；2)窗口(receptive field)滑动：即利用滤波器对局部数据进行计算。
卷积层中的名词：
1)深度(depth)：可以控制输出单元的深度，也就是filter的个数，或者是连接同一块区域的神经元个数。又名：深度列(depth column)；
2)步长(stride)：也称为步幅，即窗口一次滑动的长度。步长可以控制与同一深度的相邻两个隐含单元相连接的输入区域的距离。若步幅很小(例如stride＝1)，则与相邻两个隐含单元相连接的输入区域的重叠部分会很多；若步幅很大，则与相邻两个隐含单元相连接的输入区域的重叠部分变少；
3)补零(zero-padding)：即在输入单元周围补零来改变输入单元整体大小，从而控制输出单元的空间大小。
定义如下符号：
W 1：输入单元的大小(宽或高)；F：感受域；S 1：步幅；P：补零的数量；K：深度，即输出单元的深度。则可以用以下公式计算输出隐含单元的数目：
Figure PCTCN2021100637-appb-000010
如果计算结果不是一个整数，则说明现有参数不能正好适合输入，例如，可能是步幅设置的不合适，在此情况下，可以补零，或者重新设置步骤。
下面以一维卷积举例进行说明：
图5示出了一个一维卷积的例子。如图5所示，左边模型中输入单元有5个，即W 1＝5；输入单元的左右边界各补了一个零，即P＝1；步幅是1，即S 1＝1；因为每个输出隐含单元连接3个输入单元，因此感受域是，即F＝3。根据公式(8)可以计算出输出隐含单元的个数是：K＝(5-3+2)/1+1＝5，与图示吻合。右边的模型与左边的模型相比只是把步幅变为S 1＝2，其余不变，根据公式(8)可以算出输出隐含单元的个数为：K＝(5-3+2)/2+1＝3，也与图示吻合。若把步幅改为S 1＝3，则公式不能整除，说明步幅为3不能恰好吻合输入单元大小。
其中，卷积中包括权值参数，权值参数可以是F维的向量或矩阵，权值向量或者权值矩阵的元素取值可以为整数、实数或者复数等。例如。图5中示出的权值参数为3维的向量：[1 0-1]。隐含单元输出的数据的计算方法与上文所述的普通神经网络输出数据的计算方法可以是一样的。例如，图5中左边模型的第一个隐含单元输出数据为：0*1+1*0+2*(-1)＝-2；第二个隐含单元的输出数据为：2*1+-1*0+1*(-1)＝1；第三个隐含单元的输出数据为1*1+-3*0+0*(-1)＝1。
下面以二维卷积举例说明：
在二维卷积中，W 1可以是w 1*w 2的矩阵，F可以是f 1*f 2的矩阵，S 1可以是s 1*s 2的矩阵，P可以是p 1*p 2的矩阵，K可以是k 1*k 2的矩阵。其中，w 1，f 1，s 1，p 1，k 1分别代表矩阵的行数，w 2，f 2，s 2，p 2，k 2分别代表矩阵的列数。则可以用以下公式计算输出隐含单 元的大小：
Figure PCTCN2021100637-appb-000011
图6示出了一个二维卷积的例子。如图6所示，输入单元是5*7维，即W 1为5*7的矩阵，w 1＝5，w 2＝7；输入单元的左右便界各补一列，即P为1*0的矩阵，p 1＝1，p 2＝0；步幅S 1是1*1的矩阵，即s 1＝1，s 2＝1；因为每个输出隐含单元与9个输入单元连接，因此感受域F是3*3的矩阵，即f 1＝3，f 2＝3。根据公式(9)可以计算出输出隐含单元的个数是：k 1＝(5-3+2*1)/1+1＝5，k 2＝(7-3+2*0)/1+1＝5，因此如图6所示，输出隐含单元的是5*5维，即K是5*5的矩阵。
其中，卷积中包括权值参数，可以是F维的向量或者矩阵，权值向量或者权值矩阵的元素取值可以为整数、实数或者复数等。例如，图6中示出的权值参数为3*3维的矩阵：
Figure PCTCN2021100637-appb-000012
隐含单元输出的数据的计算方法与上文所述的普通神经网络输出数据的计算方法可以是一样的。例如，图6中第一行第一个隐含单元输出数据为：-8；第二行第二个隐含单元的输出数据为：8。
该步骤涉及的配置参数可以是输入单元的大小(宽或高)；感受域；步幅；补零的数量；深度，输出单元的深度，权值矩阵等中的一项或多项。
(3)激励层：对卷积层的输出结果做非线性映射。
卷积神经网络采用的激励函数一般可以为Relu函数。Relu函数的特点是收敛快，求梯度简单。计算公式也很简单(公式(4))，即对于输入的负值，输出全为0，对于输入的正值，则原样输出。如图7所示，第一行第一个输入值为0.77，则取max(0,0.77)得到输出值为0.77；第二行第二个值为-0.11，则取max(0,-0.11)得到输出值为0，以此类推，得到所有的输出结果。
该步骤涉及的配置参数可以是激活函数算法，激活函数可以是上述介绍的激活函数中一个或多个。
(4)池化层：池化层夹在连续的卷积层中间，用于压缩数据和参数的量，减小过拟合。简而言之，如果输入是高维数据的话，那么池化层的最主要作用就是数据压缩。因此，池化即下采样，目的是为了减少特征图。
特征不变性，类似图像处理中经常提到的特征的尺度不变性，池化操作就是对图像的尺寸进行调整。例如，一张狗的图像被缩小了一倍还能认出是一张狗的照片，这说明这张图像中仍保留着狗最重要的特征，我们一看就能判断图像中画的是一只。也就是说，图像压缩时去掉的信息只是一些无关紧要的信息，而留下的信息则是具有尺度不变性的特征，也是最能表达图像的特征。
特征降维，我们知道初始输入信道含有的信息是很大的，特征也很多，但是有些信息对于做信道学习任务没有太多用途或者有重复，因此可以把这类冗余信息去除，把最重要的特征抽取出来，这也是池化操作的一大作用。因此，池化操作在一定程度上防止过拟合，更方便优化。
池化层进行的运算一般采用以下一种或多种：1)最大池化(max pooling)：即取最 大值。例如将N 1维的输入值池化为1个输出值，则该输出值为N 1个取值中的最大值。2)均值池化(mean pooling)：即取均值。例如将N 1维的输入值池化为1个输出值，则该输出值为N 1个取值的均值。3)高斯池化：借鉴高斯模糊的方法。4)可训练池化：即训练函数f，接受N 1个点为输入，输出N 2个点，且N 2小于N 1，实现降维。5)重叠池化。6)空金字塔池化。
下面以最大池化为例进行说明：如图8所示，以输入矩阵为4*4，输出矩阵为2*2举例说明，则在此过程中，需要将4个输入值化为一个输出值。如图8中，将每个2*2窗口中的最大值作为输出值。例如，输入矩阵第一个2*2窗口中最大值是6，则得到输出矩阵的第一个元素就是6，如此类推，得到所有的输出结果。
池化操作将保存深度大小不变。如果池化层的输入单元大小不是2的整数倍，一般可以采取边缘补零(zero-padding)的方式补成2的倍数，然后再池化。
接收单元大小为：W 1*H 1*D 1；需要两个参数(hyperparameters)：空间范围(spatial extent)F 1，步幅(stride)S 2。输出大小：W 2*H 2*D 2，可以不需要引入新权重。
该步骤涉及的配置参数可以是空间范围，步幅，池化算法，输入大小(或接收单元大小)，输出大小等中一项或多项。
(5)全连接层：两层之间所有神经元都有权重连接，通常全连接层在卷积神经网络尾部。也就是跟传统的神经网络神经元的连接方式是一样的。如上文所述，每个全连接层神经元输出的值，都是对上一层神经元输出的值经过运算(例如加权求和、加权求和加偏置等)与非线性变换而得到的。
将上述层的一层或多层连接起来可以构成一个卷积神经网络，而且每个层可以有多次操作，从而构成深度神经网络。例如，卷积神经网络可以包括一层或多层的卷积层，一层或多层的激励层，一层或多层的池化层，一层或多层的全连接层，并且卷积层、激励层、池化层、全连接层的排列顺序不做限定。图9示出了一个卷积神经网络的结构示意图。
该步骤涉及的信道学习模型的配置参数可以是各层的数目，各层的先后顺序，信道学习模型的结构等中的一项或多项。
下面将结合附图详细说明本申请实施例提供的通信的方法。
应理解，下文仅为便于理解和说明，以第一通信装置与第二通信装置之间的交互为例详细说明本申请实施例提供的方法。但这不应对本申请提供的方法执行主体构成限定。例如，下文实施例示出的第一通信装置可以替换为配置于第一通信装置中的部件(如芯片或芯片系统等)。下文实施例示出的第二通信装置也可以替换为配置于第二通信装置中的部件(如芯片或芯片系统等)。
下文示出的实施例并未对本申请实施例提供的方法的执行主体的具体结构特别限定，只要能够通过运行记录有本申请实施例提供的方法的代码的程序，以根据本申请实施例提供的方法进行通信即可，例如，本申请实施例提供的方法的执行主体可以是第一通信装置或第二通信装置，或者，是第一通信装置或第二通信装置中能够调用程序并执行程序的功能模块。
下面结合图10至图17，详细说明本申请实施例提供的通信的方法。
需要说明的是，下文实施例中提及的第一通信装置可以是终端设备，也可以是配置于终端设备中的部件(如芯片或芯片系统等)。第二通信装置可以是网络设备，也可以是配 置于网络设备中的部件(如芯片或芯片系统等)。
或者，第一通信装置可以是网络设备，也可以是配置于网络设备中的部件(如芯片或芯片系统等)。第二通信装置可以是终端设备，也可以是配置于终端设备中的部件(如芯片或芯片系统等)。
或者，第一通信装置可以是终端设备，也可以是配置于终端设备中的部件(如芯片或芯片系统等)。第二通信装置可以是终端设备，也可以是配置于终端设备中的部件(如芯片或芯片系统等)。
或者，第一通信装置可以是网络设备，也可以是配置于网络设备中的部件(如芯片或芯片系统等)。第二通信装置可以是网络设备，也可以是配置于网络设备中的部件(如芯片或芯片系统等)。
还需要说明的是，下文实施例中提及的第一信道学习模型部署在第一通信装置侧，第二信道学习模型部署在第二通信装置侧。下文实施例中提及的信道学习模型在没有明确指明是第一信道学习模型还是第二信道学习模型的情况下，可以是指第一信道学习模型和/或第二信道学习模型。
还需要说明的是，本申请中的各个实施例可以相互独立，也可以相互结合，具体的，本申请对此不做限定。
还需要说明的是，本申请实施例中提及的信道学习模型训练也可以简称为信道学习训练。其中，信道学习模型训练可以是包括如下至少一项：确定第一信道学习模型，确定第二信道学习模型，确定第一信道信息，确定第二信道信息等。
还需要说明的是，针对本申请实施例中的表格，实际应用中可以是采用表格中的一行或者多行，一列或者多列，例如至少一行，至少一列。
还需要说明的是，在本申请实施例中，“大于”可以与“大于或等于”替换，“小于”可以与“小于或等于”替换。
图10是从设备交互的角度示出的本申请实施例提供的通信的方法200的示意性流程图。图10示出的方法200可以包括S210至S240。下面详细说明方法200中的各个步骤。
S210，第一通信装置确定第一信道学习模型是否适用。
第一信道学习模型用于基于目标信道信息确定第一信道信息，第一信道信息的数据量小于目标信道信息的数据量，因此也可以说第一信道学习模型用于对目标信道信息进行压缩以获得第一信道信息。
可选地，信道信息的数据量可以是指信道信息的维度。
例如，假设发送端(例如，可以是第一通信装置或第二通信装置)的天线端口数为A 2，接收端(例如，可以是第一通信装置或第二通信装置)的天线端口数为A 3，则发送端和接收端间的目标信道信息可以是A 2*A 3维的矩阵，则目标信道信息的数据量可以用A 2*A 3表示。若目标信道信息的矩阵中的元素为复数，且每个元素的实部和虚部分开表示，则目标信道信息的数据量也可以表示为A 2*A 3*2。
例如，若目标信道信息的矩阵经过第一信道学习模型处理得到的第一信道信息的矩阵的维度为B 2，则第一信道信息的数据量可以用B 2表示。
可选地，信道信息的数据量也可以是指信道信息所包含的信息量等。
可选的，目标信道信息可以看做第一信道学习模型的输入，第一信道信息可以看做第 一信道学习模型的输出。目标信道信息的数据量可以为输入的信息维度，第一信道信息的数据量可以为输出的信息维度。
第一信道信息用于通过第二信道学习模型获得第二信道信息，第二信道信息与目标信道信息的数据量相同或相近。可选的，第二信道信息可以用于进行数据传输，例如第二通信装置可以根据第二信道信息确定数据传输的调度信息，或者确定数据传输的预编码等。
应理解，第一信道学习模型于第二信道学习模型是对应的，因此，第一通信装置确定第一信道学习模型是否适用可以理解为，第一通信装置确定第一信道学习模型和第二信道学习模型是否适用。即，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定第二信道学习模型也不适用；第一通信装置在确定第一信道学习模型适用的情况下，可以确定第二信道学习模型也适用。
本申请实施例对目标信道信息不做限定。例如，在第一通信装置是终端设备的情况下，目标信道信息可以是下行信道信息；又例如，在第一通信装置是网络设备的情况下，目标信道信息可以是上行信道信息。又例如，在第一通信装置是终端设备的情况下，目标信道信息可以是上行信道信息，或者，目标信道信息可以是上行信道信息和下行信道信息，第一通信装置可以基于上下行信道的部分互易性，根据上行信道信息和下行信道信息确定第一信道学习模型和/或第二信道学习模型是否适用，或者确定新的第一信道学习模型和/或第二信道学习模型。又例如，在第一通信装置是网络设备的情况下，目标信道信息可以是下行信道信息，或者，目标信道信息可以是上行信道信息和下行信道信息，第二通信装置可以基于上下行信道的部分互易性，根据上行信道信息和下行信道信息确定第一信道学习模型和/或第二信道学习模型是否适用，或者确定新的第一信道学习模型和/或第二信道学习模型。
第一通信装置可以周期性地确定第一信道学习模型是否适用。例如，第一通信装置在第i次确定第一信道学习模型是否适用之后启动定时器，进一步地，在定时器超时的情况下，第一通信装置第i+1次确定第一信道学习模型是否适用。
该方式下，可以降低第一通信装置与第二通信装置之间交互信令的开销，并且第一通信装置定时地确定第一信道学习模型是否适用，可以避免第一信道学习模型不适用的情况下导致的通信性能的下降。
或者，第一通信装置可以在接收第一指示信息的情况下，确定第一信道学习模型是否适用。在此情况下，在S210之前，方法200还可以包括S230：第二通信装置发送第一指示信息；相应地，在S230中，第一通信装置接收第一指示信息。
该方式下，第二通信装置可以通过信令指示第一通信装置确定第一信道学习模型是否适用，以便于第二通信装置在发现第一信道学习模型不适用的情况下，及时通知第一通信装置进行验证，从而可以降低确定第一信道学习模型是否适用的时延。此外，在第二通信装置及时发现第一信道学习不适用的情况下，可以对第一信道学习模型进行及时更新，避免第一信道学习模型不适用的情况下导致的通信性能的下降。
第一指示信息用于指示第一通信装置进行信道学习模型训练，即可以指示第一通信装置确定第一信道学习模型是否适用。
可选地，第一指示信息还可以用于指示以下一项或多项：
用于传输第一消息的资源、第一消息的内容、发送第一消息的形式、信道学习模型的 训练参数。其中，第一消息用于指示第一信道学习模型不适用，训练参数可以包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
用于传输第一消息的资源可以是物理上行共享信道(physical uplink shared channel，PUSCH)/物理下行共享信道(physical downlink shared channel，PDSCH)上的资源，或者可以是物理上行控制信道(physical uplink control channel，PUCCH)/物理下行控制信道(physical downlink control channel，PDCCH)上的资源，或者可以是其他特定的资源，例如，用于传输第一消息的资源可以是资源1或资源2。
第一消息的内容可以包括以下一项或多项：秩(rank)值、信道质量指示(channel quality index，CQI)、第一信道信息、信道学习模型训练的结果(例如，指示第一信道学习模型是否适用，和/或，用于更新第二信道学习模型的一个或多个参数)。
第一通信装置向第二通信装置反馈rank值、CQI和第一信道信息的方式可以参考现有技术，为了简洁，本申请实施例不再详述；以及下文会详细说明第一通信装置反馈信道学习模型训练的结果的方式，此处暂不详述。
发送第一消息的形式可以包括以下一项或多项：周期性发送、半持续性发送、非周期性发送、差分值反馈、绝对值反馈、相对值反馈。
周期性发送：第一通信装置周期性地发送第一消息，例如，以周期T发送第一消息。
半持续性发送：第一通信装置在一段时间内持续发送第一消息，例如，第一通信装置在接收到第一指示信息之后的10s内持续发送第一消息。
非周期发送：例如，第一通信装置在收到第一指示信息的情况下，发送第一消息。
下文会详细说明差分值反馈、绝对值反馈和相对值反馈者三种形式，为了简洁，此处暂不详述。
信道学习模型的训练参数：即第一通信装置用于对第一信道学习模型和/或第二信道学习模型进行训练的参数。其中，信道学习模型训练的时间可以是周期性或非周期性。
在周期性训练信道学习模型的情况下，第一指示信息指示的信道学习模型训练的时间可以是训练的周期。具体地，可以通过指示子帧数、时隙数或无线帧数来指示信道学习模型训练的周期，例如，第一指示信息指示的子帧数是7，则信道学习模型训练的周期是7个子帧；或者可以通过指示毫秒、秒、分钟或者小时来指示信道学习模型训练的周期，例如，若第一指示信息指示的时间是5秒，则信道学习模型训练的周期是5秒。
信道学习模型训练的周期可以是和第一参数相关的。例如，在有些场景下(例如，高速运动)，第一通信装置与第二通信装置之间的信道环境变化较快，则可以以较小的周期训练信道学习模型；又例如，在有些场景下(例如，室内场景)，第一通信装置与第二通信装置之间的信道环境变化较慢，则可以以较大的周期训练信道学习模型。
在非周期性训练信道学习模型的情况下，第一指示信息指示的信道学习模型训练的时间可以是训练的周期和周期的个数。可选地，在此情况下，用于信道模型训练的其它参数可以由高层信令预先配置，第一指示信息可以指示信道学习模型训练的时间，以触发信道学习模型训练。
可选地，在S210之前，方法200还可以包括S240：第一通信装置发送第一请求消息；相应地，在S240中，第二通信装置接收第一请求消息。
第一请求消息用于请求进行以下一项或多项：信道学习模型训练、发送第一消息、第 一指示信息。
该方式下，第一通信装置可以通过信令请求确定第一信道学习模型是否适用，以便于在第一通信装置发现第一信道学习模型不适用的情况下，及时向第二通信装置请求进行验证，以降低确定第一信道学习模型是否适用的时延。此外在第一通信装置及时发现第一信道学习模型不适用的情况下，可以对第一信道学习模型及时更新，从而避免第一信道学习模型不适用的情况下导致的通信性能的下降。
本申请实施例对第一通信装置确定第一信道学习模型是否适用的方法不做限定。第一通信装置可以采用如下实现方式中的一种或多种确定第一信道学习模型是否适用。在第一通信装置确定第一信道学习模型不适用的情况下，第一通信装置和/或第二通信装置可以及时调整第一信道学习模型，提高信道学习模型的准确性和适用性，进而提升通信性能。
下文提供了第一通信装置确定第一信道学习模型和/或第二信道学习模型是否适用的方式，第一通信装置确定信道学习模型是否适用的方式可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请实施例对此不做限定。如下一种或多种确定信道学习模型是否适用的方式可以单独使用，也可以联合使用，具体的，本申请实施例对此不做限定。
需要说明的是，本申请实施例中提及的信道学习模型是否适用，也可以是指信道学习模型是否匹配、信道学习模型是否准确、信道学习模型是否过时的或信道学习模型是否错误等。
在一种实现方式中，第一通信装置可以根据目标信道的长期统计特性确定信道学习模型是否适用。例如，当第一通信装置确定目标信道的长期统计特性变化较大时，表明第一通信装置和第二通信装置间的信道特征或者信道环境发生了较大的变化，因此第一通信装置可以确定信道学习模型不适用。
例如，第一通信装置可以根据目标信道的长期统计特性的变化量是否大于或等于第一预设阈值确定第一信道学习模型是否适用。
其中，第一预设阈值可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
其中，目标信道的长期统计特性可以包括如下至少一项：rank值、大尺度特性、信道协方差矩阵、信道相关矩阵、相干时间、相干带宽等。
信道的大尺度特性可以为以下一项或多项：时延扩展(delay spread)、多普勒扩展(Doppler spread)、多普勒频移(Doppler shift)、平均信道增益(average gain)和平均时延(average delay)、接收到达角(angle of arrival，AOA)、到达角扩展(angle of arrival spread，AAS)、发射离开角(angle of departure，AOD)、离开角扩展(angle of departure spread，ADS)、空间接收参数(spatial RX parameters)和空间相关性(spatial correlation)。
第一通信装置可以根据来自第二通信装置的信号确定目标信道的长期统计特性，并进一步根据目标信道的长期统计特性确定第一信道学习模型是否适用。其中，来自第二通信装置的信号可以是参考信号或数据信号，参考信号可以是解调参考信号(demodulation reference signal，DMRS)、信道状态信息参考信号(channel state information reference signal，CSI-RS)、相位跟踪参考信号(phase tracking reference signal，PTRS)、追踪参考信号(tracking reference signal，TRS)、同步信号和物理广播信道(physical broadcast channel)PBCH块 (synchronization signal and PBCH channel，SSB)、信道探测参考信号(sounding reference signal，SRS)等，数据信号可以是在PDSCH、PDCCH、PUSCH或PUCCH上传输的信号。
例如，若第一通信装置为终端设备，第二通信装置为网络设备，则终端设备可以根据网络设备发送的参考信号和/或数据信号确定目标信道的长期统计特性。其中，参考信号可以是CSI-RS，DMRS，PTRS，TRS，SSB中至少一项，数据信号可以是在PDSCH上传输的信号、在PDCCH上传输的信号中的至少一项。目标信道可以是指下行信道。
可选的，在目标信道的长期统计特性的变化量大于或等于(或者大于)第一预设阈值的情况下，表明第一通信装置和第二通信装置之间的信道特征或信道环境发生了较大的变化，因此第一通信装置可以确定第一信道学习模型不适用；在目标信道的长期统计特性小于(或者小于或等于)第一预设阈值的情况下，表明第一通信装置和第二通信装置之间的信道特征或信道环境比较稳定，因此第一通信装置可以确定第一信道学习模型适用。
以目标信道的长期统计特性是rank值为例，第一通信装置在确定rank值变化较大的情况下，可以确定第一信道学习模型不适用。例如，第一通信装置可以在确定rank值的变化量大于或等于预设门限#1(第一预设阈值的一例)的情况下，确定第一信道学习模型不适用。其中，预设门限#1可以为R 1，其中R 1为正整数。例如R 1为2，即当rank值的变化大于或等于2时，第一通信装置可以确定第一信道学习模型不适用。例如，第一通信装置从反射体稀疏的环境运动到反射体丰富的环境，目标信道的径增多，进而rank值会发生变化，在此情况下，第一信道学习模型有可能不再适用。rank值越大，适用的信道学习模型可能越复杂，例如，信道学习模型的层数可能越高。例如当第一通信装置从室内变为室外，信道的统计特性也会发生变化，则第一通信装置可以确定第一信道学习模型有可能不适用。
以目标信道的长期统计特性为多普勒频移为例，第一通信装置在确定多普勒频移变化较大的情况下，可以确定第一信道学习模型不适用。多普勒频移可以是指当移动台以恒定的速率沿某一方向移动时，由于传播路程差的原因，会造成相位和频率的变化，通常将这种变化称为多普勒频移。多普勒频移揭示了波的属性在运动中发生变化的规律。例如，第一通信装置可以在多普勒频移的变化量大于或等于预设门限#2(第一预设阈值的一例)的情况下，确定第一信道学习模型不适用。其中，预设门限#2可以为F 2，其中F 2为实数。例如F 2为2，即当多普勒频移的变化量大于或等于2时，第一通信装置可以确定第一信道学习模型不适用。例如，多普勒频移可以反映第一通信装置的移动速度，当第一通信装置从步行变为车载时，第一通信装置的移动速度变大，则第一信道学习模型有可能不再适用。举例来说，多普勒频移与第一通信装置的运动速度、第一通信装置运动方向以及无线电波入射方向之间的夹角有关。多普勒频移的计算公式可以如下所示：
Figure PCTCN2021100637-appb-000013
其中，v为运动速度，λ为波长，θ为运动方向与无线电波入射方向之间的夹角。
可选地，第一通信装置可以根据自身运动的速度的变化量确定第一信道学习模型是否适用。例如，第一通信装置可以在运动的速度的变化量大于预设门限#3(第一预设阈值的一例)的情况下，确定第一信道学习模型不适用。其中，预设门限#3可以为S 3，其中S 3 为实数，S 3的单位可以是m/s，或者km/h，即当第一通信装置的速度的变化量大于或等于S 3时，第一通信装置可以确定第一信道学习模型不适用。例如，当第一通信装置从步行变为车载时，速度的变化量较大时，第一信道学习模型有可能不再适用。
该实现方式下，第一通信装置可以根据目标信道的长期统计特性确定信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第一信道学习模型是否适用，可以降低第一通信装置的处理复杂度，同时也不需要第二通信装置的协助即可确定信道学习模型是否适用，降低信令交互，因此该方式简单快速。
在另一种实现方式中，第一通信装置可以根据接收的第一调度信息确定第一信道学习模型是否适用。第一调度信息可以是第二通信装置根据第二信道信息确定的，第二通信装置可以向第一通信装置发送第一调度信息。
其中，第一调度信息可以包括如下至少一项：调制和编码方案(modulation and coding scheme，MCS)指示、传输块(transport block size，TBS)指示、rank指示、天线端口指示等。第一调度信息可以是物理层的下行控制信息(downlink control information，DCI)，也可以是高层信令中的调度信息。具体地，第一调度信息包含的内容和指示方式可以参考现有技术，为了简洁，本申请实施例不再详述。
MCS指示可以用于指示数据传输的调制方式(或者调制阶数)和码率。例如MCS指示可以采用如下表格中的指示方式(表1和表2)：通过指示如下表格第一列的MCS标识(MCS index)可以确定第二列的调制阶数(modulation order)和目标码率(target code rate)。其中，调制阶数2表示正交相移键控(quadrature phase shift keying，QPSK)，调制阶数4表示16正交幅度调制(16QAM，quadrature amplitude modulation)，调制阶数6表示64QAM，调制阶数6表示256QAM。或者，MCS指示也可以采用其他的指示方式，具体的，本申请对此不做限定。例如，调制方式指示和码率指示可以分开指示等。
表1
Figure PCTCN2021100637-appb-000014
表2
Figure PCTCN2021100637-appb-000015
传输块指示可以用于指示数据传输的比特大小。例如在一定时频资源内，传输块的比特大小可以取决于调制方式和码率。例如传输块的比特大小＝资源单元数*调制方式*码率*层数。
层数指示可以用于指示数据传输的层数或者流数。层数或者流数也可以对应码字数，例如，若层数或者流数小于或等于4，则对应一个码字的传输，若层数或者流数大于4，则对应两个码字的传输等。
天线端口指示可以用于指示数据的DMRS的天线端口。终端设备根据该天线端口指示可以确定天线端口数目。例如天线端口指示可以采用如下表格中的指示方式(表3～表4)：通过指示如下表格的第一列的取值(value)可以确定DMRS端口，其中DMRS端口包括DMRS天线端口号。根据DMRS天线端口号的个数可以确定DMRS天线端口数。该DMRS天线端口数可以对应数据的层数(流数)。
例如表3中的取值为2时，DMRS端口为0,1，即DMRS天线端口数为2，数据的层数(流数)为2；例如表4中的一个码字情况下取值为10时，DMRS端口为0～3，即DMRS天线端口数为4，数据的层数(流数)为4；例如表4中的两个码字情况下取值为1时，DMRS端口为0,1,2,3,4,6，即DMRS天线端口数为5，数据的层数(流数)为5。
表3
Figure PCTCN2021100637-appb-000016
表4
Figure PCTCN2021100637-appb-000017
可选的，天线端口指示也可以采用其他的表格中的指示方式，例如3GPP技术规范(technical specification，TS)38.212中的表格7.3.1.2.2-3～表格7.3.1.2.2-4，或者天线端口指示可以采用其他的表格中的指示方式，本申请对此不做限定。
作为一个示例第一通信装置可以根据测得的CQI信息与第一调度信息的差异确定第一信道学习模型是否适用。
例如第一通信装置测得的CQI信息可以反映出数据合适的调制方式和/或码率。因此第一通信装置可以根据测得的CQI信息与第一调度信息的差异确定第一信道学习模型是否适用。
作为另一个示例，第一通信装置可以根据第一调度信息和第二调度信息确定第一信道学习模型是否适用。第二调度信息是第一通信装置根据目标信道确定的。
作为一个示例，第一通信装置可以根据第一调度信息与第二调度信息的差异值是否大于或等于预设差异门限，确定第一信道学习模型是否适用。其中，预设差异门限可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
上述数据的调度信息可以反映出信道的质量情况。因此第一通信装置可以根据测得的目标信道(或实际信道)确定数据的合适的调度信息(即第二调度信息)。而第二通信装置可以根据第二信道学习模型以及第一通信装置反馈第一信道信息确定第二信道信息，并且可以根据第二信道信息确定数据的调度信息(即第一调度信息)。当第一通信装置确定的第二调度信息与第二通信装置指示给第一通信装置的第一调度信息之间的差异较大时，则表明第二信道信息和目标信道信息的差异较大，即第一信道学习模型不适用。因此第一通信装置可以根据如下方式中的至少一种确定第一信道学习模型是否适用。
例如，第一通信装置在第一调度信息与第二调度信息的差异值大于或等于预设差异门限的情况下，确定第一信道学习模型不适用；第一通信装置在第一调度信息与第二调度信息的差异值小于预设差异门限的情况下，确定第一信道学习模型适用。
预设差异门限可以是调制阶数的差异门限、码率的差异门限、MCS索引(index)值的差异门限、rank值的差异门限、TBS的差异门限、天线端口的差异门限中的至少一种。
以预设差异门限是调制阶数的差异门限为例，调制阶数的差异门限可以为N 3阶，N 3为实数，例如N 3为1,2,3,4,1/2,3/2,5/2等。
例如，在第一通信装置确定的调制方式的阶数与第一调度信息指示的调制方式的阶数的差异值大于或等于(大于)调制阶数的差异门限的情况下，第一通信装置确定第一信道学习模型不适用。
又例如，在第一通信装置确定的调制方式的阶数与第一调度信息指示的调制方式的阶数的差异值小于(小于或等于)调制阶数的差异门限的情况下，第一通信装置确定第一信道学习模型适用。
例如，若调制阶数的差异门限为2阶，第一通信装置确定的适用于数据传输的调制方式为正交相移键控(quadrature phase shift keying，QPSK)(阶数为2)，第一调度信息指示的调制方式为64相正交振幅调制(quadrature amplitude modulation，QAM)(阶数为4)，则第一通信装置确定第一信道学习模型不适用。
例如，若调制阶数的差异门限为4阶，第一通信装置确定的适用于数据传输的调制方 式为64QAM(阶数为4)，第一调度信息指示的调制方式为QPSK(阶数为2)，则第一通信装置确定第一信道学习模型适用。
例如，若调制阶数的差异门限为2阶，第一通信装置确定的适用于数据传输的调制方式为64QAM(阶数为4)，第一调度信息指示的调制方式为QPSK(阶数为2)，则第一通信装置确定第一信道学习模型不适用。
例如，第一通信装置可以在确定的调制方式的阶数低于第一调度信息指示的调制方式的阶数的差异值大于或等于调制阶数的差异门限的情况下，确定第一信道学习模型不适用。也就是说，第一通信装置在确定的调制方式的阶数与第一调度信息指示的调制方式的阶数的差异值大于或等于(大于)调制阶数的差异门限，且确定的调制方式的阶数低于第一调度信息指示的调制方式的阶数的情况下，确定第一信道学习模型不适用。
又例如，第一通信装置可以在确定的调制方式的阶数高于或等于第一调度信息指示的调制方式的阶数的的情况下，确定第一信道学习模型适用。
例如，若调制阶数的差异门限为2阶，第一通信装置确定的适用于数据传输的调制方式为64QAM(阶数为4)，第一调度信息指示的调制方式为QPSK(阶数为2)，则第一通信装置确定第一信道学习模型适用。
例如若当调制阶数的差异门限为2阶，第一通信装置确定的适用于数据传输的调制方式为QPSK(阶数2)，而第一通信装置接收到的第一调度信息中指示的调制方式为64QAM(阶数4)，则第一通信装置可以确定信道学习模型不适用。在此情况下，第二通信装置基于根据信道学习模型确定的第二信道信息向第一通信装置传输的数据的调制阶数较高，第一通信装置可能不能正确接收该数据。以预设差异门限是码率的差异门限为例，码率的差异门限可以是M 1，M 1为实数，例如M 1为200/1024,250/1024,300/1024,500/1024等。
例如，在第一通信装置确定的码率与第一调度信息指示的码率的差异值大于或等于(大于)码率的差异门限的情况下，第一通信装置确定第一信道学习模型不适用。
又例如，在第一通信装置确定的码率与第一调度信息指示的码率的差异值小于(小于或等于)码率的差异门限的情况下，第一通信装置确定第一信道学习模型适用。
例如，若码率的差异门限为200/1024，第一通信装置确定适用于数据传输的码率为400/1024，而第一调度信息中指示的码率为658/1024时，则第一通信装置可以确定第一信道学习模型不适用。
例如，第一通信装置可以在确定的码率低于第一调度信息指示的码率的差异值大于或等于码率的差异门限的情况下，确定第一信道学习模型不适用。
也就是说，第一通信装置在确定的码率与第一调度信息指示的码率的差异值大于或等于(大于)码率的差异门限，且确定码率值低于第一调度信指示的码率的情况下，确定第一信道学习模型不适用。
又例如，第一通信装置可以在确定的码率高于或等于第一调度信息指示的码率的情况下，确定第一信道学习模型适用。
例如，若码率的差异门限为200/1024，第一通信装置确定的适用于数据传输的码率为400/1024，而第一通信装置接收到的第一调度信息中指示的码率为658/1024，则第一通信装置可以确定信道学习模型不适用。在此情况下，第二通信装置基于根据第一信道学习模型确定的第二信道信息向第一通信装置传输的数据的码率较高，第一通信装置可能不能正 确接收该数据。
以预设差异门限是MCS index的差异门限为例，MCS index的差异门限可以是P 1，P 1为整数，例如P 1为1,2,3,4,5,6,8,10等。
例如，在第一通信装置确定的MCS index与第一调度信息指示的MCS index的差异值大于或等于(大于)MCS index的差异门限的情况下，第一通信装置确定第一信道学习模型不适用。
又例如，在第一通信装置确定的MCS index与第一调度信息指示的MCS index的差异值小于(小于或等于)MCS index的差异门限的情况下，第一通信装置确定第一信道学习模型适用。例如，若MCS索引值的差异门限为4，第一通信装置确定的适用于数据传输的MCS索引值为4，而第一调度信息中指示的MCS索引值为10，则第一通信装置可以确定第一信道学习模型不适用。
例如，第一通信装置可以在确定的MCS index低于第一调度信息指示的MCS index的差异值大于或等于MCS index的差异门限的情况下，确定第一信道学习模型不适用。
也就是说，第一通信装置在确定的MCS index与第一调度信息指示的MCS index的差异值大于或等于(大于)MCS index的差异门限，且确的MCS index的取值低于第一调度信息指示的MCS index的取值的情况下，第一通信装置确定第一信道学习模型不适用。
又例如，第一通信装置可以在确定的MCS index高于或等于第一调度信息指示的MCS index的情况下，确定第一信道学习模型适用。
例如若当MCS index的差异门限为4，第一通信装置确定的适用于数据传输的MCS index为4，而第一通信装置接收到的第一调度信息中指示的MCS index为10，则第一通信装置可以确定信道学习模型不适用。在此情况下，第二通信装置基于第二信道信息向第一通信装置传输的数据的MCS index较高，第一通信装置可能不能正确接收该数据。
以预设差异门限是TBS的差异门限为例，TBS的差异门限可以是Q，Q为整数，例如Q为32,64,128,256,612,1024等。
例如，在第一通信装置确定的TBS与第一调度信息指示的TBS的差异值大于或等于(大于)TBS的差异门限的情况下，第一通信装置确定第一信道学习模型不适用。又例如，在第一通信装置确定的TBS与第一调度信息指示的TBS的差异值小于(小于或等于)TBS的差异门限的情况下，第一通信装置确定第一信道学习模型适用。
例如，若TBS的差异门限为64，第一通信装置确定的适用于数据传输的TBS为288，而第一调度信息中指示的TBS为522，则第一通信装置可以确定第一信道学习模型不适用。
例如，第一通信装置可以在确定的TBS低于第一调度信息指示的TBS的差异值大于或等于TBS的差异门限的情况下，确定第一信道学习模型不适用。
也就是说，第一通信装置在确定的TBS与第一调度信息指示的TBS的差异值大于或等于(大于)TBS的差异门限，且确定的TBS低于第一调度信息指示的TBS的情况下，第一通信装置确定第一信道学习模型不适用。
又例如，第一通信装置可以在确定的TBS高于或等于第一调度信息指示的TBS的情况下，确定第一信道学习模型适用。
例如，若TBS的差异门限为64，第一通信装置确定的适用于数据传输的TBS为288，而第一通信装置接收到的第一调度信息中指示的TBS为522，则第一通信装置可以确定第 一信道学习模型不适用。在此情况下，第二通信装置基于根据该第一信道学习模型确定的第二信道信息向第一通信装置传输的数据的TBS过大，第一通信装置不能正确接收该数据。
应理解，上文仅以第一通信装置根据第一调度信息与第二调度信息中的其中一项的差异值确定第一信道学习模型是否适用为例进行说明，可选地，第一通信装置还可以根据第一调度信息与第二调度信息中的多项的差异值确定第一信道学习模型是否适用。例如，第一通信装置可以在确定的码率与第一调度信息指示的码率的差异值大于或等于码率的差异门限，且确定的TBS与第一调度信息指示的TBS的差异值大于或等于TBS差异门限的情况下，确定第一信道学习模型不适用。
还应理解，上文仅以第一通信装置根据第一调度信息和第二调度信息的差异值是否大于或等于预设差异门限确定第一信道学习模型是否适用为例进行说明。可选地，第一通信装置可以根据第一调度信息和第二调度信息的相似性是否小于预设相似门限确定第一信道学习模型是否适用。其中，预设相似门限值可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
作为一个示例，第一通信装置可以根据第一调度信息、第二调度信息和第一映射关系确定第一信道学习模型是否适用，第一映射关系用于指示调度信息与信道学习模型是否适用的对应关系。
表5中示出了第一映射关系的一例。例如第一映射关系可以是下表中的一行或者多行，第一映射关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请实施例对此不做限定。
表5
Figure PCTCN2021100637-appb-000018
例如，若第一通信装置确定的调制方式为QPSK，第一调度信息指示的调度方式为QPSK，则第一通信装置根据表5确定第一信道学习模型适用；又例如，若第一通信装置确定的码率为2/3，第一调度信息指示的码率为3/4，则第一通信装置根据表5可以确定第一信道学习模型不适用。
该实现方式下，第一通信装置可以根据第二通信装置发送的第一调度信息确定信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第一信道学习模型是否适用，可以降低第一通信装置的处理复杂度。同时根据第一通信装置和第二通信装置确定的调度信息确定第一信道学习模型是否适用，也可以保证基于第一信道学习模型下的通信性能。在第一通信装置和第二通信装置有统一的理解和认识的情况下，确定信道学习模型是否适用的方式比较简单快速。
在又一种实现方式中，第一通信装置可以根据数据传输性能确定第一信道学习模型是否适用。
因为数据传输性能和第二信道信息是相关的。因此数据传输的性能可以反映第二信道信息与目标信道信息的相似程度，进而可以反映出第一信道学习模型和/或第二信道学习模型的性能，即第一信道学习模型和/或第二信道学习模型是否适用。
其中数据传输性能可以包括第一数据的传输性能和/或第二数据的传输性能。例如，第一数据是第一通信装置根据目标信道信息发送的，第二数据是第二通信装置根据第二信道信息发送的。例如，在第一通信装置是终端设备，第二通信装置是网络设备的情况下，第一数据可以是上行数据，第二数据可以是下行数据。
第一通信装置可以根据数据传输的正确与否的性能，或者数据传输的确认(acknowledgement，ACK)/否定性确认(negative acknowledgement，NACK，)性能确定第一信道学习模型是否适用。
其中，数据传输的正确与否的性能可以通过一段时间内的数据传输的正确率来衡量。数据传输的ACK/NACK的性能可以通过一段时间内接收或发送的ACK/NACK的比例来衡量。
具体的，一段时间的时间长度的单位可以是毫秒(ms)，秒(s)，分钟(min)，小时(h)，天，月等。一段时间的时间长度的单位还可以是时隙(slot)，子帧(subframe)，无线帧(frame)，传输时间间隔(transmission time interval，TTI)等。一段时间可以是指T个时间长度的单位，T可以是实数。例如，一段时间可以是指10ms，20ms，0.5s，1s，10s等时间长度，也可以是指10个时隙，20个子帧，10个无线帧等。
作为一个示例，第一通信装置可以根据一段时间内数据传输的正确率，或者接收或发送的ACK的比例与预设门限#4的关系确定第一信道学习模型是否适用。其中，预设门限#4可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
例如，第一通信装置确定在一段时间内数据传输的正确率，或者接收或发送ACK的比例小于(或者小于或等于)预设门限#4的情况下，确定第一信道学习模型不适用。又例如，第一通信装置确定在一段时间内数据传输的正确率，或者接收或发送ACK的比例大于或等于(或大于)预设门限#4的情况下，确定第一信道学习模型适用。
例如，若一段时间为10s，预设门限#4为90％，则第一通信装置确定在10s内的数据传输的正确率，或者接收或发送ACK的比例小于90％的情况下，确定第一信道学习模型不适用。
作为另一个示例，第一通信装置根据一段时间内数据传输的失败次数，或者接收或发送NACK的个数与预设门限#5的关系确定第一信道学习模型是否适用。其中，预设门限 #5可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
例如，第一通信装置确定在一段时间内数据传输的失败次数，或者接收或发送NACK的个数大于(或者大于或等于)预设门限#5的情况下，确定第一信道学习模型不适用。又例如，第一通信装置确定在一段时间内数据传输的失败次数，或者接收或发送NACK的个数小于或等于(或小于)预设门限#5的情况下，确定第一信道学习模型适用。
例如，若一段时间为10s，预设门限#5为5次，则第一通信装置确定在10s内的数据传输的失败次数，或者接收或发送NACK的个数大于5的情况下，确定第一信道学习模型不适用。
可选地，数据传输的性能也可以是指吞吐量，吞吐率，频谱效率等性能中的至少一项。例如，第一通信装置可以在数据传输的性能小于(或者，小于或等于)预设门限#6的情况下，确定第一信道学习模型不适用。其中，预设门限#6可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
该实现方式下，第一通信装置可以根据数据传输性能确定第一信道学习模型是否适用，即可以在不进行信道学习模型训练的情况下，确定第一信道学习模型是否适用，可以降低第一通信装置的处理复杂度。同时根据数据传输性能确定第一信道学习模型是否适用的方式是以最终的通信性能来衡量第一信道学习模型的准确性，因此也可以保证基于该第一信道学习模型下的通信性能，即，该方式有助于提高通信性能。
在又一种实现方式中，在第一通信装置是终端设备的情况下，第一通信装置可以根据所处的场景是否发生变化确定第一信道学习模型是否适用。其中，场景可以是以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、V2X场景、3GPP协议定义的场景等。
不同的场景可以对应不同的信道环境，进而导致第一信道学习模型不适用。例如有些场景下，第一通信装置和第二通信装置的信道为直射径；而有些场景下，第一通信装置和第二通信装置的信道为非直射径。又例如，有些场景下，第一通信装置与第二通信装置之间的反射体比较少，相对应地，第一通信装置与第二通信装置之间的信道简单；而有些场景下，第一通信装置与第二通信装置之间的反射体比较多，相对应地，第一通信装置与第二通信装置之间的信道复杂。
因此，在第一通信装置所处的场景发生变化的情况下，第一信道学习模型可能不适用。例如，若第一通信装置从室内运动到室外，则第一通信装置可以确定第一信道学习模型不适用；又例如，第一通信装置从宏站运动到微站，则第一通信装置可以确定第一信道学习模型不适用。
该实现方式下，第一通信装置可以根据所处的场景确定第一信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第一信道学习模型是否适用，可以降低第一通信装置的处理复杂度。同时根据所处的场景的确定第一信道学习模型是否适用的方式也可以不依于第二通信装置的辅助，第一通信装置即可单独确定第一信道学习模型是否适用，因此该方式简单快捷。
可选地，在第一通信装置是网络设备，第二通信装置是终端设备的情况下，第一通信装置可以根据第二通信装置所处的的场景是否发生变化确定第一信道学习模型是否适用。 可选地，在此情况下，第一通信装置还可以接收第二通信装置发送的场景信息。
在又一种实现方式中，第一通信装置可以根据第一信道学习模型的性能指标确定第一信道学习模型是否适用。
第一通信装置可以根据目标信道信息与第一信道信息的差异性或者相似性确定第一信道学习模型是否适用。例如，第一通信装置可以根据第一信道学习模型对目标信道信息进行压缩得到第一信道信息。通过比较目标信道信息和第一信道信息的特征，进而确定第一信道学习模型的误差，从而确定第一信道学习模型是否适用。
例如，第一通信装置可以根据第一信道学习模型的性能指标是否小于第二预设阈值确定第一信道学习模型是否适用。其中，第二预设阈值可以是协议预定义的，也可以是第二通信装置向第一通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
作为一个示例，第一信道学习模型的性能指标是连续性(continuity，CT)。CT的值用于衡量高维数据点映射到低维数据点的空间中是否可以保持相对距离关系。
假设，目标信道信息中的数据点u i的K 1个最邻近点集合为
Figure PCTCN2021100637-appb-000019
目标信道信息中的数据点u i映射到第一信道信息中的数据点v i，v i的K 1个最邻近点集合为
Figure PCTCN2021100637-appb-000020
则CT表征的是u i,k的映射点v i,k在
Figure PCTCN2021100637-appb-000021
中的排序是否与u i,k在
Figure PCTCN2021100637-appb-000022
中的排序相同。CT可以用公式(1)表示：
Figure PCTCN2021100637-appb-000023
其中，k表示数据点u i,k在对于u i的集合为
Figure PCTCN2021100637-appb-000024
中的排序为k，
Figure PCTCN2021100637-appb-000025
表示u i的K 1个最邻近点集合在映射空间中的数据点集合(但是该集合中的点并不一定是在u i对应的v i在映射空间中的K 1个最邻近点集合中)，
Figure PCTCN2021100637-appb-000026
表示数据点u i,k的映射点v i,k在对于v i的集合
Figure PCTCN2021100637-appb-000027
中的排序，L为数据点的个数。
CT的取值范围为0～1，如果CT的值比较小，则表明在目标信道信息(高维数据点)中的邻近点映射到第一信道信息(低维数据点)中是不邻近的。也可以说，在此情况下，根据第一信道学习模型确定的第一信道信息并没有保留目标信道信息的有效特征，因此第一信道学习模型的性能不好，即第一信道学习模型不适用。如果CT的值比较大(接近1)，则表明目标信道信息中的数据点映射到第一信道信息之后，邻近的特征被保持。也可以说，在此情况下根据第一信道学习模型确定的第一信道信息保留了目标信道信息的有效特征，因此第一信道学习模型的性能较好，即第一信道学习模型适用。
因此，第一通信装置可以在确定CT的值大于或等于预设门限#7的情况下，确定第一信道学习模型适用；在确定CT的值小于预设门限#7的情况下，确定第一信道学习模型不适用。
作为另一个示例，第一信道学习模型的性能指标是真实性(trustworthiness，TW)。TW的值用于衡量高维数据点的空间中的邻近点映射到低维数据点的空间中是否还是邻近点。
假设，目标信道信息中的数据点u i的K 1个最邻近点集合为
Figure PCTCN2021100637-appb-000028
目标信道信息中的数据点u i映射到第一信道信息中的数据点v i，v i的K 1个最邻近点集合为
Figure PCTCN2021100637-appb-000029
则TW表征的是u i的集合
Figure PCTCN2021100637-appb-000030
中的u i,k的映射点v i,k是否还在 v i的集合
Figure PCTCN2021100637-appb-000031
中。TW可以用公式(2)表示：
Figure PCTCN2021100637-appb-000032
其中，k表示数据点u i,k的映射点v i,k在对于v i的集合
Figure PCTCN2021100637-appb-000033
中的排序为k，
Figure PCTCN2021100637-appb-000034
表示错误邻近点，即这些点在v i的集合
Figure PCTCN2021100637-appb-000035
中，但是不在u i的集合
Figure PCTCN2021100637-appb-000036
中，r(i,k)表示映射点v i,k对应的数据点u i,k在对于u i的集合
Figure PCTCN2021100637-appb-000037
中的排序，L为数据点的个数。
TW的取值范围为0～1，如果TW的值比较小，则表明在第一信道信息(低维数据点)中的邻近点在目标信道信息(高维数据点)中是不邻近的。也可以说，在此情况下，根据第一信道学习模型确定的第一信道信息并没有保留目标信道信息的有效特征，因此第一信道学习模型的性能不好，即第一信道学习模型不适用。如果TW的值比较大(接近1)，则表明第一信道信息中的数据点在目标信道信息中是邻近的。也可以说，在此情况下根据第一信道学习模型确定的第一信道信息保留了目标信道信息的有效特征，因此第一信道学习模型的性能较好，即第一信道学习模型适用。
因此，第一通信装置可以在确定TW的值大于或等于预设门限#8的情况下，确定第一信道学习模型适用；在确定TW的值小于预设门限#8的情况下，确定第一信道学习模型不适用。
该实现方式下，第一通信装置可以根据第一信道学习模型的性能指标确定第一信道学习模型是否适用，即通过进行信道学习模型训练确定第一信道学习模型是否适用。该方式可以不依赖于第二通信装置的辅助，第一通信装置即可确定信道学习模型是否适用，因此该方式简单快捷。
又例如，第一通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用。
具体地，第一通信装置根据第一信道学习模型和目标信道信息确定第一信道信息之后，根据第一信道信息和第二信道学习模型确定第二信道信息，进一步地，根据目标信道信息和第二信道信息的误差确定第一信道学习模型是否适用。
应理解，在此情况下，第一通信装置侧既部署了第一信道学习模型，也部署了第二信道学习模型。例如，第一通信装置可以预先确定第一信道学习模型的配置参数和第二信道学习模型的配置参数；进一步地，确定第一信道学习模型和第二信道学习模型。又例如，第二通信装置可以预先确定第一信道学习模型的配置参数和第二信道学习模型的配置参数；进一步地，将第一信道学习模型的配置参数和第二信道学习模型的配置参数发送至第一通信装置；相应地，第一通信装置根据第一信道学习模型的配置参数和第二信道学习模型的配置参数确定第一信道学习模型和第二信道学习模型。
可选的，在该实现方式中，第一通信装置确定第一信道学习模型是否适用的方案也可以应用在第一通信装置确定第一信道学习模型和第二信道学习模型是否适用，或者，也可以应用在第一通信装置确定第二信道学习模型是否适用。具体的，本申请对此不做限定。
第一通信装置可以在目标信道信息与第二信道信息的误差大于或等于预设门限#9的情况下，确定第一信道学习模型不适用；在目标信道信息与第二信道信息的误差小于预设门限#9的情况下，确定第一信道学习模型适用。
本申请实施例对目标信道信息与第二信道信息的误差的计算方式不做限定。
作为一个示例，第一通信装置可以根据目标信道信息与第二信道信息之间的均方误差(mean square error，MSE)确定第一信道学习模型是否适用。
假设第t次信道测量中得到的第a 2个发送端口与第a 3个接收端口之间的信道数据表示为
Figure PCTCN2021100637-appb-000038
(目标信道信息的一例)，第一通信装置根据第二信道学习模型恢复的信道数据表示为
Figure PCTCN2021100637-appb-000039
(第二信道信息的一例)，则目标信道信息与第二信道信息之间的误差可以表示为：
Figure PCTCN2021100637-appb-000040
其中，A 2表示发送端口数，0≤a 2≤A 2-1；A 3表示接收端口数，0≤a 3≤A 3-1；T表示信道测量次数，即训练样本数，0≤t≤T-1。
可选地，目标信道信息与第二信道信息之间的MSE还可以用以下公式计算：
Figure PCTCN2021100637-appb-000041
Figure PCTCN2021100637-appb-000042
Figure PCTCN2021100637-appb-000043
其中，
Figure PCTCN2021100637-appb-000044
(目标信道信息的一例)表示
Figure PCTCN2021100637-appb-000045
归一化后的数据，例如
Figure PCTCN2021100637-appb-000046
为
Figure PCTCN2021100637-appb-000047
除以矩阵H t的模(或者范数)，其中H t表示由
Figure PCTCN2021100637-appb-000048
(a 2＝1,2,…,A 2-1；a 3＝1,2,…,A 3-1)组成的矩阵，
Figure PCTCN2021100637-appb-000049
(目标信道信息的一例)表示由
Figure PCTCN2021100637-appb-000050
(a 2＝1,2,…,A 2-1；a 3＝1,2,…,A 3-1)组成的信道矩阵；
Figure PCTCN2021100637-appb-000051
(第二信道信息的一例)表示
Figure PCTCN2021100637-appb-000052
归一化后的数据，例如
Figure PCTCN2021100637-appb-000053
为
Figure PCTCN2021100637-appb-000054
除以矩阵
Figure PCTCN2021100637-appb-000055
的模(或者范数)，其中
Figure PCTCN2021100637-appb-000056
表示由
Figure PCTCN2021100637-appb-000057
(a 2＝1,2,…,A 2-1；a 3＝1,2,…,A 3-1)组成的矩阵，
Figure PCTCN2021100637-appb-000058
(第二信道信息的一例)表示由
Figure PCTCN2021100637-appb-000059
(a 2＝1,2,…,A 2-1；a 3＝1,2,…,A 3-1)组成的信道矩阵；“||||”表示范数，“|||| 2”表示范数的平方。
作为另一个示例，第一通信装置可以根据目标信道信息与第二信道信息之间的归一化均方误差(normalized mean square error，NMSE)确定第一信道学习模型是否适用。
目标信道信息与第二信道信息之间的NMSE可以用以下公式计算：
Figure PCTCN2021100637-appb-000060
Figure PCTCN2021100637-appb-000061
Figure PCTCN2021100637-appb-000062
其中，“|||| F”表示F范数。
作为另一个示例，第一通信装置可以根据目标信道信息与第二信道信息的归一化平均 相关误差(normalized mean correlation error，NMCE)确定第一信道学习模型是否适用。
目标信道信息与第二信道信息之间的NMCE可以用以下公式计算：
Figure PCTCN2021100637-appb-000063
其中，H(目标信道信息的一例)表示由
Figure PCTCN2021100637-appb-000064
(a 2＝1,2,…,A 2-1；a 3＝1,2,…,A 3-1)组成的信道矩阵，
Figure PCTCN2021100637-appb-000065
表示第a 2个发送端口与第a 3个接收端口之间的信道数据；
Figure PCTCN2021100637-appb-000066
(第二信道信息的一例)表示由
Figure PCTCN2021100637-appb-000067
(a 2＝1,2,…,A 2-1；a 3＝1,2,…,A 3-1)组成的信道矩阵，
Figure PCTCN2021100637-appb-000068
表示第一通信装置根据第二信道学习模型恢复的信道数据。
另外，参考现有技术的类型(Type)I和Type II码本的构造思路(例如LTE协议的36.系列协议或者NR协议的38.系列协议)，可以发现如果将真实下行信道的相关矩阵H HH的主次特征向量(或者H的右奇异向量)作为下行信道的估计值，就可以达到非常好的性能。换言之，信道学习模型的核心点在于对下行信道相关矩阵的特征向量的刻画，基于此提出以下两种损失函数的形式，该损失函数可以作为衡量第一信道学习模型的指标。
作为一个示例，第一通信装置可以根据目标信道信息与第二信道信息之间的归一化平均相关奇异值误差(normalized mean correlation singular error，NMCSE)确定第一信道学习模型是否适用。
对H(目标信道信息的一例)进行奇异值分解(singular value decomposition，SVD)可以得到H＝UDV H，其中D为H的奇异值矩阵，U和V为正交矩阵，则可得H HH＝VD HDV H，进一步地，NMSCE可以用以下公式表示：
Figure PCTCN2021100637-appb-000069
更为直接的是以目标信道的主次特征向量作为标签(记为H eig)；此时可以采用如下的损失函数形式：
Figure PCTCN2021100637-appb-000070
Figure PCTCN2021100637-appb-000071
可选的，第一通信装置可以根据信道学习模型的上述衡量指标中一项或多项与误差门限的对应关系确定第一信道学习模型是否适用。例如，当衡量指标的性能大于或等于某一门限时，第一通信装置确定第一信道学习模型不适用；或者，当衡量指标的性能小于某一门限时，第一通信装置确定第一信道学习模型适用。
该实现方式下，第一通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用，即通过进行信道学习模型训练确定第一信道学习模型是否适用，该方式可以不依赖于第二通信装置的辅助，第一通信装置即可确定信道学习模型是否适用，因此该方式简单快捷。在同时考虑目标信道信息和第二信道信息的特征的情况下确定第一信道学习模型是否适用，可以保证第一通信装置和第二通信装置获得相同或相近的信 道信息的特征，有助于后续利用该信道信息进行数据传输时提高数据传输的性能。
S220，第一通信装置发送第一消息，第一消息用于指示第一信道学习模型是否适用。相应地，在S220中，第二通信装置接收第一消息，并根据第一消息确定第一信道学习模型是否适用。
其中，第一通信装置在确定所述第一信道学习模型不适用的情况下，第一通信装置发送第一消息，第一消息用于指示所述第一信道学习模型不适用。
相应地，第二通信装置接收第一消息，第一消息用于指示所述第一信道学习模型不适用时，第二通信装置可以根据所述第一消息确定第一信道学习模型不适用。
可以理解，第一信道学习模型与第二信道学习模型是对应的，第二通信装置根据第一消息确定第一信道学习模型不适用，即也可以确定第二信道学习模型不适用。
可选地，第一通信装置可以在确定第一信道学习模型不适用的情况下，发送第一消息。
本申请实施例对第一消息指示第一信道学习模型不适用的方式不做限定。
作为一个示例，第一消息可以是布尔型(bool)变量。例如，第一消息是0，则表示第一信道学习模型不适用；第一消息是1，则表示第一信道学习模型适用。又例如，第一消息是0，则表示第一信道学习模型适用；第一消息是1，则表示第一信道学习模型不适用。相应地，第二通信装置根据第一消息的值可以确定第一信道学习模型是否适用。
作为另一个示例，第一消息可以是信道学习反馈信令，第一消息中可以包括信道信息。第一通信装置向第二通信装置发送第一消息，相应地，第二通信装置可以根据第一消息中的信道信息确定第一信道学习模型是否适用。可选的，信道信息可以包括如下至少一项：rank值、CQI值，CRI值。
例如，第一消息中可以包括rank值，在rank值为0的情况下，表示第一信道学习模型不适用。相应地，第二通信装置根据第一消息中的rank值可以确定第一信道学习模型是否适用。
目前，第一通信装置反馈的rank值可以是从1～R，R为正整数，例如，R＝8。rank的域的比特数可以根据第一通信装置最大支持的层数，以及天线端口数确定。例如，rank的域的比特数为log 2(min(层数，天线端口数))向上取整。例如为log 2(R)向上取整。例如，第一通信装置最大支持4层，则rank取值可以为1～4，则可以用2个比特指示rank值。又例如，天线端口数为8，第一通信装置最大支持4层，则rank取值可以为1～4，则可以用2个比特指示rank值。
在本申请实施例中，由于rank值可以为0～R，与现有的rank值相比，多了rank值为0的情况，按照现有的计算rank的域的比特数的方法计算出来的比特数不足以指示不同的rank值。在此情况下，rank的域的比特数可以为log 2(min(层数，天线端口数)+1)向上取整，例如为log 2(R+1)向上取整。例如，第一通信装置最大支持4层，则rank取值可以为0～4，则可以用3个比特指示rank值。例如，用“000”指示rank值为0，“001指示”rank值为1，“010”指示rank值为2，“100”指示rank值为3，“101”指示rank值为4。其中，rank值为0表示第一信道学习模型不适用。
又例如，，第一消息中可以包括信道质量指示(channel quality index，CQI)值，在CQI值为0的情况下，表示第一信道学习模型不适用。相应地，第二通信装置根据第一消息中的CQI值可以确定第一信道学习模型是否适用。
再例如，第一消息中可以包括信道状态信息参考信号资源(channel state information reference signal resource index，CRI)值，在CRI值为0的情况下，表示第一信道学习模型不适用。相应地，第二通信装置根据第一消息中的CRI值可以确定第一信道学习模型是否适用。
目前，第一通信装置反馈的CRI值对应用于发送参考信号的资源的个数，例如，发送参考信号的资源的个数为C，为C正整数，则CRI值可以是从1～C。CQI的域的比特数可以根据配置的参考信号的资源个数确定。例如，CRI的域的比特数为log 2(C)向上取整。例如，配置的参考信号的资源个数为2，则CRI取值可以为1～2，则可以用1个比特指示CRI值。
在本申请实施例中，由于CRI值可以为0～C，与现有的CRI值相比，多了CRI值为0的情况，按照现有的计算CRI的域的比特数的方法计算出来的比特数不足以指示不同的CRI值。在此情况下，CRI的域的比特数可以为log 2(C+1)向上取整。例如，配置的参考信号资源的个数为2，则CRI取值可以为0～2，则可以用2个比特指示CRI值。例如，用“00”指示CRI值为0，“01”指示CRI值为1，“10”指示CRI值为2。其中，CRI值为0表示第一信道学习模型不适用。
作为又一个示例，第一消息中可以包括目标信道的长期统计特性的变化量。在第一消息指示的变化量大于或等于第一预设阈值的情况下，则表示第一信道学习模型不适用。
可选地，第一通信装置可以以差分上报的形式上报目标信道的长期统计特性的变化量。例如，第一通信装置可以在第一消息中上报差分上报表中的差分值；相对应地，第二通信装置可以根据差分值和差分上报表确定上报的变化量与参考变量之间的偏移值，进一步地，第二通信装置可以根据偏移值和参考变量确定上报的变化量；再进一步地，第二通信装置可以在确定变化量大于或等于第一预设阈值的情况下确定第一信道学习模型不适用。表6示出了差分上报表的一例。例如差分值可以是下表中的一行或者多行，下表中的对应关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请实施例对此不做限定。
表6
差分值(differential value) 偏移值(offset level)
0 0
1 1
2 ≥2
3 ≤-1
可选地，第一通信装置可以以在第一消息中上报目标信道的长期统计特性的变化量的绝对值；相对应地，第二通信装置可以根据上报的绝对值确定上报的变化量；再进一步地，第二通信装置可以在确定变化量大于或等于第一预设阈值的情况下确定第一信道学习模型不适用。
可选地，第一通信装置可以以相对值上报的形式上报目标信道的长期统计特性的变化量。例如，第一通信装置可以在第一消息中上报相对值上报表中的相对值；相对应地，第二通信装置可以根据相对值和相对值上报表确定上报的变化量与参考变量之间的偏移量， 进一步地，第二通信装置可以根据偏移量和参考变量确定上报的变化量；再进一步地，第二通信装置可以在确定变化量大于或等于第一预设阈值的情况下确定第一信道学习模型不适用。表7示出了相对值上报表的一例。例如相对值可以是下表中的一行或者多行，下表中的对应关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请实施例对此不做限定。
表7
相对值 Offset level
0 1
1 1/2
2 1/4
3 2
作为又一个示例，第一通信装置通过在特定的资源上发送第一消息，以指示第一信道学习模型不适用。相应地，第二通信装置可以根据接收第一消息的资源确定第一信道学习模型是否适用。
例如，第一通信装置在物理上行共享信道(physical uplink shared channel，PUSCH)/物理下行共享信道(physical downlink shared channel，PDSCH)上发送第一消息，则表示第一信道学习模型不适用；第一通信装置在物理上行控制信道(physical uplink control channel，PUCCH)/物理下行共享信道(physical downlink control channel，PDCCH)上发送第一消息，则表示第一信道学习模型适用。又例如，第一通信装置在PUSCH/PDSCH上发送第一消息，则表示第一信道学习模型适用；第一通信装置在PUCCH/PDCCH上发送第一消息，则表示第一信道学习模型不适用。
又例如，第一通信装置在资源1上发送第一消息，则表示第一信道学习模型适用；第一通信装置在资源2上发送第一消息，则表示第一信道学习模型不适用。
又例如，第一通信装置发送第一消息占用的资源大小为X个资源单元，则表示第一信道学习模型适用；第一通信装置发送第一消息占用的资源大小为Y个资源单元，则表示第一信道学习模型不适用。其中资源单元可以是指资源粒子(resource element，RE)，符号(symbol)，或者资源块(resource block，RB)。其中，X为正整数，或者，X也可以是指一定范围，例如X为X 1～X 2，或者，大于X 1，或者，小于X 2等，X 1，X 2为正整数。Y为正整数，或者，Y也可以是指一定范围，例如Y为Y 1～Y 2，或者，大于Y 1，或者，小于Y 2等，Y 1，Y 2为正整数。
表8示出了发送第一消息使用的不同资源与第一信道学习模型是否适用的对应关系的示例。例如对应关系可以是下表中的一行或者多行，下表中的对应关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请实施例对此不做限定。
表8
Figure PCTCN2021100637-appb-000072
作为又一个示例，第一通信装置通过第一消息的比特数指示第一信道学习模型不适用。对应的，第二通信装置可以根据第一消息的比特数确定第一信道学习模型是否适用。
例如，第一通信装置发送的第一消息的比特数为Z，则表示第一信道学习模型不适用；第一通信装置发送的第一消息的比特数为W 3，则表示第一信道学习模型适用。其中，Z为正整数，或者，Z也可以是指一定范围，例如Z为Z 1～Z 2，或者，Z大于Z 1，或者，Z小于Z 2等，Z 1，Z 2为正整数。W 3为正整数，或者，W也可以是指一定范围，例如W为W 3，或者，W大于W 3,1，或者，W小于W 3,2等，W 3,1，W 3,2为正整数。
例如，第一通信装置发送的第一消息的比特数为20比特(bits)，则表示第一信道学习模型不适用；第一通信装置发送的第一消息的比特数为10bits，则表示第一信道学习模型适用。
又例如，第一通信装置发送的第一消息的比特数大于10bits，则表示第一信道学习模型不适用；第一通信装置发送的第一消息的比特数小于10bits，则表示第一信道学习模型适用。
表9示出了第一消息的比特数与第一信道学习模型是否适用的对应关系的示例。例如对应关系可以是下表中的一行或者多行，下表中的对应关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请对此不做限定。
表9
Figure PCTCN2021100637-appb-000073
其中，W’,W”,W”’的定义与W 3类似，Z’,Z”,Z”’的定义与Z类似，为了简洁，本申请实施例不再赘述。
可选地，第一通信装置还可以根据第一消息的比特数确定发送第一消息的资源。对应的，第二通信装置可以根据第一消息的比特数确定第一消息的资源。
例如，第一消息的比特数为Z，则第一通信装置确定发送第一消息占用的资源标识为资源1；又例如，第一消息的比特数为W 3，则第一通信装置确定发送第一消息占用的资源标识为资源1。
例如，第一消息的比特数为20bits，则第一通信装置确定发送第一消息占用的资源标识为资源1；第一消息的比特数为10bits，则第一通信装置确定发送第一消息占用的资源标识为资源2。
又例如，第一消息的比特数为大于10bits，则第一通信装置确定发送第一消息占用的资源标识为为资源1；第一消息的比特数为小于10bits，则第一通信装置确定发送第一消息占用的资源标识为资源2。
又例如，第一消息的比特数为Z，则第一通信装置确定发送第一消息占用的资源大小为X个资源单元；第一消息的比特数为W 3，则第一通信装置确定发送第一消息占用的资源大小为Y个资源单元。
表10示出了第一消息的比特数与发送第一消息的资源的对应关系的示例。例如对应关系可以是下表中的一行或者多行，下表中的对应关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请对此不做限定。
表10
Figure PCTCN2021100637-appb-000074
作为又一个示例，第一通信装置通过第一消息的内容指示第一信道学习模型不适用。对应的，第二通信装置可以根据第一消息的内容确定第一信道学习模型是否适用。
表11示出了第一消息的内容与第一信道学习模型是否适用的对应关系的示例。例如对应关系可以是下表中的一行或者多行，下表中的对应关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请对此不做限定。
表11
Figure PCTCN2021100637-appb-000075
如表11所示，若第一消息中包括rank值、CQI和第一信道信息，则表示第一信道学习模型适用，若第一消息中包括rank值和CQI，则表示第一信道学习模型不适用；若第一消息中包括CQI和第一信道信息，则表示第一信道学习模型适用，若第一消息中包括CQI，则表示第一信道学习模型不适用；若第一消息中包括第一信道信息，则表示第一信道学习模型适用，若第一消息中包括rank值，则表示第一信道学习模型不适用；若第一消息中包括CQI和第一信道信息，则表示第一信道学习模型适用，若第一消息中包括rank值，则表示第一信道学习模型不适用。
可选地，在第一通信装置确定第一信道学习模型适用的情况下，第一通信装置可以以差分上报的形式反馈第一信道信息。在第一信道学习模型适用的情况下，由于第一通信装置根据第一信道学习模型和目标信道信息确定的第一信道信息的取值的范围是相对稳定的，因此可以反馈当前确定的第一信道信息的取值与前一次确定的第一信道信息的取值的差值，以减小反馈开销。
可选地，第一消息中还包括用于更新第二信道学习模型的一个或多个配置参数。相应地，第二通信装置可以根据第一消息更新第二信道学习模型。一个或多个配置参数可以包括如下至少一项：模型类型、模型结构、模型算法、模型权值向量、模型权值矩阵、模型偏置向量、模型偏置矩阵、模型激活函数。可选的，第一消息中也可以包括用于更新第一信道学习模型的一个或多个配置参数。
可选的，用于更新第一信道学习模型的一个或多个配置参数也可以通过其他消息发送，其他消息与第一消息不同，具体的，本申请实施例对此不做限定。
其中，模型的类型包括机器学习的算法、神经网络模型或自动编码模型。
例如，在信道学习模型是机器学习的算法的情况下，配置参数可以包括模型算法，用于指示信道学习模型具体是机器学习算法中的哪一种算法。
又例如，在信道学习模型是神经网络模型的情况下，配置参数可以包括模型结构，用于指示信道学习模型具体是哪一种神经网络，以及模型结构还可以包括输入层维数、输出层维数、隐藏层层数、隐藏层神经元数、训练算法、损失函数中的一项或多项。配置参数可以包括变换算法、权值矩阵权值向量、偏置向量、偏置矩阵、激活函数中的一项或多项。
再例如，在信道学习模型是卷积神经网络模型的情况下，配置参数可以包括模型结构，用于指示各层的数目和/或各层的先后顺序。配置参数还可以包括以下一项或多项：针对数据输入层的参数(预处理操作算法、输入数据的维数、输入数据的取值范围)、针对卷积层的参数(输入单元的大小、感受域、步幅、补零的数量、深度、输出单元的深度、权值矩阵)、针对激励层的参数(激活函数)、针对池化层的参数(池化算法、空间范围、步幅、输入单元的大小、输出单元的大小)、针对全连接层的参数(权值矩阵、权值向量、偏置矩阵、偏置向量)。
第一通信装置在确定第一信道学习模型不适用的情况下，可以确定新的第一信道学习模型和新的第二信道学习模型，即确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置，以用于更新先前的第二信道学习模型。因此，新的第二信道学习模型的一个或多个配置参数可以认为是用于更新先前的第二信道学习模型的一个或多个配置参数。可选地，第一通信装置可以将新的第二信道学习模型的配置参数中不同于先前的第二信道学习 模型的一个或多个配置参数发送给第二通信装置，以用于更新先前的第二信道学习模型。
可选的，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定新的第二信道学习模型，即确定用于更新第二信道学习模型的一个或多个配置参数，并将新的第二信道学习模型的一个或多个配置参数发送给第二通信装置，以用于更新先前的第二信道学习模型。
本申请实施例对第一通信装置确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数的方式不做限定。即本申请实施例对第一通信装置确定第一信道学习模型和/或第二信道学习模型的方式不做限定。以下实施例提供了第一通信装置确定第一信道学习模型和/或第二信道学习模型的方式，第一通信装置确定信道学习模型的方式可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请实施例对此不做限定。如下一种或多种确定信道学习模型的方式可以单独使用，也可以联合使用，具体的，本申请实施例对此不做限定
作为一个示例，第一通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型。
其中，第一参数包括如下至少一项：终端设备所在的小区的小区标识、终端设备所在的场景、终端设备置的类型、终端设备所在的地理位置。
可选地，在第一通信装置是终端设备的情况下，第一通信装置可以根据自身所在的小区、所在的场景、类型或所在的地理位置确定第一参数。
可选地，在第一通信装置是网络设备、第二通信装置是终端设备的情况下，第一通信装置可以接收第二通信装置发送的与第一参数相关的信息，并确定第一参数。
终端设备所在的场景可以是室内、室外、郊区、城镇、外界环境(例如，白天、晚上、晴天、阴天、交通顺畅、交通拥堵)等。例如，终端设备处于室内，则第一通信装置可以确定与室内场景对应的第一信道学习模型和/或第二信道学习模型，进一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
终端设备的类型可以是终端设备的天线端口数、处理能力等。
终端设备所在的地理位置可以是三维坐标、二维坐标、定位数据等。如图11所示，若终端设备所在的地理位置是区域1，则第一通信装置可以确定与区域1对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；若终端设备所在的地理位置是区域2，则第一通信装置可以确定与区域2对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
可选地，在第一通信装置是网络设备的情况下，第一通信装置向第二通信装置发送的用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数可以是终端级、小区级或终端组级。以终端级为例，第一通信装置以单播的方式分别向每个第二通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以小区级为例，第一通信装置以广播的方式向小区内的多个第二通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以终端组级为例，第一通信装置以广播或组播的方式向终端组内的多个第二通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
具体地，第一通信装置可以根据第一参数从数据库#1保存的配置参数集中确定第一信道学习模型和第二信道学习模型。
其中，数据库#1可以是保存信道学习模型相关信息或者相关数据的数据库，例如数据库#1可以保存信道学习模型的配置参数等。
数据库#1可以是第一通信装置的本地数据库，也可以是存储在高层的数据库。
以数据库#1是存储在高层的数据库为例，存储在高层的数据库可以是存储在移动性管理单元、核心网、云端、中央管理器、运营商系统、第一通信装置群组管理系统或数据中心中的数据库。在此情况下，第一通信装置可以与高层进行通信交互，并从存储在高层的数据库中确定信道学习模型的配置参数。例如，第一通信装置可以根据第一参数从高层数据库下载和/或读取数据库中存储的信道学习模型的配置参数，进而确定第一信道学习模型和/或第二信道学习模型。又例如，第一通信装置可以接收高层网元发送的关于信道学习模型的配置参数，进而确定第一信道学习模型和/或第二信道学习模型。
以数据库#1是存储在第一通信装置的本地的数据库为例，第一通信装置可以从本地数据库下载和/或读取数据库中存储的信道学习模型的配置参数，进而确定第一信道学习模型和/或第二信道学习模型。
第一通信装置的本地数据库中存储的信道学习模型的配置参数可以是预配置的，也可以是第一通信装置侧通过训练学习确定并且存储在本地数据库中的。第一通信装置可以在应用信道学习模型之前，根据第一参数从本地数据库中确定信道学习模型。
该实现方式下，第一通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型，并进一步确定用于更新第二信道学习模型的配置参数，即不进行信道学习模型训练即可确定新的信道学习模型，降低了第一通信装置的处理复杂度。
作为另一个示例，第一通信装置可以对第一信道学习模型和第二信道学习模型进行训练，以获得第一信道学习模型和第二信道学习模型，并一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
第一通信装置对第一信道学习模型和第二信道学习模型进行训练的方式可以是：
1)基于接收的参考信号获得目标信道信息；
2)根据目标信道信息和第一信道学习模型确定第一信道信息；
3)根据第一信道信息和第二信道学习模型确定第二信道信息；
4)计算第二信道信息和目标信道信息的误差；
5)基于得到的误差计算损失函数，通过损失函数计算梯度信息，并将梯度信息反向传播；
6)通过梯度下降的方法更新第一信道学习模型和第二信道学习模型。
若更新第一信道学习模型和第二信道学习模型之后，损失函数的值小于预设门限#10，则将上述梯度信息作为用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置。
若损失函数的值大于或等于预设门限#10，则第一通信装置继续基于上述方法对第一信道学习模型和第二信道学习模型进行训练，直至损失函数的值小于预设门限#10。进一步地，将多次训练过程中获得的多个梯度信息累加起来作为用于更新和第二信道学习模型的一个或多个配置参数发送给第二通信装置。
具体地，第一通信装置对信道学习模型进行训练的训练参数可以是预配置的，或者可以是根据来自第二通信装置的指示信息确定的。
可选地，在用于更新信道学习模型的配置参数包括多个的情况下，第一通信装置可以采用相同的方法确定多个配置参数中的每一个，也可以采用不用的方法确定多个配置参数中的每一个。
可选地，第一通信装置可以采用一级或多级的方式确定用于更新信道学习模型的配置参数。以一级方式为例，第一通信装置可以采用上述方法中的一种或多种确定用于更新信道学习模型的配置参数；以多级方式为为例，第一通信装置可以先采用上述方法中的一种或多种确定用于更新信道学习模型的一部分配置参数，再采用上述方法中的一种或多种确定用于更新信道学习模型的另一部分配置参数。例如，第一通信装置根据第一参数从高层数据库中确定信道学习模型的结构，再根据第一参数从本地数据库中确定信道学习模型的维数、运算和/或函数，最后再根据来自第二通信装置的与配置参数相关的信息确定信道学习模型的变量。
可选地，若方法200执行了S230，则用于传输第一消息的资源、第一消息的内容或者发送第一消息的形式可以是由第一指示信息指示的；若方法200没有执行S230，则用于传输第一消息的资源、第一消息的内容或者发送第一消息的形式可以是预配置的，以便于第二通信装置可以正确接收第一消息。
可选地，在S220之后，方法200还可以包括：第二通信装置发送第二消息，第二消息用于指示用于更新第一信道学习模型的一个或多个配置参数；相应地，第一通信装置接收第二消息，并根据第二消息更新第一信道学习模型。
第二通信装置接收到第一消息，并根据第一消息确定第一信道学习模型不适用之后，则可以确定新的第一信道学习模型和新的第二信道学习模型，即确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数，并将新的第一信道学习模型的一个或多个配置参数发送给第一通信装置，以用于更新第一信道学习模型。因此，新的第一信道学习模型的一个或多个配置参数可以认为是用于更新先前的第一信道学习模型的一个或多个配置参数。可选地，第二通信装置可以将新的第一信道学习模型的配置参数中不同于先前的第一信道学习模型的一个或多个配置参数发送给第一通信装置，以用于更新先前的第一信道学习模型。
本申请实施例对第二通信装置确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数的方式不做限定。即本申请实施例对第二通信装置确定第一信道学习模型和/或第二信道学习模型的方式不做限定。以下实施例提供了第二通信装置确定第一信道学习模型和/或第二信道学习模型的方法，第二通信装置确定信道学习模型的方法可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请实施例对此不做限定。如下一种或多种确定信道学习模型的方法可以单独使用，也可以联合使用，具体的，本申请对此不做限定。
作为一个示例，第二通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型。
其中，第一参数包括如下至少一项：终端设备所在的小区的小区标识、终端设备所在的场景、终端设备的类型、终端设备所在的地理位置。
可选地，在第二通信装置是终端设备的情况下，第二通信装置可以根据自身所在的小区、所在的场景、类型或所在的地理位置确定第一参数。
可选地，在第二通信装置是网络设备、第一通信装置是终端设备的情况下，第二通信装置可以接收第一通信装置发送的与第一参数相关的信息，并确定第一参数。
终端设备所在的场景可以是室内、室外、郊区、城镇、外界环境(例如，白天、晚上、晴天、阴天、交通顺畅、交通拥堵)等。例如，终端设备处于室内，则第二通信装置可以确定与室内场景对应的第一信道学习模型和/或第二信道学习模型，进一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
终端设备的类型可以是终端设备的天线端口数、处理能力等。
终端设备所在的地理位置可以是三维坐标、二维坐标、定位数据等。如图11所示，若终端设备所在的地理位置是区域1，则第二通信装置可以确定与区域1对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；若终端设备所在的地理位置是区域2，则第二通信装置可以确定与区域2对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
可选地，在第二通信装置是网络设备的情况下，第二通信装置向第一通信装置发送的用于更新第一信道学习模型的一个或多个配置参数可以是终端级、小区级或终端组级。以终端级为例，第二通信装置以单播的方式分别向每个第一通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以小区级为例，第二通信装置以广播的方式向小区内的多个第一通信装置发送用于更新第二信道学习模型和/或第一信道学习模型的一个或多个配置参数；以终端组级为例，第二通信装置以广播或组播的方式向终端组内的多个第一通信装置发送用于更新第二信道学习模型和/或第一信道学习模型的一个或多个配置参数。
具体地，第二通信装置可以根据第一参数从数据库#2保存的配置参数集中确定第一信道学习模型和/或第二信道学习模型。
其中，数据库#2可以是携带信道学习模型相关信息或者相关数据的数据库，例如数据库#2中可以保存信道学习模型的配置参数等。
数据库#2可以是第二通信装置的本地数据库，也可以是存储在高层的数据库。
以数据库#2是存储在高层的数据库为例，存储在高层的数据库可以是存储在移动性管理单元、核心网、云端、中央管理器、运营商系统、第二通信装置群组管理系统、数据中心中的数据库。在此情况下，第二通信装置可以与高层进行通信交互，并从高层的数据库中确定信道学习模型的配置参数。例如，第二通信装置可以根据第一参数从高层数据库下载和/或读取数据库中存储的信道学习模型，进而确定新的第一信道学习模型和新的第二信道学习模型。又例如，第二通信装置可以接收高层网元发送的关于信道学习模型的配置参数，进而确定新的第一信道学习模型和新的第二信道学习模型。
以数据库#2是存储在第二通信装置的本地的数据库为例，第二通信装置可以从本地数据库下载和/或读取数据库中信道学习模型，进而确定新的第一信道学习模型和新的第二信道学习模型。
第二通信装置的本地数据库中存储的信道学习模型的配置参数库预配置的，也可以是 第二通信装置侧通过训练学习确定并且存储在本地数据库中的。第二通信装置可以在应用信道学习模型之前，根据第一参数从本地数据库确定信道学习模型。
该实现方式下，第二通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型，即不进行信道学习模型训练即可确定新的信道学习模型，降低了第二通信装置的处理复杂度。
作为另一个示例，第二通信装置可以对第一信道学习模型和第二信道学习模型进行训练，以获得第一信道学习模型和第二信道学习模型，并进一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
第二通信装置对第一信道学习模型和第二信道学习模型进行训练的方式可以是：
1)根据第一信道信息和第二信道学习模型确定第二信道信息；
2)计算第二信道信息和目标信道信息的误差；
3)基于得到的误差计算损失函数，通过损失函数计算梯度信息，并将梯度信息反向传播；
4)通过梯度下降的方法更新第一信道学习模型和第二信道学习模型。
若更新第一信道学习模型和第二信道学习模型之后，损失函数的值小于预设门限值，则将上述梯度信息作为用于更新第一信道学习模型的一个或多个配置参数发送给第一通信装置。
若损失函数的值大于或等于预设门限值，则第二通信装置继续基于上述方法对第一信道学习模型和第二信道学习模型进行训练，直至损失函数的值小于预设门限值。进一步地，将多次训练过程中获得的多个梯度信息累加起来作为用于更新第一信道学习模型的一个或多个配置参数发送给第二通信装置。
其中，第一信道信息可以是第一通信装置向第二通信装置发送的。目标信道信息可以是第一通信装置向第二通信装置发送的，也可以是第二通信装置根据上下行信道的互易性获得的。
具体地，第一通信装置对信道学习模型进行训练的训练参数可以是预配置的，或者可以是第一通信装置指示的。
可选地，在用于更新信道学习模型的配置参数包括多个的情况下，第二通信装置可以采用相同的方法确定多个配置参数中的每一个，也可以采用不用的方法确定多个配置参数中的每一个。
可选地，第二通信装置可以采用一级或多级的方式确定用于更新信道学习模型的配置参数。以一级方式为例，第二通信装置可以采用上述方法中的一种或多种确定用于更新信道学习模型的配置参数；以多级方式为为例，第二通信装置可以先采用上述方法中的一种或多种确定用于更新信道学习模型的一部分配置参数，再采用上述方法中的一种或多种确定用于更新信道学习模型的另一部分配置参数。例如，第二通信装置根据第一参数从高层数据库中确定信道学习模型的结构，再根据第二参数从本地数据库中确定信道学习模型的维数、运算和/或函数，最后再根据来自第一通信装置的与配置参数相关的信息确定信道学习模型的变量。
可选地，若第二通信装置接收到多个第一通信装置发送的第一消息，则第二通信装置可以根据虚警概率确定第一信道学习模型是否适用。
例如当前小区中有J个第一通信装置进行信道学习模型训练，第二通信装置接收到J 1个第一通信装置发送的第一消息，其中J 2个第一通信装置反馈第一信道学习模型适用(或者J 2个第一消息指示第一信道学习模型适用)，J 3个第一通信装置反馈第一信道学习模型不适用(或者J 3个第一消息指示第一信道学习模型不适用)。
则虚警概率可以是反馈第一信道学习模型适用的第一通信装置数(或指示第一信道学习模型适用的第一消息数)占进行信道学习模型训练的第一通信装置数的比例(即J 2/J)；或者可以是反馈第一信道学习模型适用的第一通信装置数(或指示第一信道学习模型适用的第一消息数)占接收到第一消息的总数的比例(即J 2/J 1)；或者可以是反馈第一信道学习模型不适用的第一通信装置数(或指示第一信道学习模型不适用的第一消息数)占进行信道学习模型训练的第一通信装置数的比例(即J 3/J)；或者可以是反馈第一信道学习模型不适用的第一通信装置数(或指示第一信道学习模型不适用的第一消息数)占接收到第一消息的总数的比例(即J 3/J 1)；或者可以是反馈第一信道学习模型适用的第一通信装置数(或指示第一信道学习模型适用的第一消息数)与反馈第一信道学习模型不适用的第一通信装置数(或指示第一信道学习模型不适用的第一消息数)的比例(即J 2/J 3)。
具体的，若虚警概率大于或等于(或大于)预设门限#11，则第二通信装置确定第一信道学习模型适用；若虚警概率小于(或者小于或等于)预设门限#11，则第二通信装置确定第一信道学习模型不适用。或者，若虚警概率大于或等于(或大于)预设门限#12，则第二通信装置确定第一信道学习模型不适用；若虚警概率小于(或者小于或等于)预设门限#12，则第二通信装置确定第一信道学习模型适用。
例如，预设门限#11是70％，若指示第一信道学习模型适用的第一消息数与第二通信装置接收到的第一消息的总数之比大于70％，则第二通信装置确定第一信道学习模型适用。
又例如，预设门限#12是50％，若指示第一信道学习模型不适用的第一消息数与第二通信装置接收到的第一消息的综上数之比大于50％，则第二通信装置确定第一信道学习模型不适用。
第二通信装置通过上述方式确定信道学习模型是否适用，可以综合考虑多个第一通信装置的信道学习模型的情况，避免因为个别第一通信装置的反馈做出错误的决定，可以兼顾多个第一通信装置的信道学习模型的情况，提高确定信道学习模型是否适用的准确性，提高通信性能。
进一步地，第二通信装置确定第一信道学习模型不适用之后，则可以确定新的第一信道学习模型和新的第二信道学习模型，即确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数，并将用于更新第一信道学习模型的配置参数发送给第一通信装置。
可选地，在S210之前，方法200还可以包括：第一通信装置根据第一参数确定第一信道学习模型。
具体的，第一通信装置确定第一信道学习模型的方式可以参考上述实施例中关于确定信道学习模型的描述，为了简洁，本申请实施例不再赘述。
如下，本申请实施例提供了一种可以由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S210’至S230’中的至少一项：
S210’(即S210)，第一通信装置确定第一信道学习模型是否适用。具体的该步骤描述可以参考上文S210中的描述，为了简洁，此处不再详述。
S220’(即S220)，第一通信装置发送第一消息，第一消息用于指示第一信道学习模型是否适用。具体的该步骤描述可以上文S220中的描述，为了简洁，此处不再详述。
S230a’，第一通信装置发送用于更新第二信道学习模型的一个或多个配置参数。具体的该步骤描述可以上文S220中的描述，为了简洁，此处不再详述。可选的，该步骤可以由S230b’代替
S230b’，第一通信装置接收用于更新第一信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文方法200中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以包括S210”至S230”中的至少一项：
S210”(即S220)，第二通信装置接收第一消息，第一消息用于指示第一信道学习模型是否适用。具体的该步骤描述可以参考上文S210中的描述，为了简洁，此处不再详述。
S220”(即S220)，第二通信装置根据第一消息确定第一信道学习模型是否适用。具体的该步骤描述可以参考上文S220中的描述，为了简洁，此处不再详述。
S230a”，第二通信装置接收用于更新第二信道学习模型的一个或多个配置参数。具体的该步骤描述可以上文S220中的描述，为了简洁，此处不再详述。可选的，该步骤也可以由S230b”代替。
S230b”，第二通信装置发送用于更新第一信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文方法200中的描述，为了简洁，此处不再详述。
本申请实施例提供了一种第一通信装置确定信道学习模型不适用并告知第二通信装置的通信方法，本方法可以实现当信道学习模型不适用时及时上报告知第二通信装置，进而实现对信道学习模型的更新，从而可以提高信道学习模型的准确性和通信性能。
图12示出了本申请另一实施例提供的通信的方法。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。图12示出的方法300可以包括S310至S320。下面详细说明方法300中的各个步骤。
S310，第一通信装置确定第一信道学习模型是否适用。
第一信道学习模型用于基于目标信道信息确定第一信道信息，第一信道信息的数据量小于目标信道信息的数据量，因此也可以说第一信道学习模型用于对目标信道信息进行压缩以获得第一信道信息。
可选地，信道信息的数据量可以是指信道信息的维度。
例如，假设发送端(例如，可以是第一通信装置或第二通信装置)的天线端口数为A 2，接收端(例如，可以l是第一通信装置或第二通信装置)的天线端口数为A 3，则发送端和接收端间的目标信道信息可以是A 2*A 3维的矩阵，则目标信道信息的数据量可以用A 2*A 3表示。若目标信道信息的矩阵中的元素为复数，且每个元素的实部和虚部分开表示，则目标信道信息的数据量也可以表示为A 2*A 3*2。
例如，若目标信道信息的矩阵经过第一信道学习模型处理得到的第一信道信息的矩阵的维度为B 2，则第一信道信息的数据量可以用B 2表示。
可选的，信道信息的数据量也可以是指信道信息所包含的信息量等。
可选的，目标信道信息可以看做第一信道学习模型的输入，第一信道信息可以看做第一信道学习模型的输出。目标信道信息的数据量可以为输入的信息维度，第一信道信息的数据量可以为输出的信息维度。第一信道信息用于通过第二信道学习模型获得第二信道信息，第二信道信息与目标信道信息的数据量相同或相近。可选的，第二信道信息可以用于进行数据传输，例如第二通信装置可以根据第二信道信息确定数据传输的调度信息等，或者确定数据传输的预编码等。
应理解，第一信道学习模型于第二信道学习模型是对应的，因此，第一通信装置确定第一信道学习模型是否适用可以理解为，第一通信装置确定第一信道学习模型和第二信道学习模型是否适用。即，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定第二信道学习模型也不适用；第一通信装置在确定第一信道学习模型适用的情况下，可以确定第二信道学习模型也适用。
本申请实施例对目标信道信息不做限定。例如，在第一通信装置是终端设备的情况下，目标信道信息可以是下行信道信息；又例如，在第一通信装置是网络设备的情况下，目标信道信息可以是上行信道信息。又例如，在第一通信装置是终端设备的情况下，目标信道信息可以是上行信道信息，或者，目标信道信息可以是上行信道信息和下行信道信息，第一通信装置可以基于上下行信道的部分互易性，根据上行信道信息和下行信道信息确定第一信道学习模型和/或第二信道学习模型是否适用，或者确定新的第一信道学习模型和/或第二信道学习模型。又例如，在第一通信装置是网络设备的情况下，目标信道信息可以是下行信道信息，或者，目标信道信息可以是上行信道信息和下行信道信息，基于上下行信道的部分互易性，根据上行信道信息和下行信道信息确定第一信道学习模型和/或第二信道学习模型是否适用，或者确定新的第一信道学习模型和/或第二信道学习模型。
第一通信装置可以周期性地确定第一信道学习模型是否适用，例如，第一通信装置在第i次确定第一信道学习模型是否适用之后启动定时器，在定时器超时的情况下，第一通信装置第i+1次确定第一信道学习模型是否适用。
该方式下，可以降低第一通信装置与第二通信装置之间交互信令的开销，并且第一通信装置定时地确定第一信道学习模型是否适用，可以避免第一信道学习模型不适用的情况下导致的通信性能的下降。
本申请实施例对第一通信装置确定第一信道学习模型是否适用的方法不做限定。第一通信装置可以采用如下实现方式中的一种或多种确定第一信道学习模型是否适用。在第一通信装置确定第一信道学习模型不适用的情况下，第一通信装置和/或第二通信装置可以及时调整第一信道学习模型，提高信道学习模型的准确性和适用性，进而提升通信性能。
下文提供了第一通信装置确定第一信道学习模型和/或第二信道学习模型是否适用的方式，第一通信装置确定信道学习模型是否适用的方法可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请实施例对此不做限定。如下一种或多种确定信道学习模型是否适用的方式可以单独使用，也可以联合使用，具体的，本申请实施例对此不做限定。
需要说明的是，本申请实施例中提及的信道学习模型是否适用，也可以是指信道学习模型是否匹配、信道学习模型是否准确、信道学习模型是否过时的或信道学习模型是否错误等。
在一种实现方式中，第一通信装置可以根据目标信道的长期统计特性确定信道学习模型是否适用。例如，第一通信装置可以根据目标信道的长期统计特性的变化量是否大于或等于第一预设阈值确定第一信道学习模型是否适用。具体地，可以参见上文S210中的描述，为了简洁，此处不再赘述。
在另一种实现方式中，第一通信装置可以根据接收的第一调度信息确定第一信道学习模型是否适用。第一调度信息是第二通信装置根据第二信道信息发送的。具体地，可以参见上文S210中的描述，为了简洁，此处不再赘述。
在又一种实现方式中，第一通信装置可以根据数据传输性能确定第一信道学习模型是否适用。具体地，可以参见上文S210中的描述，为了简洁，此处不再赘述。
在又一种实现方式中，第一通信装置可以根据所处的场景是否发生变化确定第一信道学习模型是否适用。具体地，可以参见上文S210中的描述，为了简洁，此处不再赘述。
在又一种实现方式中，第一通信装置可以根据第一信道学习模型的性能指标是否小于第二预设阈值确定第一信道学习模型是否适用。具体地，可以参见上文S210中的描述，为了简洁，此处不再赘述。
在又一种实现方式中，第一通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用。具体地，可以参见上文S210中的描述，为了简洁，此处不再赘述。
在又一种实现方式中，第一通信装置可以根据是否接收到第一指示信息，确定第一信道学习模型是否适用。
第一指示信息用于指示第一通信装置进行信道学习模型训练。
第一通信装置可以在接收第一指示信息的情况下，确定第一信道学习模型不适用。
第一指示信息是第二通信装置发送的，第二通信装置可以在确定第二信道学习模型不适用的情况下，发送第一指示信息。
该实现方式下，第一通信装置可以根据是否接收到第一指示信息确定第一信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第一信道学习模型是否适用，可以降低第一终端设备的处理复杂度。同时根据是否接收到第一指示信息，可以在两个通信装置有统一的理解和认识的情况下确定第一信道学习模型是否适用，因此该方式简单快速。
可选的，本申请实施例中，第一通信装置或第二通信装置确定第一信道学习模型不适用时，因为第一信道学习模型和第二信道学习模型是对应的，因此可以推断出，第一通信装置或第二通信装置确定第二信道学习模型不适用。
可选的，本申请实施例中，第一通信装置或第二通信装置确定第二信道学习模型不适用时，因为第一信道学习模型和第二信道学习模型是对应的，因此可以推断出，第一通信装置或第二通信装置确定第一信道学习模型不适用。
可选的，本申请实施例中，第一通信装置或第二通信装置确定第一信道学习模型或第二信道学习模型不适用时，因为第一信道学习模型和第二信道学习模型是对应的，因此可以推断出，第一通信装置或第二通信装置确定第一信道学习模型和第二信道学习模型不适用。
本申请实施例对第二通信装置确定第二信道学习模型是否适用的方法不做限定。第二 通信装置可以采用如下实现方式中的一种或多种确定第二信道学习模型是否适用。在第二通信装置确定第二信道学习模型不适用的情况下，第一通信装置和/或第二通信装置可以及时调整第二信道学习模型，提高信道学习模型的准确性和适用性，进而提升通信性能。
下文提供了第二通信装置确定第一信道学习模型和/或第二信道学习模型是否适用的方式，第二通信装置确定信道学习模型是否适用的方法可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请实施例对此不做限定。如下一种或多种确定信道学习模型是否适用的方式可以单独使用，也可以联合使用，具体的，本申请实施例对此不做限定。
需要说明的是，本申请实施例中提及的信道学习模型是否适用，也可以是指信道学习模型是否匹配、信道学习模型是否准确、信道学习模型是否过时的或信道学习模型是否错误等。
在一种实现方式中，第二通信装置可以根据目标信道的长期统计特性确定信道学习模型是否适用。例如，当第二通信装置确定目标信道的长期统计特性变化较大时，表明第一通信装置和第二通信装置间的信道特征或者信道环境发生了较大的变化，因此第二通信装置可以确定信道学习模型不适用。
例如，第二通信装置可以根据目标信道的长期统计特性的变化量是否大于或等于第三预设阈值确定第一信道学习模型是否适用。其中，第三预设阈值可以是协议预定义的，也可以是预配置的，本申请实施例对此不做限定。
其中，目标信道的长期统计特性可以包括如下至少一项：rank值、大尺度特性、信道协方差矩阵、信道相关矩阵、相干时间、相干带宽等。
信道的大尺度特性可以为以下一项或多项：时延扩展、多普勒扩展、多普勒频移、平均信道增益和平均时延、接收到达角、到达角扩展、发射离开角、离开角扩展、空间接收参数和空间相关性。
作为一个示例，第二通信装置可以根据来自第一通信装置的信号确定目标信道的长期统计特性，并进一步根据目标信道的长期统计特性确定第二信道学习模型是否适用。其中，来自第二通信装置的信号可以是参考信号或数据信号，参考信号可以是DMRS、CSI-RS、PTRS、TRS、SBB、SRS等，数据信号可以是在PDSCH、PDCCH、PUSCH或PUCCH上传输的信号。
例如，若第一通信装置为终端设备，第二通信装置为网络设备，则网络设备可以根据终端设备发送的参考信号和/或数据信号确定目标信道的长期统计特性。其中，参考信号可以是SRS，DMRS，PTRS中至少一项，数据信号可以是在PUSCH上传输的信号、在PUCCH上传输的信号中的至少一项。目标信道可以是指下行信道，网络设备可以根据上行信道的长期统计特性推断下行信道的长期统计特性。
作为另一个示例，第二通信装置可以根据第一通信装置反馈的信道状态信息确定目标信道的长期统计特性的变化量。
例如第一通信装置上报的信道状态信息(channel state information，CSI)中可以包括如下至少一项：rank值、信道协方差矩阵、相关矩阵等。第二通信装置可以根据第一通信装置的上报信息确定目标信道的长期统计特性。
在目标信道的长期统计特性的变化量大于或等于第三预设阈值的情况下，表明第一通 信装置和第二通信装置之间的信道特征或信道环境发生了较大的变化，因此第二通信装置可以确定第二信道学习模型不适用；在目标信道的长期统计特性小于第三预设阈值的情况下，表明第一通信装置和第二通信装置之间的信道特征或信道环境比较稳定，因此第二通信装置可以确定第二信道学习模型适用。
以目标信道的长期统计特性是rank值为例，第二通信装置在确定rank值变化较大的情况下，可以确定第二信道学习模型不适用。例如，第二通信装置可以在确定rank值的变化量大于或等于预设门限#12(第三预设阈值的一例)的情况下，确定第二信道学习模型不适用。其中，预设门限#12可以为R 1，其中R 1为正整数。例如R 1为2，即当rank值的变化大于或等于2时，第二通信装置可以确定第二信道学习模型不适用。例如第一通信装置从反射体稀疏的环境运动到反射体丰富的环境，目标信道的径增多，进而rank值会发生变化，在此情况下，第二信道学习模型有可能不再适用。rank值越大，使用的信道学习模型可能越复杂，例如，信道学习模型的层数可能越高。例如当第一通信装置从室内变为室外，目标信道的长期统计特性(rank值)也会发生变化，则第二通信装置可以确定第二信道学习模型有可能不适用。
以目标信道的长期统计特性为多普勒频移为例，第二通信装置在确定多普勒频移变化较大的情况下，可以确定第二信道学习模型不适用。例如，第二通信装置可以在多普勒频移的变化量大于预设门限#13(第三预设阈值的一例)的情况下，确定第二信道学习模型不适用。其中，预设门限#13可以为F 2，其中F 2为实数。例如F 2为2，即当多普勒频移的变化量大于或等于2时，第二通信装置可以确定第二信道学习模型不适用。例如多普勒频移可以反映第一通信装置的运动速度，当第一通信装置从步行变为车载时，第一通信装置的移动速度变大，则第二信道学习模型有可能不再适用。例如，若第一通信装置的运动速度的变化量大于预设门限#14，则第二通信装置确定第二信道学习模型不适用。
该实现方式下，第二通信装置可以根据目标信道的长期统计特性确定信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第二信道学习模型是否适用，可以降低第二通信装置的处理复杂度，同时也不需要第一信装置的协助即可确定信道学习模型是否适用，降低信令交互，因此该方式简单快速。
在另一种实现方式中，第二通信装置可以根据第一调度信息确定第二信道学习模型是否适用。第一调度信息可以是第二通信装置根据第二信道信息确定的。
其中，第一调度信息可以包括如下至少一项：MCS指示、TBS指示、rank指示、天线端口指示等。第一调度信息可以是物理层的下行控制信息(downlink control information，DCI)，也可以是高层信令中的调度信息。具体地，第一调度信息包含的内容和指示方式可以参考现有技术，为了简洁，本申请实施例不再详述。
第二通信装置可以根据多次确定的第一调度信息的变化量确定第二信道学习模型是否适用。例如，第二通信装置可以根据前B’次确定的第一调度信息与当前确定的第一调度信息的差异，确定第二信道学习模型是否适用。在多次确定的第一调度信息的变化量大于或等于预设差异门限的情况下，第二通信装置可以确定第二信道学习模型不适用。B’为整数。
数据的调度信息可以反映出信道的质量情况，因此第二通信装置可以根据多次调度信息的差异确定第二信道学习模型是否适用。例如，当第二通信装置前B’次指示给第一通 信装置的第一调度信息和当前指示的第一调度信息的差异较大时，则表明第一通信装置的信道变化较大，即第二信道学习模型不适用。因此第二通信装置可以根据如下方式中的至少一种确定第二信道学习模型是否适用。
作为一个示例，第二通信装置可以根据前B’次确定的第一调度信息与当前确定的第一调度信息的差异值是否大于或等于预设差异门限，确定第二信道学习模型是否适用。其中，预设差异门限可以是协议预定义的，也可以是第一通信装置向第二通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
例如，第二通信装置在前B’次确定的第一调度信息与当前确定的第一调度信息的差异值大于或等于预设差异门限的情况下，确定第二信道学习模型不适用；第二通信装置在前B’次确定的第一调度信息与当前确定的第一调度信息的差异值小于预设差异门限的情况下，确定第二信道学习模型适用。
预设差异门限可以是调制阶数的差异门限、码率的差异门限、MCS索引(index)值的差异门限、rank值的差异门限、TBS的差异门限、天线端口的差异门限中的至少一种。
以预设差异门限是调制阶数的差异门限为例，调制阶数的差异门限可以为N 3阶，N 3为实数，例如N 3为1,2,3,4,1/2,3/2,5/2等。在前B’次确定的调制方式的阶数与当前确定的调制方式的阶数的差异值大于或等于调制阶数的差异门限的情况下，第二通信装置确定第二信道学习模型不适用。
例如，若调制阶数的差异门限为2阶，第二通信装置前B’次确定的调制方式为QPSK(阶数为2)，当前确定的调制方式为64QAM)(阶数为4)，则第一通信装置确定第一信道学习模型不适用。
以预设差异门限是码率的差异门限为例，码率的差异门限可以是M 1，M 1为实数，例如M 1为200/1024,250/1024,300/1024,500/1024等。在前B’次确定的码率与当前确定的码率的差异值大于或等于码率的差异门限的情况下，第二通信装置确定第二信道学习模型不适用。
例如，若码率的差异门限为200/1024，第二通信装置前B’次确定的码率为400/1024，而第一调度信息中指示的码率为658/1024时，则第二通信装置可以确定第二信道学习模型不适用。
以预设差异门限是MCS索引值的差异门限为例，MCS索引值的差异门限可以是P 1，P 1为整数，例如P 1为1,2,3,4,5,6,8,10等。在前B次确定的MCS索引值与当前确定的MCS索引值的差异值大于或等于MCS索引值的差异门限的情况下，第二通信装置确定第二信道学习模型不适用。
例如，若MCS索引值的差异门限为4，第二通信装置前B’次确定的MCS索引值为4，而当前确定的MCS索引值为10，则第二通信装置可以确定第二信道学习模型不适用。
以预设差异门限是TBS的差异门限为例，TBS的差异门限可以是Q，Q为整数，例如Q为32,64,128,256,612,1024等。在前B’次确定的TBS与当前确定的TBS的差异值大于或等于TBS的差异门限的情况下，第二通信装置确定第二信道学习模型不适用。
例如，若TBS的差异门限为64，第二通信装置前B’次确定的TBS为288，而当前确定的TBS为522，则第二通信装置可以第二确定信道学习模型不适用。
应理解，上文仅以第二通信装置根据前B’次确定的第一调度信息与当前确定的第一 调度信息中的其中一项的差异值确定第二信道学习模型是否适用为例进行说明，可选地，第二通信装置还可以根据前B’次确定的第一调度信息与当前确定的第一调度信息中的多项的差异值确定第二信道学习模型是否适用。例如，第二通信装置可以在前B’次确定的码率与当前确定的码率的差异值大于或等于码率的差异门限，且前B’次确定的TBS与当前确定的TBS的差异值大于或等于TBS差异门限的情况下，确定第二信道学习模型不适用。
还应理解，上文仅以第二通信装置根据第一调度信息的差异值是否大于或等于预设差异门限确定第二信道学习模型是否适用为例进行说明。可选地，第二通信装置可以根据第一调度信息的相似性是否小于预设相似门限确定第二信道学习模型是否适用。其中，预设相似门限值可以是协议预定义的，也可以是第一通信装置向第二通信装置指示的，也可以是预配置的，本申请实施例对此不做限定。
第二通信装置还可以根据第一通信装置反馈的CQI信息与第一调度信息之间的差异确定第二信道学习模型是否适用。具体地，第二通信装置可以在第一通信装置反馈的CQI信息与第一调度信息之间的差异大于或等于预设差异门限的情况下，确定第二信道学习模型不适用。
例如，第二通信装置可以根据第一调度信息、第二通信装置反馈的CQI信息和第二映射关系确定第二信道学习模型是否适用，第二映射关系用于指示调度信息与信道学习模型是否适用的对应关系。
表12中示出了第二映射关系的一例。例如第二映射关系可以是下表中的一行或者多行，第二映射关系可以是协议预定义的，也可以是第二通信装置通过信令告知第一通信装置的，具体的，本申请实施例对此不做限定。
表12
Figure PCTCN2021100637-appb-000076
例如，若第一通信装置反馈的CQI信息指示的调制方式为QPSK，第一调度信息指示的调度方式为QPSK，则第二通信装置根据表12确定第二信道学习模型适用；又例如，若第一通信装置反馈的CQI信息指示的码率为2/3，第一调度信息指示的码率为3/4，则 第一通信装置根据表12可以确定第二信道学习模型不适用。
该实现方式下，第二通信装置可以根据第一调度信息确定信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第二信道学习模型是否适用，可以降低第二通信装置的处理复杂度。同时根据确定第一调度信息的确定第一信道学习模型是否适用，也可以保证基于第一信道学习模型下的通信性能。此外，在该实现方式中，不需要第一通信装置的协助，因此可以减小信令交互。
在又一种实现方式中，第二通信装置可以根据数据传输性能确定第二信道学习模型是否适用。具体地，可以参考S210中，关于第一通信装置根据数据传输性能确定第一信道学习模型是否适用的描述，为了简洁，此处不再详述。
数据传输性能可以是指与第二通信装置通信的某一个第一通信装置的数据传输性能，根据数据传输性能可以确定与第一通信装置对应的第二信道学习模型是否适用。对第一通信装置对应的第二信道学习模型，即与第一通信装置侧部署的第一信道学习模型对应的第二信道学习模型。
数据传输性能也可以是指第二通信装置的小区的数据传输性能，例如小区边缘吞吐量，小区中心吞吐量，小区总吞吐量等。通过该小区的数据传输性能，第二通信装置可以确定该小区对应的第二信道学习模型是否适用。与小区对应的第二信道学习模型，即与小区中的第一通信装置对应的第二信道学习模型。
该实现方式下，第二通信装置可以根据数据传输性能确定第二信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第二信道学习模型是否适用，可以降低第二通信装置的处理复杂度。同时根据数据传输性能确定第二信道学习模型是否适用的方式是以最终的通信性能来衡量第二信道学习模型的准确性，因此也可以保证基于该第二信道学习模型下的通信性能，即该方式有助于提高通信性能。
在又一种实现方式中，第二通信装置可以根据第一通信装置所处的场景是否发生变化确定第一信道学习模型是否适用。
其中，场景可以是以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、V2X场景、3GPP协议定义的场景等。
不同的场景可以对应不同的信道环境，进而导致第二信道学习模型不适用。例如有些场景下，第一通信装置和第二通信装置的信道为直射径；而有些场景下，第一通信装置和第二通信装置的信道为非直射径。又例如，有些场景下，第一通信装置与第二通信装置之间的反射体比较少，相对应地，第一通信装置与第二通信装置之间的信道简单；而有些场景下，第一通信装置与第二通信装置之间的反射体比较多，相对应地，第一通信装置与第二通信装置之间的信道复杂。
因此，在第一通信装置所处的场景发生变化的情况下，第二信道学习模型可能不适用。例如，若第一通信装置从室内运动到室外，则第二通信装置可以确定第二信道学习模型不适用；又例如，第一通信装置从宏站运动到微站，则第二通信装置可以确定第二信道学习模型不适用。
可选的，第一通信装置可以向第二通信装置发送位置信息，第二通信装置根据接收到的位置信息确定第一通信装置所处的场景，进而根据第一通信装置所处的场景确定第二信道学习模型是否适用。
可选的，第一通信装置可以向第二通信装置发送场景信息，第二通信装置根据接收到的第一通信装置所处的场景确定第二信道学习模型是否适用。
该实现方式下，第二通信装置可以根据第一通信装置所处的场景确定第二信道学习模型是否适用，即在不进行信道学习模型训练的情况下，确定第二信道学习模型是否适用，可以降低第二通信装置的处理复杂度，且该方式简单快捷。
可选地，在第二通信装置是终端设备的情况下，第二通信装置可以根据自身所处的场景确定第二信道学习模型是否适用。
在又一种实现方式中，第二通信装置可以根据第二信道学习模型的性能指标确定第二信道学习模型是否适用。
作为一个示例，第二通信装置可以根据目标信道信息(或者第一信道信息)与第二信道信息的差异性或者相似性确定信道学习模型是否适用。例如，第二通信装置可以根据第二信道学习模型对目标信道信息进行压缩以及解压得到第二信道信息。例如，第二通信装置可以根据第二信道学习模型对第一信道信息进行解压得到第二信道信息。通过比较目标信道信息和第二信道信息的特征，或者通过比较第一信道信息和第二信道信息的特征，进而确定第二信道学习模型的误差，从而确定第二信道学习模型是否适用。具体地，可以参考S210中，关于第一通信装置根据第一信道学习模型的性能指标确定第一信道学习模型是否适用的描述，为了简洁，此处不再详述。
其中，目标信道信息可以是第一通信装置向第二通信装置发送的，也可以是第二通信装置根据上下行信道的互易性获得的。
其中，第一信道信息可以是第一通信装置向第二通信装置发送的。
该实现方式下，第二通信装置可以根据信道学习模型的性能指标确定第二信道学习模型是否适用，即通过进行信道学习模型训练确定第二信道学习模型是否适用，该方式可以不依赖于第一通信装置的辅助，第二通信装置即可确定第二信道学习模型是否适用，因此该方式简单快捷
作为另一个示例，第二通信装置可以根据目标信道信息与第二信道信息的误差确定第二信道学习模型是否适用。具体地，可以参考S210中，关于第一通信装置可以根据目标信道信息与第二信道信息的误差确定第一信道学习模型是否适用，为了简洁，此处不再详述。
其中，目标信道信息可以是第一通信装置向第二通信装置发送的，也可以是第二通信装置根据上下行信道的互易性获得的。
该实现方式下，第二通信装置可以根据目标信道信息与第二信道信息的误差确定第二信道学习模型是否适用，即通过进行信道学习模型训练确定第二信道学习模型是否适用，该方式可以不依赖于第一通信装置的辅助，第二通信装置即可确定信道学习模型是否适用，因此该方式简单快捷。在同时考虑目标信道信息和第二信道信息的特征的情况下确定第二信道学习模型是否适用，即可以保证第一通信装置和第二通信装置获得相同或相近的信道信息的特征，有助于后续利用该信道信息进行数据传输时提高数据传输的性能。
在又一种实现方式中，第二通信装置可以根据第一通信装置发送的信道学习模型的性能指标确定第二信道学习模型是否适用。其中，第一通信装置确定信道学习模型的性能指标的方式可以参考S210中的描述，为了简洁，此处不再详述。
该实现方式下，第二通信装置可以根据第一通信装置发送的信道学习模型的性能指标确定第二信道学习模型是否适用，即第二通信装置不需要进行信道学习模型训练即可确定第二信道学习模型是否适用，降低了第二通信装置的处理复杂度，因此，该方式简单快捷。
S320，第一通信装置发送第一消息，第一消息用于指示用于更新第二信道学习模型的一个或多个配置参数。相应地，在S320中，第二通信装置接收第一消息，并根据第一消息确定第二信道学习模型不适用，并且更新第二信道学习模型。
可选的，第一消息中也可以包括用于更新第一信道学习模型的一个或多个配置参数。
可选的，用于更新第一信道学习模型的一个或多个配置参数也可以通过其他消息发送，其他消息与第一消息不同，具体的，本申请实施例对此不做限定。
第一通信装置在确定第一信道学习模型不适用的情况下，可以确定新的第一信道学习模型和新的第二信道学习模型，并将新的第二信道学习模型的一个或多个配置参数发送给第二通信装置，以用于更新第二信道学习模型。可选地，第一通信装置可以将新的第二信道学习模型的配置参数中不同于先前的第二信道学习模型的一个或多个配置参数发送给第二通信装置，以用于更新先前的第二信道学习模型。
可选的，第一通信装置在确定第一信道学习模型不适用的情况下，可以确定新的第二信道学习模型，并将新的第二信道学习模型的一个或多个配置参数发送给第二通信装置，以用于更新第二信道学习模型，并将用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置。
该方式下，第一消息包括用于更新第二信道学习模型的一个或多个配置参数。相应地，第二通信装置可以根据第一消息更新第二信道学习模型。一个或多个配置参数可以包括如下至少一项：模型类型、模型结构、模型算法、模型权值向量、模型权值矩阵、模型偏置向量、模型偏置矩阵、模型激活函数。
其中，模型的类型包括机器学习的算法、神经网络模型或自动编码模型。
例如，在信道学习模型是机器学习的算法的情况下，配置参数可以包括模型算法，用于指示信道学习模型具体是机器学习算法中的哪一种算法。
又例如，在信道学习模型是神经网络模型的情况下，配置参数可以包括模型结构，用于指示信道学习模型具体是哪一种神经网络，以及模型结构还可以包括输入层维数、输出层维数、隐藏层层数、隐藏层神经元数、训练算法、损失函数中的一项或多项。配置参数可以包括变换算法、权值矩阵权值向量、偏置向量、偏置矩阵、激活函数中的一项或多项。
再例如，在信道学习模型是卷积神经网络模型的情况下，配置参数可以包括模型结构，用于指示各层的数目和/或各层的先后顺序。配置参数还可以包括以下一项或多项：针对数据输入层的参数(预处理操作算法、输入数据的维数、输入数据的取值范围)、针对卷积层的参数(输入单元的大小、感受域、步幅、补零的数量、深度、输出单元的深度、权值矩阵)、针对激励层的参数(激活函数)、针对池化层的参数(池化算法、空间范围、步幅、输入单元的大小、输出单元的大小)、针对全连接层的参数(权值矩阵、权值向量、偏置矩阵、偏置向量)。
本申请实施例对第一通信装置确定用于更新第一信道学习模型和第二信道学习模型的一个或多个配置参数的方式不做限定。即本申请实施例对第一通信装置确定第一信道学习模型和/或第二信道学习模型的方式不做限定。以下实施例提供了第一通信装置确定第 一信道学习模型和/或第二信道学习模型的方式，第一通信装置确定信道学习模型的方式可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请对此不做限定。如下一种或多种确定信道学习模型的方式方式可以单独使用，也可以联合使用，具体的，本申请对此不做限定。
作为一个示例，第一通信装置可以根据第一参数确定第一信道学习模型和第二信道学习模型。
其中，第一参数包括如下至少一项：终端设备所在的小区的小区标识、终端设备所在的场景、终端设备的类型、终端设备所在的地理位置。
可选地，在第一通信装置是终端设备的情况下，第一通信装置可以根据自身所在的小区、所在的场景、类型或所在的地理位置确定第一参数。
可选地，在第一通信装置是网络设备、第二通信装置是终端设备的情况下，第一通信装置可以接收第二通信装置发送的与第一参数相关的信息，并确定第一参数。
终端设备所在的场景可以是室内、室外、郊区、城镇、外界环境(例如，白天、晚上、晴天、阴天、交通顺畅、交通拥堵)等。例如，终端设备处于室内，则第一通信装置可以确定与室内场景对应的第一信道学习模型和/或第二信道学习模型，进一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
终端设备的类型可以是终端设备的天线端口数、处理能力等。
终端设备所在的地理位置可以是三维坐标、二维坐标、定位数据等。如图11所示，若终端设备所在的地理位置是区域1，则第一通信装置可以确定与区域1对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；若终端设备所在的地理位置是区域2，则第一通信装置可以确定与区域2对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
可选地，在第一通信装置是网络设备的情况下，第一通信装置向第二通信装置发送的用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数可以是终端级、小区级或终端组级。以终端级为例，第一通信装置以单播的方式分别向每个第二通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以小区级为例，第一通信装置以广播的方式向小区内的多个第二通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以终端组级为例，第一通信装置以广播或组播的方式向终端组内的多个第二通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
具体地，第一通信装置可以根据第一参数从数据库#1保存的配置参数集中确定第一信道学习模型和第二信道学习模型。
其中，数据库#1可以是保存信道学习模型相关信息或者相关数据的数据库，例如数据库#1可以保存信道学习模型的配置参数等。
数据库#1可以是第一通信装置的本地数据库，也可以是存储在高层的数据库。
以数据库#1是存储在高层的数据库为例，存储在高层的数据库可以存储在移动性管理单元、核心网、云端、中央管理器、运营商系统、第一通信装置群组管理系统、数据中心中的数据库。在此情况下，第一通信装置可以与高层进行通信交互，并从高层的数据库 中确定信道学习模型的配置参数。例如，第一通信装置可以从高层数据库下载和/或读取数据库中存储的信道学习模型的配置参数，进而确定第一信道学习模型和/或第二信道学习模型。或者，第一通信装置可以接收高层网元发送的关于信道学习模型的配置参数，进而确定第一信道学习模型和/或第二信道学习模型。
以数据库#1是存储在第一通信装置的本地数据库为例，，第一通信装置可以从本地数据库下载和/或读取数据库中存储的信道学习模型的配置参数，进而确定第一信道学习模型和/或第二信道学习模型。
第一通信装置的本地数据库中存储的信道学习模型的配置参数可以是预配置的，也可以是第一通信装置侧通过训练学习确定并且存储在本地数据库中的。第一通信装置可以在应用信道学习模型之前，根据第一参数从本地数据库中确定信道学习模型。
该实现方式下，第一通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型，并进一步确定用于更新第二信道学习模型的配置参数，即不进行信道学习模型训练即可确定新的信道学习模型，降低了第一通信装置的处理复杂度。
作为另一个示例，第一通信装置可以对第一信道学习模型和第二信道学习模型进行训练，以获得第一信道学习模型和第二信道学习模型，并进一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
第一通信装置对第一信道学习模型和第二信道学习模型进行训练的方式可以是：
1)基于接收的参考信号获得目标信道信息；
2)根据目标信道信息和第一信道学习模型确定第一信道信息；
3)根据第一信道信息和第二信道学习模型确定第二信道信息；
4)计算第二信道信息和目标信道信息的误差；
5)基于得到的误差计算损失函数，通过损失函数计算梯度信息，并将梯度信息反向传播；
6)通过梯度下降的方法更新第一信道学习模型和第二信道学习模型。
若更新第一信道学习模型和第二信道学习模型之后，损失函数的值小于预设门限值，则将上述梯度信息作为用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置。
若损失函数的值大于或等于预设门限值，则第一通信装置继续基于上述方法对第一信道学习模型和第二信道学习模型进行训练，直至损失函数的值小于预设门限值。进一步地，将多次训练过程中获得的多个梯度信息累加起来作为用于更新第二信道学习模型的一个或多个配置参数发送给第二通信装置。
具体地，第一通信装置对信道学习模型进行训练的训练参数可以是预配置的，或者可以是根据来自第二通信装置的指示信息确定的。
可选地，在用于更新信道学习模型的配置参数包括多个的情况下，第一通信装置可以采用相同的方法确定多个配置参数中的每一个，也可以采用不用的方法确定多个配置参数中的每一个。
可选地，第一通信装置可以采用一级或多级的方式确定用于更新信道学习模型的配置参数。以一级方式为例，第一通信装置可以采用上述方法中的一种或多种确定用于更新信道学习模型的配置参数；以多级方式为为例，第一通信装置可以先采用上述方法中的一种 或多种确定用于更新信道学习模型的一部分配置参数，再采用上述方法中的一种或多种确定用于更新信道学习模型的另一部分配置参数。例如，第一通信装置根据第一参数从高层数据库中确定信道学习模型的结构，再根据第一参数从本地数据库中确定信道学习模型的维数、运算和/或函数，最后再根据来自第二通信装置的与配置参数相关的信息确定信道学习模型的变量。
如下，本申请实施例提供了一种可以由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S310’至S320’中的至少一项：
S310’(即S310)，第一通信装置确定第一信道学习模型是否适用。具体的该步骤描述可以参考上文S310中的描述，为了简洁，此处不再详述。
S320’(即S320)，第一通信装置发送第一消息，第一消息用于指示用于更新第二信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文S320中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以包括S310”至S320”中的至少一项：
S310”(即S320)，第二通信装置接收第一消息，第一消息用于指示用于更新第二信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文S320中的描述，为了简洁，此处不再详述。
S320”(即S320)，第二通信装置根据第一消息确定第二信道学习模型不适用，并且更新第二信道学习模型。具体的该步骤描述可以参考上文S320中的描述，为了简洁，此处不再详述。
本申请实施例提供了一种第一通信装置确定信道学习模型不适用并告知第二通信装置的通信方法，本方法可以实现当信道学习模型不适用时及时上报告知第二通信装置，进而实现对信道学习模型的更新，从而可以提高信道学习模型的准确性和通信性能。
图13示出了本申请另一实施例提供的通信的方法。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。图13示出的方法400可以包括S410至S420。下面详细说明方法400中的各个步骤。
S410，第二通信装置确定第二信道学习模型是否适用。
第二信道学习模型用于根据第一信道信息确定第二信道信息，第二信道信息与目标信道信息的数据量相同或相近。可选的，第二信道信息可以用于进行数据传输，例如第二通信装置可以根据第二信道信息确定数据传输的调度信息等，或者确定数据传输的预编码等。
其中，第一信道信息可以是根据第一信道学习模型和目标信道信息确定的，第一信道信息的数据量小于目标信道信息的数据量，因此也可以说第一信道学习模型用于对目标信道信息进行压缩以获得第一信道信息。
具体地，关于信道信息的数据量的描述可以参数上文S210中的描述，为了简洁，此处不再详述。
应理解，第一信道学习模型于第二信道学习模型是对应的，因此，第二通信装置确定第二信道学习模型是否适用可以理解为，第二通信装置确定第二信道学习模型和第一信道学习模型是否适用。即，第二通信装置在确定第二信道学习模型不适用的情况下，可以确 定第一信道学习模型也不适用；第二通信装置在确定第二信道学习模型适用的情况下，可以确定第一信道学习模型也适用。
第二通信装置可以周期性地确定第二信道学习模型是否适用，例如，第二通信装置在第i次确定第二信道学习模型是否适用之后启动定时器，在定时器超时的情况下，第二通信装置第i+1次确定第二信道学习模型是否适用。
该方式下，可以降低第一通信装置与第二通信装置之间交互信令的开销，并且第二通信装置定时地确定第二信道学习模型是否适用，可以避免第二信道学习模型不适用的情况下导致的通信性能的下降。
本申请实施例对第二通信装置确定第二信道学习模型是否适用的方法不做限定。具体地，第二通信装置确定第二信道学习模型是否适用的方法可以参考上文S310中的描述，为了简洁，此处不再详述。
S420，第二通信装置发送第二消息，第二消息用于指示用于更新第一信道学习模型的一个或多个配置参数。相应地，在S420中，第一通信装置接收第二消息，并根据第二消息确定第一信道学习模型不适用，并且更新第一信道学习模型。
可选的，第二消息中也可以包括用于更新第二信道学习模型的一个或多个配置参数。
可选的，用于更新第二信道学习模型的一个或多个配置参数也可以通过其他消息发送，所述其他消息与第二消息不同，具体的，本申请实施例对此不做限定。
第二通信装置在确定第二信道学习模型不适用的情况下，可以确定新的第一信道学习模型和新的第二信道学习模型，并将新的第一信道学习模型的一个或多个配置参数发送给第一通信装置，以用于更新第一信道学习模型。可选地，第二通信装置可以将新的第一信道学习模型的配置参数中不同于先前的第一信道学习模型的一个或多个配置参数发送给第一通信装置，以用于更新先前的第一信道学习模型。
可选的，第二通信装置在确定第二信道学习模型不适用的情况下，可以确定新的第一信道学习模型，并将新的第一信道学习模型的一个或多个配置参数发送给第二通信装置，以用于更新第一信道学习模型。
该方式下，第二消息包括用于更新第一信道学习模型的一个或多个配置参数。相应地，第一通信装置可以根据第二消息更新第一信道学习模型。一个或多个配置参数可以包括如下至少一项：模型类型、模型结构、模型算法、模型权值向量、模型权值矩阵、模型偏置向量、模型偏置矩阵、模型激活函数。
其中，模型的类型包括机器学习的算法、神经网络模型或自动编码模型。
例如，在信道学习模型是机器学习的算法的情况下，配置参数可以包括模型算法，用于指示信道学习模型具体是机器学习算法中的哪一种算法。
又例如，在信道学习模型是神经网络模型的情况下，配置参数可以包括模型结构，用于指示信道学习模型具体是哪一种神经网络，以及模型结构还可以包括输入层维数、输出层维数、隐藏层层数、隐藏层神经元数、训练算法、损失函数中的一项或多项。配置参数可以包括变换算法、权值矩阵权值向量、偏置向量、偏置矩阵、激活函数中的一项或多项。
再例如，在信道学习模型是卷积神经网络模型的情况下，配置参数可以包括模型结构，用于指示各层的数目和/或各层的先后顺序。配置参数还可以包括以下一项或多项：针对数据输入层的参数(预处理操作算法、输入数据的维数、输入数据的取值范围)、针对卷 积层的参数(输入单元的大小、感受域、步幅、补零的数量、深度、输出单元的深度、权值矩阵)、针对激励层的参数(激活函数)、针对池化层的参数(池化算法、空间范围、步幅、输入单元的大小、输出单元的大小)、针对全连接层的参数(权值矩阵、权值向量、偏置矩阵、偏置向量)。
本申请实施例对第二通信装置确定第一信道学习模型和第二信道学习模型的方式不做限定。
以下实施例提供了第二通信装置确定第一信道学习模型和/或第二信道学习模型的方式，第二通信装置确定信道学习模型的方式可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请实施例对此不做限定。如下一种或多种方式确定信道学习模型的方式可以单独使用，也可以联合使用，具体的，本申请实施例对此不做限定。
作为一个示例，第二通信装置可以根据第一参数确定新的第一信道学习模型和新的第二信道学习模型。
具体地，第二通信装置可以根据第一参数从数据库#2保存的配置参数集中确定第一信道学习模型和第二信道学习模型。
其中，第一参数包括如下至少一项：终端设备所在的小区的小区标识、终端设备所在的场景、终端设备的类型、终端设备所在的地理位置。
可选地，在第二通信装置是终端设备的情况下，第二通信装置可以根据自身所在的小区、所在的场景、类型或所在的地理位置确定第一参数。
可选地，在第二通信装置是网络设备、第一通信装置是终端设备的情况下，第二通信装置可以接收第一通信装置发送的与第一参数相关的信息，并确定第一参数。
终端设备所在的场景可以是室内、室外、郊区、城镇、外界环境(例如，白天、晚上、晴天、阴天、交通顺畅、交通拥堵)等。例如，终端设备处于室内，则第二通信装置可以确定与室内场景对应的第一信道学习模型和/或第二信道学习模型，进一步确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
终端设备的类型可以是终端设备的天线端口数、处理能力等。
终端设备所在的地理位置可以是三维坐标、二维坐标、定位数据等。如图11所示，若终端设备所在的地理位置是区域1，则第二通信装置可以确定与区域1对应的第一信道学习模型和第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；若终端设备所在的地理位置是区域2，则第二通信装置可以确定与区域2对应的第一信道学习模型和/或第二信道学习模型，并确定用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
可选地，在第二通信装置是网络设备的情况下，第二通信装置向第一通信装置发送的用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数可以是终端级、小区级或终端组级。以终端级为例，第二通信装置以单播的方式分别向每个第一通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以小区级为例，第二通信装置以广播的方式向小区内的多个第一通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数；以终端组级为例，第二通信装置以广播或组播的方式向终端组内的多个第一通信装置发送用于更新第一信道学习模型和/或第二信道学习模型的一个或多个配置参数。
其中，数据库#2可以是携带信道学习模型相关信息或者相关数据的数据库，例如数据库#2中可以保存信道学习模型的配置参数等。
数据库#2可以是第二通信装置的本地数据库，也可以是存储在高层的数据库。
以数据库#2是存储在高层的数据库为例，存储在高层的数据库可以是存储在移动性管理单元、核心网、云端、中央管理器、运营商系统、第二通信装置群组管理系统、数据中心中的数据库。在此情况下，第二通信装置可以与高层进行通信交互，并从高层的数据库中确定信道学习模型的配置参数。例如，第二通信装置可以根据第一参数从高层数据库下载和/或读取数据库中存储的信道学习模型，进而确定新的第一信道学习模型和新的第二信道学习模型。又例如，第二通信装置可以接收高层网元发送的关于信道学习模型的配置参数，进而确定新的第一信道学习模型和新的第二信道学习模型。
以数据库#2是存储在第二通信装置的本地的数据库为例，第二通信装置可以从本地数据库下载和/或读取数据库中信道学习模型，进而确定第一信道学习模型和第二信道学习模型。
第二通信装置的本地数据库中存储的信道学习模型的配置参数库预配置的，也可以是第二通信装置侧通过训练学习确定并且存储在本地数据库中的。第二通信装置可以在应用信道学习模型之前，根据第一参数从本地数据库确定信道学习模型。
该实现方式下，第二通信装置可以根据第一参数确定第一信道学习模型和/或第二信道学习模型，并进一步确定用于更新第一信道学习模型的配置参数，即不进行信道学习模型训练即可确定新的信道学习模型，降低了第二通信装置的处理复杂度。
作为另一个示例，第二通信装置可以对第一信道学习模型和第二信道学习模型进行训练，以获得第一信道学习模型和第二信道学习模型，并进一步确定用于更新第一信道学习模型的一个或多个配置参数。
第二通信装置对第一信道学习模型和第二信道学习模型进行训练的方式可以是：
1)根据第一信道信息和第二信道学习模型确定第二信道信息；
2)计算第二信道信息和目标信道信息的误差；
3)基于得到的误差计算损失函数，通过损失函数计算梯度信息，并将梯度信息反向传播；
4)通过梯度下降的方法更新第一信道学习模型和第二信道学习模型。
若更新第一信道学习模型和第二信道学习模型之后，损失函数的值小于预设门限值，则将上述梯度信息作为用于更新第一信道学习模型的一个或多个配置参数发送给第一通信装置。
若损失函数的值大于或等于预设门限值，则第二通信装置继续基于上述方法对第一信道学习模型和第二信道学习模型进行训练，直至损失函数的值小于预设门限值。进一步地，将多次训练过程中获得的多个梯度信息累加起来作为用于更新第一信道学习模型的一个或多个配置参数发送给第二通信装置。
其中，第一信道信息可以是第一通信装置向第二通信装置发送的。目标信道信息可以是第一通信装置向第二通信装置发送的，也可以是第二通信装置根据上下行信道的互易性获得的。
具体地，第一通信装置对信道学习模型进行训练的训练参数可以是预配置的，或者可 以是第一通信装置指示的。
可选地，在用于更新信道学习模型的配置参数包括多个的情况下，第二通信装置可以采用相同的方法确定多个配置参数中的每一个，也可以采用不用的方法确定多个配置参数中的每一个。
可选地，第二通信装置可以采用一级或多级的方式确定用于更新信道学习模型的配置参数。以一级方式为例，第二通信装置可以采用上述方法中的一种或多种确定用于更新信道学习模型的配置参数；以多级方式为为例，第二通信装置可以先采用上述方法中的一种或多种确定用于更新信道学习模型的一部分配置参数，再采用上述方法中的一种或多种确定用于更新信道学习模型的另一部分配置参数。例如，第二通信装置根据第一参数从高层数据库中确定信道学习模型的结构，再根据第二参数从本地数据库中确定信道学习模型的维数、运算和/或函数，最后再根据来自第一通信装置的与配置参数相关的信息确定信道学习模型的变量。
如下，本申请实施例提供了一种可以由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S410’至S420’中的至少一项：
S410’(即S420)，第一通信装置接收第二消息，第二消息用于指示用于更新第一信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文S420中的描述，为了简洁，此处不再详述。
S420’(即S420)，第一通信装置根据第二消息确定第一信道学习模型不适用，并更新第一信道学习模型。具体的该步骤描述可以参考上文S420中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以包括S410”至S420”中的至少一项：
S410”(即S410)，第二通信装置确定第二信道学习模型是否适用。具体的该步骤描述可以参考上文S410中的描述，为了简洁，此处不再详述。
S420”(即S420)，第二通信装置发送第二消息，第二信息用于指示用于更新第一信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文S420中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种通信的方法。下文中的实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。
如下，本申请实施例提供了一种由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S410*至S420*中的至少一项：
S410*，第一通信装置接收第一指示信息，第一指示信息用于指示进行信道学习模型训练。具体的该步骤描述可以参考上文S230中的描述，为了简洁，此处不再详述。
S420*，第一通信装置发送第一消息，第一消息用于指示用于更新信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考本申请中其他实施例的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以包括S410**至S430**中的至少一项：
S410**(即S410)，第二通信装置确定第二信道学习模型是否适用。具体的该步骤 描述可以参考上文S410中的描述，为了简洁，此处不再详述。
S420**，第二通信装置发送第一指示信息，第一指示信息用于指示进行信道学习模型训练。具体的该步骤描述可以参考上文S230中的描述，为了简洁，此处不再详述。
S430**，第二通信装置接收第一消息，第一消息用于指示用于更新第二信道学习模型的一个或多个配置参数。具体的该步骤描述可以参考上文S320中的描述，为了简洁，此处不再详述。
本申请实施例提供了一种第二通信装置确定信道学习模型不适用并告知第一通信装置的通信方法，本方法可以实现当信道学习模型不适用时及时上报告知第一通信装置，进而实现对信道学习模型的更新，从而可以提高信道学习模型的准确性和通信性能。
图14示出了本申请实施例提供的一种通信的方法的示意性流程图。具体的，本申请实施例中提供了通信中确定信道信息的方法的示意性流程图。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。在图14示出的方法中，以第一通信装置是终端设备、第二通信装置是网络设备为例进行说明。如图14所示，方法500可以包括S510至S550，下面详细说明各个步骤。
S510，终端设备和网络设备确定第一信道学习模型和第二信道学习模型。
具体的，本步骤中包括终端设备确定第一信道学习模型，或者，终端设备确定第一信道学习模型和第二信道学习模型。相应的，本步骤中包括网络设备确定第二信道学习模型，或者，网络设备确定第一信道学习模型和第二信道学习模型。
第一信道学习模型可以部署在终端设备侧，用于根据测得的下行信道信息确定第一信道信息，第一信道信息的数据量小于测得的下行信道信息的数据量。部署也可以是指在实际应用中，第一信道学习模型可以在终端设备侧存储，终端设备侧可以基于第一信道学习模型进行信息处理或计算。
第二信道学习模型可以部署在网络设备侧，用于根据第一信道信息确定第二信道信息，第二信道信息的数据量与测得的下行信道信息的数据量相同或相近。可选地，第二信道信息可以用于下行传输。部署也可以是指在实际应用中，第二信道学习模型可以在网络设备侧存储，网络设备侧可以基于第二信道学习模型进行信息处理或计算。
确定信道学习模型可以包括确定信道学习模型中的配置参数。
其中，信道学习模型中的配置参数包括模型的结构、模型的维数、模型中的运算、模型中的函数、模型中的变量、模型中的参数等中的一项或多项。
模型的结构可以包括该模型是神经网络、主成分分析、自动编码等结构中的一项或多项。模型的结构还可以包括模型的输入信息的特征、模型的输出信息的特征等。
其中，模型的输入信息的特征可以是指对高维信道信息(例如，测得的下行信道信息)进行运算之后的输入信息特征。对高维信道信息进行的运算可以是去均值、归一化运算、离散傅里叶变换、时延-角度域变换、或者实部和虚部分离等中的一项或者多项。输入信息可以是高维信道信息的实部和虚部联合的信息，也可以是高维信道信息的独立实部或独立虚部，或者可以是时延-角度域信息等。对高维信道信息进行运算也可以称为数据预处理。例如，确定该模型的输入信息的特征为归一化，时域-角度域变换，以及独立实部和虚部等。
其中，模型的输出信息的特征可以是指编码或者降维之后的信息特征。例如该编码或 者降维之后的信息可以是指归一化运算之后的信息、离散傅里叶变换之后的信息、时延-角度域变换之后的信息、实部和虚部联合的信息、或者实部和虚部分离之后的信息。
模型的维数可以是模型中的层数、每层的维数、输入层/隐藏层/输出层的维数等中的一项或多项。例如，确定该模型的维数为N层卷积神经网络，输入层的维数为N 1，输出层的维数为N 2。其中，N，N 1，N 2为正整数。
模型中的运算可以是线性运算、非线性运算等中的一项或多项。例如，确定该模型的运算为线性运算。该运算也可以用复杂度进行衡量或者表征。
模型中的函数可以是数学运算、逻辑运算等，例如加减乘除、加权求和、加权求和加偏置、激活函数等中的一项或多项。例如，确定该模型的函数为加权求和加偏置以及Relu激活函数。该激活函数可以是针对模型中的每一层分别确定，或可以一层或者多层采用相同的激活函数。
模型中的变量可以是指模型中涉及的参数的信息，例如可以是参数的个数、参数的取值范围、参数的取值、参数的类型等中的一项或多项。参数可以是常量，也可以是变量。例如，确定该模型的权值矩阵为W，偏置矩阵为b，激活函数λ的变量取值等。
第一信道学习模型可以包括一个或多个模型，对应的，第二信道学习模型也可以包括一个或多个模型。例如针对信道信息的实部有信道学习模型1，针对信道信息的虚部有信道学习模型2。
例如，第一信道学习模型的结构为卷积神经网络，则高维信道信息实部和虚部可以分别对应一个卷积神经网络，即卷积神经网络1对应的输入信息为高维信道信息的实部，编码输出信息为低维信道信息(例如，第一信道信息)的实部，卷积神经网络2对应的输入信息为高维信道信息的虚部，编码输出信息为低维信道信息的虚部。在高维信道信息的实部和虚部对应2个神经网络时，该2个神经网络的配置参数可以相同，也可以不同。
又例如，第二信道学习模型的结构为卷积神经网络，则低维信道信息实部和虚部可以分别对应一个卷积神经网络，即卷积神经网络3对应的输入信息为低维信道信息的实部，译码输出信息为高维信道信息(例如，第二信道信息)的实部，卷积神经网络4对应的输入信息为低维信道信息的虚部，译码输出信息为高维信道信息的虚部。在低维信道信息的实部和虚部对应2个神经网络时，该2个神经网络的配置参数可以相同，也可以不同。
确定信道学习模型可以是确定一个或多个信道学习模型的配置参数。
作为一个示例，网络设备确定第一信道学习模型和第二信道学习模型，并将第一信道学习模型的配置参数发送给终端设备。具体地，网络设备确定第一信道学习模型和第二信道学习模型的方式，可以参考上文S420中关于第二通信装置确定新的第一信道学习模型和新的第二信道学习模型的描述，为了简洁，此处不再详述。
作为另一个示例，终端设备确定第一信道学习模型和第二信道学习模型，并将第二信道学习模型的配置参数发送给网络设备。具体地，终端设备确定第一信道学习模型和第二信道学习模型的方式，可以参考上文S320中关于第一通信装置确定新的第一信道学习模型和新的第二信道学习模型的描述，为了简洁，此处不再详述。
可选地，在S510之前，方法500还可以包括：终端设备接收网络设备发送的配置信息#1，该配置信息#1用于指示发送第一信道学习模型和/或第二信道学习模型的配置参数的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容等中的一项或多项。相 应地，终端设备根据该配置信息#1可以确定发送第一信道学习模型和/或第二信道学习模型的配置参数的资源、码率、调制方式、比特数或反馈顺序。可选地，网络设备可以通过高层信令发送该配置信息#1，也可以通过物理层信令发送该配置信息#1。其中，高层信令可以是指RRC信令或MAC信令，物理层信令是指DCI信令。
可选地，在S510之前，方法500还可以包括：终端设备接收网络设备发送的配置信息#2，该配置信息#2用于指示接收第一信道学习模型和/或第二信道学习模型的配置参数的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容等中的一项或多项。相应地，终端设备根据该配置信息#2可以确定接收第一信道学习模型和/或第二信道学习模型的配置参数的资源、码率、调制方式、比特数或反馈顺序。
可选地，网络设备可以通过高层信令发送该配置信息#2，也可以通过物理层信令发送该配置信息#2。其中，高层信令可以是指RRC信令或MAC信令，物理层信令是指DCI信令。
作为另一个实例，终端设备确定第一信道学习模型，网络设备确定第二信道学习模型。应理解，在终端设备和网络设备基于相同的规则各自确定信道学习模型的情况下，终端设备确定的第一信道学习模型和网络设备确定的第二信道学习模型是对应的。例如，终端设备根据自身所处的场景确定第一信道学习模型，与网络设备根据终端设备所处的场景确定的第二信道学习模型是对应的。
例如，第一信道学习模型可以是对应自动编码模型中的编码模块，第二信道学习模型可以是对应自动编码模型中的解码模块。第一信道学习模型和第二信道学习模型是对应的。
S520，网络设备发送参考信号。相应地，在S520中，终端设备接收参考信号。
网络设备向终端设备发送参考信号的方式参考现有技术，为了简洁，本申请实施例不再赘述。
可选的，S520可以在S510之前执行，或者，S520可以在S510之后执行，本申请对此不做限定。
其中，参考信号可以为DMRS，CSI-RS，TRS，SSB等，本申请对此不做限定。
例如终端设备可以接收CSI-RS配置信息，根据配置信息可以周期性，半持续性，或者，非周期性接收CSI-RS。所述CSI-RS配置信息可以是指CSI测量配置信息，该CSI配置信息中可以包括CSI资源配置信息(CSI-ResourceConfig)，CSI上报配置信息(CSI-ReportConfig)中至少一项。
可选的，终端设备接收网络设备发送的CSI-RS配置信息，还可以包括终端设备接收网络设备发送的信道学习模型的配置参数。例如信道学习模型的配置参数可以包含在CSI-RS配置信息中。
具体的，信道学习模型的配置参数可以包含在CSI上报配置信息中。
可选的，终端设备可以基于网络设备发送的信道学习模型的配置参数确定信道学习模型。网络设备可以通过高层信令和/或物理层信令发送所述信道学习模型的配置参数。
可选的，S520和S510的前后顺序，本申请对此不做限定。例如，可以先确定信道学习模型再发送参考信号，或者先发送参考信号再确定信道学习模型。例如可以先确定信道学习模型再接收参考信号，或者先接收参考信号再确定信道学习模型。
S530，终端设备基于目标信道信息和第一信道学习模型确定第一信道信息。
其中，目标信道信息可以是终端设备基于接收的参考信号获得的下行信道信息。具体的，目标信道信息可以参考本申请中的描述，为了简洁，本申请实施例不再赘述。
终端设备获得下行信道信息的方式可以参考现有技术，为了简洁，本申请实施例不再赘述。
终端设备将测得的下行信道信息作为第一信道学习模型的输入数据，根据第一信道学习模型对下行信道信息进行编码或压缩，得到第一信道信息。
例如，若第一信道学习模型是神经网络模型，终端设备可以将下行信道信息的实部进行编码，得到低维信道信息的实部，再将下行信道信息的虚部进行编码，得到低维信道信息的虚部。在此情况下，第一信道信息可以包括低维信道信息的实部和低维信道信息的虚部。
例如，终端设备根据第一信道学习模型对64*1维复数的下行信道信息进行编码后，可以得到2维的实部信息和2维的虚部信息，则终端设备可以将2维的实部信息和2维的虚部信息进行量化后作为第一信道信息。
又例如，若第一信道学习模型是神经网络模型，终端设备可以将下行信道信息的实部和虚部联合编码，得到低维信道信息。在此情况下，第一信道信息可以包括低维实数信息。
例如，终端设备根据第一信道学习模型对64*1维复数的下行信道信息进行编码后，可以得到4维的实数信息，则终端设备可以将4维的实数信息进行量化后作为第一信道信息。
S540，终端设备发送第一信道信息。相应地，在S540中，网络设备接收第一信道信息。
终端设备发送第一信道信息的方式，可以参考现有技术中终端设备反馈CSI的方式，为了简洁，本申请实施例不再赘述。
可选地，在S540之前，方法500还可以包括：终端设备接收网络设备发送的配置信息#3，该配置信息#3用于指示发送第一信道信息的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容、第一信道信息的内容等中的一项或多项。相应地，终端设备根据该配置信息#2可以确定发送第一信道信息的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容或第一信道信息的内容。即，方法500包括：网络设备发送配置信息#3，终端设备接收配置信息#3。
可选地，网络设备可以通过高层信令发送该配置信息#3，也可以通过物理层信令发送该配置信息#3。其中，高层信令可以是指RRC信令或MAC信令，物理层信令是指DCI信令。
可选地，在终端设备确定第二信道学习模型的情况下，终端设备可以将第二信道学习模型的配置参数和第一信道信息一起发送，也可以将第二信道学习模型的配置参数和第一信道信分开发送。例如，终端设备在信道#1发送第二信道学习模型的配置参数，在信道#2发送第一信道信息；或者，终端设备在信道#3发送第二信道学习模型的配置参数和第一信道信息。其中，信道#1～信道#3可以是PUCCH、PUSCH或物理反馈信道(physical feedback channel，PFCH)，或者可以是其他上行信道。
可选地，在终端设备确定第一信道学习模型和第二信道学习模型的情况下，终端设备 可以将第一信道学习模型的配置参数，第二信道学习模型的配置参数和第一信道信息一起发送，也可以将第一信道学习模型的配置参数，第二信道学习模型的配置参数和第一信道信分开发送。例如，终端设备在信道#1发送第一信道学习模型的配置参数，第二信道学习模型的配置参数，在信道#2发送第一信道信息；或者，终端设备在信道#3发送第一信道学习模型的配置参数，第二信道学习模型的配置参数和第一信道信息。其中，信道#1～信道#3可以是PUCCH、PUSCH或PFCH，或者可以是其他上行信道。
可选地，在S540之前，方法500还可以包括：终端设备接收网络设备发送的配置信息#4，该配置信息#4用于指示发送第一信道信息和信道学习模型的配置参数的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容、第一信道信息的内容等中的一项或多项。相应地，终端设备根据该配置信息#4可以确定发送第一信道信息和信道学习模型的配置参数的资源、码率、调制方式、比特数、反馈顺序或第一信道信息的内容。即，方法500包括：网络设备发送配置信息#4，终端设备接收配置信息#4。
可选地，网络设备可以通过高层信令发送该配置信息#4，也可以通过物理层信令发送该配置信息#4。其中，高层信令可以是指RRC信令或MAC信令，物理层信令是指DCI信令。
可选地，在S540之前，方法500还可以包括：终端设备接收网络设备发送的配置信息#5，该配置信息#5用于指示发送信道学习模型的配置参数的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容等中的一项或多项。相应地，终端设备根据该配置信息#5可以确定发送信道学习模型的配置参数的资源、码率、调制方式、比特数、反馈顺序、反馈配置参数内容。即，方法500包括：网络设备发送配置信息#5，终端设备接收配置信息#5。
可选地，网络设备可以通过高层信令发送该配置信息#5，也可以通过物理层信令发送该配置信息#5。其中，高层信令可以是指RRC信令或MAC信令，物理层信令是指DCI信令。
S550，网络设备根据第一信道信息和第二信道学习模型确定第二信道信息。
如上文所述，第二信道信息的数据量与测得的下行信道信息的下行信道信息的数据量相同或相近，也就是说，网络设备根据第二信道学习模型对第一信道信息进行解码或解压缩得到第二信道信息。
例如，若第二信道学习模型是神经网络模型，网络设备可以将第一信道信息中的低维信道信息的实部信息进行解码或者解压缩，得到高维实部信息，将第一信道信息中的低维信道信息的虚部信息进行解码或解压缩，得到高维虚部信息。第二信道信息可以包括高维实部信息和高维虚部信息。即第二信道信息可以是高维复数信道信息。
例如，网络设备将2维实部信息进行解码或解压缩后，可以得到64*1维的高维实部信息，将2维虚部信息进行解码或解压缩后，可以得到64*1维的高维虚部信息，则网络设备可以将64*1维的实部信息和64*1维的虚部信息作为第二信道信息。
又例如，若第二信道学习模型是神经网络模型，网络设备可以将低维信道信息的实部和虚部联合进行解码或解压缩，得到高维实数信息。即第二信道信息可以是包括高维实数信息。
例如，网络设备将4维实数的低维信道信息进行解码或解压缩后，可以得到64*2维 实数信息。网络设备可以将64*2维实数信息作为第二信道信息。
进一步地，网络设备确定第二信道信息之后，可以根据第二信道信息进行预编码，并根据确定的预编码进行数据传输，可以提高预编码的准确性，降低小区间和/或多个终端设备间的信号干扰，进而提高通信性能。
进一步地，网络设备确定第二信道信息之后，可以根据第二信道信息确定数据的调度信息，并根据确定的调度信息进行数据传输，可以提高数据传输的准确性，进而提高通信性能。
如下，本申请实施例提供了一种可以由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S510’至S540’中的至少一项：
S510’(即S510)，第一通信装置确定信道学习模型。具体的该步骤描述可以参考上文S510中的描述，为了简洁，此处不再详述。
S520’(即S520)，第一通信装置接收参考信号。具体的该步骤描述可以参考上文S520中的描述，为了简洁，此处不再详述。其中，S510’和S520’的顺序也可以是先执行S520’，后执行S510’，本申请对此不做限定。
S530’(即S530)，第一通信装置基于目标信道信息和第一信道学习模型确定第一信道信息。具体的该步骤描述可以参考上文S530中的描述，为了简洁，此处不再详述。
S540’(即S540)，第一通信装置发送第一信道信息。具体的该步骤描述可以参考上文S540中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以包括S510”至S540”中的至少一项：
S510”(即S510)，第二通信装置确定信道学习模型。具体的该步骤描述可以参考上文S510中的描述，为了简洁，此处不再详述。
S520”(即S520)，第二通信装置发送参考信号。具体的该步骤描述可以参考上文S520中的描述，为了简洁，此处不再详述。其中，S510”和S520”的顺序也可以是先执行S520’，后执行S510’，本申请对此不做限定。
S530’(即S540)，第二通信装置接收第一信道信息。具体的该步骤描述可以参考上文S540中的描述，为了简洁，此处不再详述。
S540’(即S550)，第二通信装置根据第一信道信息和第二信道学习模型确定第二信道信息。具体的该步骤描述可以参考上文S550中的描述，为了简洁，此处不再详述。
本方法中，终端设备根据第一信道学习模型确定第一信道信息，并向网络设备发送第一信道信息，可以降低信道信息反馈的开销，提高通信性能。另外，网络设备根据收到的第一信道信息和第二信道学习模型可以确定第二信道信息，进一步，根据第二信道信息确定数据传输的预编码和调度信息，可以提高信道获取的准确性，即提高预编码的准确性，有利于降低小区或者多个终端间的干扰，提高数据传输的准确性，提高通信性能。
如下本申请实施例提供了一种通信的方法。本实施例中提供了一种根据终端设备的能力确定信道学习模型的训练和/或反馈和/或更新方法。如下方法可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。
因为无线信道是随时间变化的，因此网络设备和/或终端设备确定信道学习模型后，随着时间的变化，信道学习模型可能不适用于当前的信道状况，此时网络设备和/或终端 设备需要对信道学习模型进行训练和/或反馈和/或更新等。
例如，终端设备的能力可以是指如下终端类型中的至少一种，：主动学习类终端、网络设备指示类终端、被动接收类终端等。
其中，主动学习类终端是指可以主动进行信道学习模型训练，并向网络设备发送信道学习反馈信令的终端设备。网络设备指示类终端是指可以根据网络设备发送的信道学习训练指示，进行信道学习模型训练并向网络设备发送信道学习反馈信令的终端设备。被动接收类终端是指没有能力进行信道学习模型训练，可通过接收网络设备发送的信道学习模型的配置参数的信令确定信道学习模型的终端设备。
其中，信道学习反馈信令中可以包括信道学习模型训练的结果，例如，第一信道学习模型是否适用，和/或，用于更新第一信道学习模型的配置参数，和/或，用于更新第二信道学习模型的配置参数。具体的可以参考本申请中S210中关于第一消息的描述，为了简洁，此处不再详述。
例如，终端设备的能力可以是指如下类型的终端中的至少一种：终端设备支持信道学习模型训练、终端设备不支持信道学习模型训练、终端设备使能信道学习模型训练、终端设备不使能信道学习模型训练等。
进一步，终端设备可以向网络设备发送关于终端设备能力的信息，即终端设备可以向网络设备上报终端设备的能力。例如，终端设备可以上报是否支持信道学习模型训练，或者上报终端类型。
该方法中，根据终端设备的能力确定信道学习模型的方法可以更适应在实际通信系统中有多种终端设备的情况，根据不同的终端设备的能力采用不同的方法确定和/或训练和/或反馈和/或更新信道学习模型，可以提高通信系统的灵活性，更兼顾了不同能力的终端设备，提高通信性能。
下面针对每一种类型的终端设备如何进行信道学习模型的训练和/或反馈和/或更新进行说明。例如，针对主动学习类终端可以按照图16的方法中的至少一个步骤与网络设备进行通信，针对网络设备指示类终端可以按照图15的方法中的至少一个步骤与网络设备进行通信，针对被动接收类终端可以按照图17的方法中的至少一个步骤与网络设备进行通信。例如，针对不支持(或不使能)信道学习模型训练的终端设备可以按照图15或图16的方法中的至少一个步骤与网络设备进行通信，针对支持(或使能)信道学习模型训练的终端设备可以按照图17的方法中的至少一个步骤与网络设备进行通信。
可选的，在图15至图17所述的方法中，第一通信装置和/或第二通信装置可以先确定终端设备的能力。具体的确定方法可以按照上述描述。即可以包括第一通信装置发送终端设备能力的信息，以及第二通信装置接收终端设备能力的信息的步骤。
图15示出了本申请实施例提供的通信的方法的示意性流程图。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。在图15示出的方法中，以第一通信装置是终端设备、第二通信装置是网络设备为例进行说明。如图15所示，方法600可以包括S610至S660中的至少一项，下面详细说明各个步骤。
S610，网络设备发送信道学习训练信令(第一指示信息的一例)。相应地，在S610中，终端设备接收信道学习训练信令。
信道学习训练信令用于指示终端设备进行信道学习模型训练，即用于指示终端设备确定第一信道学习模型是否适用，和/或，指示终端设备确定信道学习模型。
可选地，在由网络设备确定第一信道学习模型和第二信道学习模型的情况下，信道学习训练信令中还可以包括第一信道学习模型的配置参数。第一信道学习模型用于根据测得的下行信道信息确定第一信道信息，第一信道信息的数据量小于测得的下行信道信息的数据量。
可选地，信道学习训练信令中还可以包括第二信道学习模型的配置参数。第二信道学习模型用于根据第一信道信息确定第二信道信息。
可选地，信道学习训练信令中还可以包括信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息、信道学习训练的反馈内容相关的信息中的一项或多项。
可选地，第一信道学习模型的配置参数、第二信道学习模型的配置参数、信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息、信道学习模型训练的反馈内容相关的信息中的一项或多项，可以通过不同的信令发送给终端设备。
可选的，信道学习训练信令可以通过高层信令发送，例如MAC层信令，RRC层信令，或者，可以通过物理层信令发送，例如DCI信令。
更多的关于信道学习训练信令的描述可以参考上文S230中关于第一指示信息的描述，为了简洁，此处不再详述。
S620，终端设备根据信道学习训练信令进行信道学习模型训练。
终端设备根据接收到的信道学习训练信令可以进行信道学习模型训练，并确定第一信道学习模型是否适用，或者是否需要更新。终端设备确定第一信道学习模型是否适用的方法可以参考上文S210中的描述。
可选地，在终端设备确定第一信道学习模型不适用的情况下，终端设备还可以确定新的第一信道学习模型和新的第二信道学习模型。终端设备确定第一信道学习模型和第二信道学习模型的方法可以参考上文S320中的描述。
S630，终端设备发送信道学习反馈信令(第一消息的一例)。相应地，在S630中，网络设备接收信道学习反馈信令。
信道学习反馈信令中可以包括信道学习模型训练的结果，例如，第一信道学习模型是否适用，和/或，用于更新第一信道学习模型的配置参数，和/或，用于更新第二信道学习模型的配置参数。
可选的，终端设备在向网络设备发送信道学习反馈信令之前，终端设备可以向网络设备发送信道学习反馈信令的发送请求。该发送请求用于指示终端设备向终端设备请求发送信道学习反馈信令。该发送请求可以类似现有技术中的调度请求(scheduling request，SR)。当网络设备接收该发送请求后，网络设备可以发送调度信息，该调度信息指示终端设备发送信道学习反馈信令。该调度信息中可以包括信道学习反馈信令的相关配置信息，例如时频资源，反馈内容，反馈码率，反馈比特数等中的一项或多项。
可选地，若终端设备和网络设备预先配置了几组第二信道学习模型，以及预先配置了与第二信道学习模型对应的模型标识，则终端设备在反馈用于更新第二信道学习模型的配置参数的情况下，可以仅反馈确定的第二信道学习模型的标识。例如，终端设备和网络设备预先配置了4组信道学习模型，4组信道学习模型分别对应模型标识1、模型标识2、 模型标识3和模型标识4。若终端设备确定的第二信道学习模型是第3组信道学习模型，则终端设备将模型标识3携带在信道学习反馈信令中，发送给网络设备。相应地，网络设备根据模型标识3可以确定终端设备适用的第二信道学习模型。
可选地，若终端设备和网络设备预先配置了几组第一信道学习模型，以及预先配置了与第一信道学习模型对应的模型标识，则终端设备在反馈用于更新第一信道学习模型的配置参数的情况下，可以仅反馈确定的第一信道学习模型的标识。例如，终端设备和网络设备预先配置了4组信道学习模型，4组信道学习模型分别对应模型标识1、模型标识2、模型标识3和模型标识4。若终端设备确定的第一信道学习模型是第3组信道学习模型，则终端设备将模型标识3携带在信道学习反馈信令中，发送给网络设备。相应地，网络设备根据模型标识3可以确定终端设备适用的第一信道学习模型。
可选地，若终端设备和网络设备预先配置了几组信道学习模型(包括第一信道学习模型和第二信道学习模型)，以及预先配置了与信道学习模型对应的模型标识，则终端设备在反馈用于更新第一信道学习模型和第二信道学习模型的配置参数的情况下，可以仅反馈确定的信道学习模型的标识。例如，终端设备和网络设备预先配置了4组信道学习模型，4组信道学习模型分别对应模型标识1、模型标识2、模型标识3和模型标识4。若终端设备确定的信道学习模型是第3组信道学习模型，则终端设备将模型标识3携带在信道学习反馈信令中，发送给网络设备。相应地，网络设备根据模型标识3可以确定终端设备适用的信道学习模型。
本方法中，在信道学习反馈信令中，通过反馈信道学习模型标识的方式确定信道学习学习模型，可以降低反馈比特数，降低反馈开销，提高通信性能。
具体地，关于信道学习反馈信令的反馈内容、反馈形式、反馈资源等内容，可以参考上文S210中关于第一消息的描述，为了简洁，此处不再详述。
可选地，在S630之前，方法600还可以包括：终端设备接收网络设备发送的配置信息#6，该配置信息#6用于指示发送信道学习反馈信令的资源、码率、调制方式、比特数、反馈顺序、反馈内容等中的一项或多项。相应地，终端设备根据该配置信息#3可以确定发送信道学习反馈信令的资源、码率、调制方式、比特数、反馈顺序或反馈的内容。
可选地，网络设备可以通过高层信令发送该配置信息#6，也可以通过物理层信令发送该配置信息#6。其中，高层信令可以是指RRC信令或MAC信令，物理层信令是指DCI信令。
可选的，信道学习反馈信令也可以与第一信道信息一起发送，也可以与第一信道信息分开发送。例如终端设备可以通过一条信令发送信道学习反馈信令和第一信道信息，或者终端设备可以通过多条信令发送信道学习反馈信令和第一信道信息。具体的发送方式可以是预定义的，也可以是通过网络设备指示的，具体的，在此不做限定。
可选的，当终端设备确定信道学习模型不适用，并且确定了信道学习模型的配置参数之后，终端设备可以将信道学习模型的配置参数发送和/或存储到数据库或者其他网元。具体的数据库描述可以参考上文S220中关于数据库的描述，为了简洁，此处不再详述。
S640，网络设备确定信道学习模型。
网络设备根据接收的信道学习反馈信令，可以确定当前的信道学习模型是否适用，和/或，确定信道学习模型的配置参数。
可选的，网络设备可以参考上文S410中关于第二通信装置确定信道学习模型是否适用的描述确定信道学习模型是否适用，为了简洁，此处不再详述。
可选的，网络设备可以参考上文S420中关于第二通信装置确定信道学习模型的描述确定信道学习模型，为了简洁，此处不再详述。
例如，若信道学习反馈信令用于指示当前的信道学习模型不适用，且信道学习反馈信令中包括用于更新第一信道学习模型和/或第二信道学习模型的配置参数，则网络设备根据该信道学习反馈信令可以确定当前的信道学习模型不适用，并可以对当前的第一信道学习模型和/或第二信道学习模型进行更新，以确定新的第二信道学习模型。
又例如，若信道学习反馈信令用于指示当前的信道学习模型不适用，则网络设备根据该信道学习反馈信令可以确定当前的信道学习模型不适用。进一步地，方法600还可以包括S650和S660。
可选的，S650和S660也可以作为独立的实施例，也可以与其他实施例相结合，具体的，本申请对此不做限定。本实施例可以提供了一种确定信道学习模型的方法。
S650，网络设备发送第一信道学习模型的配置信息。相应地，在S650中，终端设备接收第一信道学习模型的配置信息。
网络设备确定当前信道学习模型不适用之后，可以确定新的第一信道学习模型和新的第二信道学习模型，并将新的第一信道学习模型的配置参数发送给终端设备，以更新当前的第一信道学习模型。即网络设备可以向终端设备发送用于更新第一信道学习模型的配置参数。
可选地，网络设备可以将新的第一信道学习模型的配置参数发送/或存储到数据库#2中，关于数据库#2的描述可以参考上文S220中的描述。
可选的，本申请实施例中，当终端设备接入该网络设备所在的小区，或终端设备与网络设备建立RRC连接时，或终端设备进行小区切换时，网络设备可以向终端设备发送信道学习模型的配置信息。
S660，终端设备确定第一信道学习模型。
终端设备根据接收到的用于更新第一信道学习模型的配置参数，更新当前的的第一信道学习模型，以确定新的第一信道学习模型。
可选地，终端设备可以将新的第一信道学习模型的配置参数发送/或存储到数据库#1中，关于数据库#1的描述可以参考上文S220中的描述。
可选的，本申请实施例中，当终端设备接入该网络设备所在的小区，或终端设备与网络设备建立RRC连接时，或终端设备进行小区切换时，终端设备可以接收网络设备发送的信道学习模型的配置信息，终端设备可以根据所述信道学习模型的配置参数确定信道学习模型。
如下，本申请实施例提供了一种可以由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S610’至S640’中的至少一项：
S610’(即S610)，第一通信装置接收信道学习训练信令。具体的该步骤描述可以参考上文S610中的描述，为了简洁，此处不再详述。
S620’(即S610)，第一通信装置根据信道学习训练信令进行信道学习模型训练。具体的该步骤描述可以参考上文S620中的描述，为了简洁，此处不再详述。
S630’(即S630)，第一通信装置发送信道学习反馈信令。具体的该步骤描述可以参考上文S630中的描述，为了简洁，此处不再详述。
S640’(即S650)，第一通信装置接收信道学习模型的配置信息。具体的该步骤描述可以参考上文S650中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以包括S610”至S640”中的至少一项：
S610”(即S610)，第二通信装置发送信道学习训练信令。具体的该步骤描述可以参考上文S610中的描述，为了简洁，此处不再详述。
S620”(即S630)，第二通信装置接收信道学习反馈信令。具体的该步骤描述可以参考上文S630中的描述，为了简洁，此处不再详述。
S630”(即S640)，第二通信装置确定信道学习模型。具体的该步骤描述可以参考上文S640中的描述，为了简洁，此处不再详述述。
S640”(即S650)，第二通信装置发送信道学习模型的配置信息。具体的该步骤描述可以参考上文S650中的描述，为了简洁，此处不再详述。
本方法中，终端设备根据网络设备的信道学习训练信令进行信道学习模型训练，可以是适用于网络设备确定信道学习模型不适用的场景。在网络设备确定信道学习模型不适用的情况下，及时告知终端设备进行信道学习模型训练，并确定适用的信道学习模型，从而可以提高信道学习模型的准确性，对信道学习模型及时更新，可以提高通信性能。
图16示出了本申请实施例提供的通信的方法的示意性流程图。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。在图16示出的方法中，以第一通信装置是终端设备、第二通信装置是网络设备为例进行说明。如图16所示，方法700可以包括S710至S770中的至少一项，下面详细说明各个步骤。
S710，终端设备向网络设备发送信道学习训练请求信令(第一请求消息的一例)。相应地，在S710中，网络设备接收终端设备发送的信道学习训练请求信令。
信道学习训练请求信令用于指示终端设备请求进行信道学习模型训练，和/或，用于指示终端设备请求反馈信道学习训练信令。
可选的，信道学习训练请求信令可以在SR资源中发送。
S720，网络设备发送信道学习训练信令(第一指示信息的一例)。相应地，在S720中，终端设备接收信道学习训练信令。
S730，终端设备根据信道学习训练信令进行信道学习模型训练。
S740，终端设备发送信道学习反馈信令(第一消息的一例)。相应地，在S740中，网络设备接收信道学习反馈信令。
S750，网络设备确定信道学习模型。
S760，网络设备发送第一信道学习模型的配置信息。相应地，在S760中，终端设备接收第一信道学习模型的配置信息。
S770，终端设备确定第一信道学习模型。
应理解，图16仅以S710在S730之前为例，不应对本申请实施例造成限定。可选地，S710可以在S730和S740之间执行。例如S710可以在S730之后执行。例如S710可以在 S740之前执行。
S720至S770可以与方法600中的S610至S660相同，为了简洁，本申请实施例不再详述。
可选地，在终端设备与网络设备预先定义的发送信道学习反馈信令的相关配置信息的情况下，方法700可以不执行S710和S720。其中，与发送信道学习反馈信令相关的配置信息可以包括：信道学习反馈信息令占用的时频资源、信道学习反馈信令的内容、信道学习反馈信令的反馈码率、信道学习反馈信令的反馈比特数。
如下，本申请实施例提供了一种可以由第一通信装置的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以包括S710’至S750’中的至少一项：
S710’(即S710)，第一通信装置发送信道学习训练请求信令。具体的该步骤描述可以上文S710中的描述，为了简洁，此处不再详述。
S720’(即S720)，第一通信装置接收信道学习训练信令。具体的该步骤描述可以参考上文S610中的描述，为了简洁，此处不再详述。
S730’(即S730)，第一通信装置根据信道学习训练信令进行信道学习模型训练。具体的该步骤描述可以参考上文S620中的描述，为了简洁，此处不再详述。
S740’(即S740)，第一通信装置发送信道学习反馈信令。具体的该步骤描述可以参考上文S630的描述，为了简洁，此处不再详述。
S750’(即S760)，第一通信装置接收信道学习模型的配置信息。具体的该步骤描述可以上文S650中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置的通信的方法。即如下方法是从第二通信装置的角度进行描述，该方法可以可以包括S710”至S750”中的至少一项：
S710”(即S710)，第二通信装置接收信道学习训练请求信令。具体的该步骤描述可以参考上文S710中描述，为了简洁，此处不再详述。
S720”(即S720)，第二通信装置发送信道学习训练信令。具体的该步骤描述可以参考上文S610中的描述，为了简洁，此处不再详述。
S730”(即S740)，第二通信装置接收信道学习反馈信令。具体的该步骤描述可以参考上文S630中的描述，为了简洁，此处不再详述。
S740”(即S750)，第二通信装置确定信道学习模型。具体的该步骤描述可以参考上文S640中的描述，为了简洁，此处不再详述。
S750”(即S760)，第二通信装置发送信道学习模型的配置信息。具体的该步骤描述可以参考上文S650中的描述，为了简洁，此处不再详述。
本方法中，终端设备可以向网络设备发送信道学习训练请求信令以请求进行信道学习模型训练，可以是适用于终端设备确定信道学习模型不适用的场景。在终端设备确定信道学习模型不适用的情况下，可以及时告知网络设备以请求进行信道学习模型训练的指示，从而确定适用的信道学习模型，提高信道学习模型的准确性，对信道学习模型及时更新，可以提高通信性能。
图17示出了本申请实施例提供的通信的方法的示意性流程图。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。在图17示出的方法中，以第一通信装置是终端设备、第二通信装置是网络设备为例进行 说明。如图17所示，方法800可以包括S810至S830中的至少一项，下面详细说明各个步骤。
S810，网络设备确定信道学习模型。
可选的，网络设备可以参考上文S410中关于第二通信装置确定信道学习模型的描述确定信道学习模型，为了简洁，此处不再详述。
S820，网络设备发送信道学习模型的配置信息。相应的，在S820中，终端设备接收信道学习模型的配置信息。
其中，本申请实施例中，信道学习模型的配置信息可以是包括信道学习模型的配置参数的信息。
可选的，网络设备发送信道学习模型的配置信息可以参考上文S420中关于第二通信装置发送第二消息，第二消息用于指示用于更新第一信道学习模型的一个或多个配置参数的描述，为了简洁，此处不再详述。
S830，终端设备根据信道学习模型的配置信息确定信道学习模型。
如下，本申请实施例提供了一种可以由第一通信装置执行的通信的方法。即如下方法是从第一通信装置的角度进行描述，该方法可以可以包括S810’至S820’中的至少一项：
S810’(即S820)，第一通信装置接收信道学习模型的配置信息。具体的该步骤描述可以参考上文S650中的描述，为了简洁，此处不再详述。
S820’(即S830)，第一通信装置根据信道学习模型的配置信息确定信道学习模型。具体的该步骤描述可以参考上文S660中的描述，为了简洁，此处不再详述。
如下，本申请实施例提供了一种可以由第二通信装置执行的通信的方法。即如下方法是从第二通信装置角度进行描述，该方法可以包括S810”至S820”中的至少一项：
S810”(即S810)，第二通信装置确定信道学习模型。具体的该步骤描述可以参考上文S410中的描述，为了简洁，此处不再详述。
S820”(即S820)，第二通信装置发送信道学习模型的配置信息。具体的该步骤描述可以参考上文S420中的描述，为了简洁，此处不再详述。
可选的，本申请实施例中，当终端设备接入该网络设备所在的小区，或终端设备与网络设备建立RRC连接时，或终端设备进行小区切换时，终端设备可以接收网络设备发送的信道学习模型的配置信息，终端设备可以根据信道学习模型的配置参数确定信道学习模型。
本方法中，终端设备根据网络设备发送信道学习模型的配置信息确定信道学习模型，可以是适用于终端设备没有信道学习模型训练能力的场景，也可以是适用于为例降低终端设备的处理复杂度的场景，即终端设备无需进行信道学习模型训练也可以确定适用的信道学习模型，可以提高信道学习模型的准确性，对信道学习模型及时更新，可以提高通信性能。
如下实施例提供了一种通信的方法。该实施例提供了一种根据有无信道学习训练信令确定信道学习反馈信令的方法。本申请实施例可以作为独立的实施例，也可以与本申请中的其他实施例相结合，具体的，本申请对此不做限定。如下一种或多种方法可以单独使用，也可以联合使用，具体的，本申请对此不做限定。
第一通信装置可以根据是否接收到信道学习训练信令确定如下信息中的至少一项：
a.信道学习反馈信令是否包括信道学习模型是否适用的信息；
b.信道学习反馈信令的反馈值的含义；
c.信道学习反馈信令的反馈内容、反馈形式、反馈资源中至少一项
第二通信装置可以根据是否发送信道学习训练信令确定如下信息中的至少一项：
a.信道学习反馈信令是否包括信道学习模型是否适用的信息；
b.信道学习反馈信令的反馈值的含义；
c.信道学习反馈信令的反馈内容、反馈形式、反馈资源中至少一项
下文以第一通信装置是终端设备、第二通信装置是网络设备为例进行说明。
作为一种可能的实现方式，信道学习反馈信令中是否包括信道学习模型是否适用的信息可以是根据网络设备是否发送信道学习训练信令确定。即网络设备根据是否发送信道学习训练信令确定信道学习反馈信令中是否包括信道学习模型是否适用的信息。
作为一种可能的实现方式，信道学习反馈信令中是否包括信道学习模型是否适用的信息可以是根据终端设备是否接收到信道学习训练信令确定。即终端设备根据是否发送信道学习训练信令确定信道学习反馈信令中是否包括信道学习模型是否适用的信息。
场景1：网络设备向终端设备发送信道学习训练信令。
当终端设备接收到网络设备发送的信道学习训练信令，终端设备可以基于目标信道信息和信道学习模型向网络设备发送信道学习反馈信令。
该信道学习反馈信令中可以包括信道学习模型是否适用的信息，或者，该信道学习反馈信令中不包括信道学习模型是否适用的信息。
例如：网络设备指示终端设备进行信道学习模型训练以确定信道学习模型是否适用，因此网络设备发送信道学习训练信令。在该场景下，网络设备不确定信道学习模型是否适用，因此终端设备发送的信道学习反馈信令中可以包括信道学习模型是否适用的信息。
又例如：网络设备确定信道学习模型不适用，因此网络设备发送信道学习训练信令是为了确定更合适的信道学习模型。此时，终端设备发送的信道学习反馈信令中可以不包括信道学习模型是否适用的信息。
可选的，信道学习反馈信令中是否包括信道学习模型是否适用的信息可以是协议预定义的，也可以是网络设备通过信令告知终端设备的，具体的，本申请对此不做限定。
场景2：网络设备不向终端设备发送信道学习训练信令。
可选的，当网络设备向终端设备发送信道学习训练信令时，网络设备确定信道学习反馈中包括信道学习模型是否适用的信息。当网络设备不向终端设备发送信道学习训练信令时，网络设备确定信道学习反馈中不包括信道学习模型是否适用的信息。
可选的，当终端设备接收到信道学习训练信令时，终端设备确定信道学习反馈中包括信道学习模型是否适用的信息。当终端设备没有接收到信道学习训练信令时，终端设备确定信道学习反馈中不包括信道学习模型是否适用的信息。
可选的，当信道学习反馈信令中包括信道学习模型是否适用的信息时，网络设备可以通过如下至少一种方式确定信道学习模型是否适用：
作为一种可能的实现方式，信道学习反馈信令中的信息指示信道学习模型是否适用。
例如1比特信息指示信道学习模型是否适用。
信道学习模型是否适用可以通过在信道学习反馈信令中反馈1比特信息，该1比特信 息用于指示信道学习模型是否适用。例如1比特信息是“0”，则表示信道学习模型不适用，1比特信息是“1”，则表示信道学习模型适用。反之亦可。
作为一种可能的实现方式，信道学习反馈信令中的反馈值指示信道学习模型是否适用。
信道学习反馈信令中的反馈值可以是如下的至少一个：rank值、CQI值，CRI值。
(1)rank值
现有技术中，终端设备反馈的rank值是可以从1～R，R为正整数。例如R为8。rank的域的比特数可以根据终端设备最大支持的层数，以及天线端口数确定，例如，rank的域的比特数为log 2(min(层数，天线端口数))向上取整，例如为log 2(R)向上取整。例如终端设备最大支持4层，则rank取值可以为1～4，则可以用2个比特指示rank值。例如天线端口数为8，终端设备最大支持4层，则rank取值可以为1～4，则可以用2个比特指示rank值。
可选的，当网络设备没有向终端设备发送信道学习训练信令时，或者，当终端设备没有接收到信道学习反馈信令时，网络设备和/或终端设备可以确定rank的域的比特数为log 2(min(层数，天线端口数))向上取整。
当网络设备向终端设备发送信道学习训练信令时，或者，当终端设备接收到网络设备发送的信道学习训练信令时，终端设备和网络设备可以确定rank值可以为0～R，R为正整数。例如R为8。此时，rank的域的比特数为log 2(min(层数，天线端口数)+1)向上取整，例如为log 2(R+1)向上取整。其中，rank值为0，表示信道学习模型不适用。
可选的，当网络设备向终端设备发送信道学习训练信令时，或者，当终端设备接收到信道学习反馈信令时，网络设备和/或终端设备可以确定rank的域的比特数为log 2(min(层数，天线端口数)+1)向上取整。
例如，当网络设备向终端设备发送信道学习训练信令时，网络设备可以根据rank值为0确定第一信道学习模型不适用。或者，rank值为0指示信道学习模型不适用。
例如，当终端设备接收到网络设备发送的信道学习训练信令时，当终端设备确定第一信道学习模型不适用时，终端设备可以发送rank值为0的信息，rank值为0指示信道学习模型不适用。
(2)CQI值
可选的，CQI值为0时(即CQI index＝0)，可以表示信道学习模型不适用。
例如，当网络设备向终端设备发送信道学习训练信令时，网络设备可以根据CQI值为0确定第一信道学习模型不适用。或者，CQI值为0指示信道学习模型不适用。
例如，当终端设备接收到网络设备发送的信道学习训练信令时，当终端设备确定第一信道学习模型不适用时，终端设备可以发送CQI值为0的信息，CQI值为0指示信道学习模型不适用。
(3)CRI值
现有技术中，终端设备反馈的CRI值是对应测量的CSI-RS资源，例如如果CSI-RS资源为C个，C为正整数。CRI值可以为1～C。CRI的域的比特数可以根据配置的CSI-RS资源个数数确定，例如CRI的域的比特数为log 2(C)向上取整。例如终端设备配置测量的CSI-RS资源数为2，则CRI仅需要1个比特指示。例如比特0代表配置的第一个CSI-RS 资源，比特2代表配置的第二个CSI-RS资源。
可选的，当网络设备没有向终端设备发送信道学习训练信令时，或者，当终端设备没有接收到信道学习反馈信令时，网络设备和/或终端设备可以确定CRI的域的比特数为log 2(C)向上取整，其中，C为CSI-RS资源的个数。
当网络指示终端设备信道学习训练信令时，CRI值可以为0～C，C为正整数。例如C为2。此时，CRI的域的比特数为log 2(C+1)向上取整。其中，CRI值为0，表示信道学习模型不适用。
可选的，当网络设备向终端设备发送信道学习训练信令时，或者，当终端设备接收到信道学习反馈信令时，网络设备和/或终端设备可以确定CRI的域的比特数为log 2(C+1)向上取整，其中，C为CSI-RS资源的个数。
作为一种可能的实现方式，信道学习反馈信令中的内容可以是根据网络设备发送的信道学习训练信令中指示的反馈内容确定。具体的，可以参考本申请中的其他实施例的描述，在此不再赘述。
可选，信道学习反馈信令中可以包括当前信道学习模型的性能指标，例如确定模型是否适用的相关性能情况。网络设备根据接收的该信道学习反馈信令可以确定信道学习模型是否适用。具体的，可以参考本申请中的其他实施例的描述，在此不再赘述。
可选地，信道学习反馈信令中的反馈形式可以是根据网络设备是否发送的信道学习训练信令确定。即网络设备根据是否发送信道学习训练信令确定信道学习反馈信令中的反馈形式。
可选地，信道学习反馈信令中的反馈形式可以是根据终端设备是否接收到信道学习训练信令确定。即终端设备根据是否接收信道学习训练信令确定信道学习反馈信令中的反馈形式。
信道学习反馈信令的反馈形式可以是包括如下至少一种：反馈的周期性/非周期性/半持续性、差分反馈/绝对值反馈/相对值反馈、反馈资源。具体的，可以参考本申请中的其他实施例的描述，在此不再赘述。
例如，当网络设备向终端设备发送信道学习训练信令时，网络设备确定信道学习反馈信令可以是非周期性发送。当网络设备没有向终端设备发送信道学习训练信令时，网络设备确定信道学习反馈信令可以是周期性发送。或者，反之亦可。
例如，当终端设备接收到信道学习训练信令时，终端设备可以确定信道学习反馈信令是非周期性发送。当终端设备没有接收到信道学习训练信令时，终端设备可以确定信道学习反馈信令是周期性发送。或者，反之亦可。
例如，当终端接收到信道学习训练信令时，终端设备确定信道学习反馈信令采用绝对值反馈的方式反馈；当终端设备没有接收到信道学习训练信令时，终端设备确定信道学习反馈信令采用差分反馈或者相对值反馈的方式反馈等。或者，反之亦可。
例如，当网络设备向终端设备发送信道学习训练信令时，网络设备确定信道学习反馈信令采用绝对值反馈的方式反馈；当网络设备没有向终端设备发送信道学习训练信令时，网络设备确定信道学习反馈信令采用差分反馈或者相对值反馈的方式反馈等。或者，反之亦可。
例如，当终端设备接收到信道学习训练信令时，信道学习反馈信令在PUSCH中传输； 当终端设备没有接收到信道学习训练信令时，信道学习反馈信令在PUCCH中传输等。或者，反之亦可。
例如，当网络设备向终端设备发送信道学习训练信令时，信道学习反馈信令在PUSCH中传输；当网络设备没有向终端设备发送信道学习训练信令时，信道学习反馈信令在PUCCH中传输等。或者，反之亦可。
本实施例，通过针对是否发送或接收信道学习训练信令设计不同的信道学习反馈信令，可以实现不同场景下的反馈信息不同，合理高效的进行信道学习反馈信令的发送和/或接收，可以降低反馈开销，提高通信性能。
如下实施例提供了一种根据信道学习模型是否适用设计信道学习反馈信令的方法。本申请实施例可以作为独立的实施例，也可以与本申请其他实施例相结合，具体的，本申请对此不做限定。如下一种或多种方法可以单独使用，也可以联合使用，具体的，本申请对此不做限定。
作为一种可能的实现方式，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用确定信道学习反馈信令的反馈内容。
下文以第一通信装置为终端设备，第二通信装置为网络设备为例进行说明。
例如，信道学习反馈信令中可以包括第一信道信息。
如果信道学习模型适用，则终端设备可以基于该信道学习模型确定第一信道信息，并反馈第一信道信息。此方式下，信道学习反馈信令中包括第一信道信息。
或者，
如果信道学习模型适用，终端设备可以发送信道学习模型适用的信息以及第一信道信息)。此方式下，信道学习反馈信令中包括信道学习模型适用的信息以及第一信道信息。
如果信道学习模型不适用，则终端设备可以发送当前信道学习模型不适用的信息给网络设备。进一步地，网络设备接收到该信息后，可以向终端设备发送信道学习模型的配置参数。进一步对，终端设备基于接收到的信道学习模型的配置参数确定信道学习模型，并基于信道学习模型确定第一信道信息，并反馈第一信道信息。此方式下，信道学习反馈信令中仅包括信道学习模型不适用的信息。
或者，
如果信道学习模型不适用，则终端设备可以发送当前信道学习模型不适用的信息给网络设备，并发送基于当前信道学习模型得到的第一信道信息。此方式下，信道学习反馈信令中包括信道学习模型不适用的信息以及基于当前信道学习模型确定的第一信道信息。
或者，
如果信道学习模型不适用，则终端设备可以基于训练得到的更新的信道学习模型得到第一信道信息，并向网络设备发送训练得到的更新的信道学习模型的配置参数和第一信道信息。此方式下，信道学习反馈信令中包括更新信道学习模型的配置参数以及基于更新的信道学习模型确定的第一信道信息。
或者，
如果信道学习模型不适用，则终端设备可以基于训练得到的更新的信道学习模型得到第一信道信息，并向网络设备发送训练得到的更新的信道学习模型的配置参数和第一信道信息。此方式下，信道学习反馈信令中包括信道学习模型不适用的信息，更新信道学习模 型的配置参数以及基于更新的信道学习模型确定的第一信道信息。
可选的，终端设备具体采用上述那种方式进行反馈可以是协议预定义的，也可以是网络设备通过信令告知终端设备的，具体的，本申请对此不做限定。
作为一种可能的实现方式，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用确定信道学习反馈信令的反馈形式。
具体的，当信道学习模型适用时，第一信道信息的反馈可以以差分的形式上报，例如终端设备当前反馈的第一信道信息的反馈量的取值可以是通过反馈与上次反馈的第一信道信息反馈量的取值的差值的方式反馈。例如，可以反馈rank的差值，CQI的差值，压缩后反馈量的差值等。当信道学习模型相同时，第一信道信息的取值的范围可以是相对稳定的，因此可以通过反馈差值的方式反馈第一信道信息，并且可以降低反馈开销。
例如，当信道学习模型适用时，信道学习反馈信令可以在PUCCH中反馈。
例如，当信道学习模型不适用时，信道学习反馈信令可以在PUSCH中反馈。
例如，当信道学习模型不适用时，第一信道信息的反馈可以不以差分的形式上报。因为信道学习模型变化之后，第一信道信息的取值反馈可能会发生变化，因此不适于通过反馈与之前反馈的第一信道信息的差值的方式进行反馈。
信道学习反馈信令中可以包括信道状态信息的内容，例如Rank值，CQI等。
当信道学习模型适用时终端可以反馈rank值和/或CQI，第一信道信息。而当信道学习模型不适用时，终端可以仅反馈rank值和/或CQI。
作为一种可能的实现方式，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用确定信道学习反馈信令的反馈内容。
例如，当信道学习模型适用时，信道学习反馈信令中包括rank值、CQI和第一信道信息。或者，当信道学习模型适用时，信道学习反馈信令中包括rank值和第一信道信息。或者，当信道学习模型适用时，信道学习反馈信令中包括CQI值和第一信道信息。
例如，当信道学习模型不适用时，信道学习反馈信令中包括rank值、CQI。或者，当信道学习模型不适用时，信道学习反馈信令中包括rank值或者CQI。
例如，信道学习模型是否适用可以与信道学习反馈信令包括的内容有对应关系。具体的，可以参考本申请中的表11的对应关系的举例，为了简洁，在此不再赘述。
可选的，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用与信道学习反馈信令包括的内容的对应关系确定信道学习反馈信令的反馈内容。
作为一种可能的实现方式，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用确定信道学习反馈信令占用的资源。
因为，信道学习模型是否适用可以对应不同的信道学习反馈信令的反馈内容。因此网络设备和/或终端设备可以根据信道学习模型是否适应确定信道学习反馈信令占用的资源。
具体的，例如网络设备和/或终端设备可以根据信道学习模型是否适用确定信道学习反馈信令占用的资源标识，和/或，信道学习反馈信令占用的资源大小。
例如当信道学习模型适用时，信道学习反馈信令占用的资源为资源1，即终端设备可以在资源1上发送信道学习反馈信令，而网络设备可以在资源1上接收信道学习反馈信令。
例如当信道学习模型不适用时，信道学习反馈信令占用的资源为资源2，即终端设备 可以在资源2上发送信道学习反馈信令，而网络设备可以在资源2上接收信道学习反馈信令。
例如当信道学习模型适用时，信道学习反馈信令占用的资源大小为X个资源单元，所述资源单元可以是指RE，符号，或者RB。其中，X为正整数，或者，X也可以是指一定范围，例如X为X 1～X 2，或者，X大于X 1，或者，X小于X 2等。其中，X 1，X 2为正整数。
例如当信道学习模型不适用时，信道学习反馈信令占用的资源大小为Y个资源单元，所述资源单元可以是指RE，符号，或者RB。其中，Y为正整数，或者，Y也可以是指一定范围，例如Y为Y 1～Y 2，或者，Y大于Y 1，或者，Y小于Y 2等。其中，Y 1，Y 2为正整数。
可选的，网络设备和/或终端设备可以根据信道学习反馈信令占用的资源大小确定信道学习反馈信令的反馈资源。即终端设备可以在所述反馈资源上发送信道学习反馈信令，而网络设备可以在相应的反馈资源上接收信道学习反馈信令。
另外，信道学习模型是否适应可以与信道学习反馈信令占用的资源有对应关系。具体的，可以参考本申请中的表8的对应关系的举例，为了简洁，在此不再赘述。
可选的，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用与信道学习反馈信令占用的资源的对应关系确定信道学习反馈信令占用的资源。
另外，网络设备和/或终端设备可以根据信道学习模型是否适用确定信道学习反馈信令的比特数。进一步地，网络设备和/或终端设备可以根据信道学习反馈信令的比特数确定信道学习反馈信令占用的资源。
例如，信道学习模型适用时，信道学习反馈信令的比特数为Z，其中，Z为正整数，或者，Z也可以是指一定范围，例如Z为Z 1～Z 2，或者，Z大于Z 1，或者，Z小于Z 2等。其中，Z 1，Z 2为正整数。
例如，信道学习模型不适用时，信道学习反馈信令的比特数为W 3，其中，W 3为正整数，或者，W 3也可以是指一定范围，例如W 3为W 3，或者，W大于W 3,1，或者，W小于W 3,2等。其中，W 3,1，W 3,2为正整数。
例如，信道学习模型适用时，信道学习反馈信令的比特数为20bits；信道学习模型不适用时，信道学习反馈信令的比特数为10bits等。
例如，信道学习模型适用时，信道学习反馈信令的比特数为大于10bits；信道学习模型不适用时，信道学习反馈信令的比特数为小于10bits等。
例如，网络设备和/或终端设备可以根据信道学习反馈信令的比特数确定信道学习反馈信令占用的资源标识，信道学习反馈信令占用的资源大小。
例如，信道学习反馈信令的比特数为Z时，信道学习反馈信令占用的资源标识为资源1。
例如，信道学习反馈信令的比特数为W 3时，信道学习反馈信令占用的资源标识为资源2。
例如，信道学习反馈信令的比特数为20bits时，信道学习反馈信令占用的资源标识为资源1；信道学习反馈信令的比特数为10bits时，信道学习反馈信令占用的资源标识为资源1。
例如，信道学习反馈信令的比特数为大于10bits时，信道学习反馈信令占用的资源标 识为资源1；信道学习反馈信令的比特数为小于10bits时，信道学习反馈信令占用的资源标识为资源2。
例如信道学习反馈信令的比特数为Z时，信道学习反馈信令占用的资源大小为X个资源单元，所资源单元可以是指RE，符号，或者RB。
例如信道学习反馈信令的比特数为W 3时，，信道学习反馈信令占用的资源大小为Y个资源单元，所资源单元可以是指RE，符号，或者RB。
可选的，信道学习模型是否适用与信道学习反馈信令的比特数有对应关系。具体的，可以参考本申请中的表9的对应关系的举例，为了简洁，在此不再赘述。
可选的，第一通信装置和/或第二通信装置可以根据信道学习模型是否适用与信道学习反馈信令的比特数确定信道学习反馈信令的比特数。
可选的，信道学习反馈信令的比特数与信道学习反馈信令占用的资源有对应关系。具体的，可以参考本申请中的表10的对应关系的举例，为了简洁，在此不再赘述。
可选的，第一通信装置和/或第二通信装置可以根据信道学习反馈信令的比特数与信道学习反馈信令占用的资源确定信道学习反馈信令占用的资源。
通过上述的设计，可以在终端设备接收到信道学习训练信令时，终端设备和/或网络设备确定信道学习模型是否适用，并根据信道学习模型是否适用确定信道学习反馈信令的反馈内容，反馈比特数，反馈资源，可以更有效的降低信道学习反馈信令的反馈开销，合理利用反馈资源，提高资源利用率，进而提升通信性能。例如，针对信道学习模型不适用的场景，可以降低信道学习反馈信令的反馈内容，进而降低反馈开销，避免不必要的反馈。
针对上位所述的场景2，即终端设备没有收到信道学习训练信令。对应的，网络设备没有发送信道学习训练信令。
在该场景下，终端设备可以根据定时器(或者周期性)进行信道学习训练并反馈信道学习反馈信令。因为是在指定的时间，指定的资源上进行训练和反馈。因此反馈的内容可以是仅包括信道学习模型的配置参数。该信道学习反馈信令中可以不包括信道学习模型是否适用的信息。例如，终端设备可以在确定信道学习模型不适用时才反馈信道学习反馈信令。
在该场景下，信道学习反馈信令可以是周期性发送。
该信道学习反馈信令中的rank值按照现有技术中的设计，取值为1～R。
该信道学习反馈信令中的反馈内容可以包括：信道学习模型的配置参数。
该信道学习反馈信令中的反馈内容可以包括：第一信道信息。因为定时器长度较长，所以第一信道信息变化较大，因此第一信道信息的反馈可以不以差分的方式反馈。第一信道信息是终端设备基于信道学习反馈信令中的配置参数确定的第一信道信息。
当终端设备自主在特定时间特定资源反馈信道学习反馈信令时，为了让网络设备可以正确接收，降低检测/接收的复杂度，该场景下，信道学习反馈信令的反馈内容，反馈比特数，反馈资源可以是预先配置的，例如根据网络设备的高层信令配置确定。
本实施例，通过针对模型是否适用设计不同的信道学习反馈信令，可以实现不同场景下的反馈信息不同，合理高效的进行信道学习反馈信令的发送和/或接收，可以降低反馈开销，提高通信性能。
上文结合图10至图17详细地描述了本申请实施例的方法，下文结合图18至图20详 细地描述本申请实施例的装置。需要说明的是，图18至图20所示的装置可以实现上述方法中各个步骤，为了简洁，在此不再赘述。
图18是本申请实施例提供的通信装置的示意性框图。如图18所示，该通信装置2000可以包括处理单元2100和收发单元2200。
在一种可能的设计中，该通信装置2000可对应于上文方法实施例中的第一通信装置，例如，可以为第一通信装置，或者配置于第一通信装置中的部件(如芯片或芯片系统等)。
应理解，该通信装置2000可对应于根据本申请实施例的方法200、方法300和方法400中的第一通信装置，该通信装置2000可以包括用于执行图10中的方法200、图12中的方法300和图13中的方法400中第一通信装置执行的方法的单元。并且，该通信装置2000中的各单元和上述其他操作和/或功能分别为了实现图10中的方法200、图12中的方法300和图13中的方法400任一方法的相应流程。应理解，各单元执行上述相应步骤的具体过程在上述方法实施例中已经详细说明，为了简洁，在此不再赘述。
在另一种可能的设计中，该通信装置2000可对应于上文方法实施例中的第二通信装置，例如，可以为第二通信装置，或者配置于第二通信装置中的部件(如芯片或芯片系统等)。
应理解，该通信装置2000可对应于根据本申请实施例的方法200、方法300和方法400中的第二通信装置，该通信装置2000可以包括用于执行图10中的方法200、图12中的方法300和图13中的方法400中第二通信装置执行的方法的单元。并且，该通信装置2000中的各单元和上述其他操作和/或功能分别为了实现图10中的方法200、图12中的方法300和图13中的方法400中任一方法的相应流程。应理解，各单元执行上述相应步骤的具体过程在上述方法实施例中已经详细说明，为了简洁，在此不再赘述。
在又一种可能的设计中，该通信装置2000可对应于上文方法实施例中的终端设备，例如，可以为终端设备，或者配置于终端设备中的部件(如芯片或芯片系统等)。
应理解，该通信装置2000可对应于根据本申请实施例的方法500至方法800中的终端设备，该通信装置2000可以包括用于执行图14至图17中的方法500至方法800中终端设备执行的方法的单元。并且，该通信装置2000中的各单元和上述其他操作和/或功能分别为了实现图14至图17中的方法500至方法800中任一方法的相应流程。应理解，各单元执行上述相应步骤的具体过程在上述方法实施例中已经详细说明，为了简洁，在此不再赘述。
还应理解，该通信装置2000为配置于终端设备中的芯片时，该通信装置2000中的收发单元2200可以通过输入/输出接口实现，该通信装置2000中的处理单元2100可以通过该芯片或芯片系统上集成的处理器、微处理器或集成电路等实现。
在又一种可能的设计中，该通信装置2000可对应于上文方法实施例中的网络设备，例如，可以为网络设备，或者配置于网络设备中的部件(如芯片或芯片系统等)。
应理解，该通信装置2000可对应于根据本申请实施例的方法500至方法800中的网络设备，该通信装置2000可以包括用于执行图14至图17中的方法500至方法800中网络设备执行的方法的单元。并且，该通信装置2000中的各单元和上述其他操作和/或功能分别为了实现图14至图17中的方法500至方法800中任一方法的相应流程。应理解，各单元执行上述相应步骤的具体过程在上述方法实施例中已经详细说明，为了简洁，在此不 再赘述。
还应理解，该通信装置2000为配置于网络设备中的芯片时，该通信装置2000中的收发单元2200可以通过输入/输出接口实现，该通信装置2000中的处理单元2100可以通过该芯片或芯片系统上集成的处理器、微处理器或集成电路等实现。
图19是本申请另一实施例的通信装置的示意性框图。图19所示的通信装置3000可以包括：存储器3100、处理器3200、以及通信接口3300。其中，存储器3100、处理器3200，通信接口3300通过内部连接通路相连，该存储器3100用于存储指令，该处理器3200用于执行该存储器3100存储的指令，以控制输入/输出接口3000接收/发送第一消息。可选地，存储器3100既可以和处理器3200通过接口耦合，也可以和处理器3200集成在一起。
需要说明的是，上述通信接口3300使用例如但不限于收发器一类的收发装置，来实现通信装置3000与其他设备或通信网络之间的通信。上述通信接口3300还可以包括输入/输出接口(input/output interface)。
在实现过程中，上述方法的各步骤可以通过处理器3200中的硬件的集成逻辑电路或者软件形式的指令完成。结合本申请实施例所公开的方法可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器3100，处理器3200读取存储器3100中的信息，结合其硬件完成上述方法的步骤。为避免重复，这里不再详细描述。
应理解，本申请实施例中，该处理器可以为中央处理单元(central processing unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
还应理解，本申请实施例中，该存储器可以包括只读存储器和随机存取存储器，并向处理器提供指令和数据。处理器的一部分还可以包括非易失性随机存取存储器。例如，处理器还可以存储设备类型的信息。
图20是本申请实施例的一种芯片系统的示意图。图20所示的芯片系统4000包括：逻辑电路4100以及输入/输出接口(input/output interface)4200，所述逻辑电路用于与输入接口耦合，通过所述输入/输出接口传输数据(例如第一消息)，以执行图10和图12至图17所述的方法。
本申请实施例还提供了一种处理装置，包括处理器和接口；所述处理器用于执行上述任一方法实施例中的方法。
应理解，上述处理装置可以是一个或多个芯片。例如，该处理装置可以是现场可编程门阵列(field programmable gate array，FPGA)，可以是专用集成芯片(application specific integrated circuit，ASIC)，还可以是系统芯片(system on chip，SoC)，还可以是中央处理器(central processor unit，CPU)，还可以是网络处理器(network processor，NP)，还可以是数字信号处理电路(digital signal processor，DSP)，还可以是微控制器(micro  controller unit，MCU)，还可以是可编程控制器(programmable logic device，PLD)或其他集成芯片。
在实现过程中，上述方法的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。结合本申请实施例所公开的方法的步骤可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器，处理器读取存储器中的信息，结合其硬件完成上述方法的步骤。为避免重复，这里不再详细描述。
应注意，本申请实施例中的处理器可以是一种集成电路芯片，具有信号的处理能力。在实现过程中，上述方法实施例的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现场可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器，处理器读取存储器中的信息，结合其硬件完成上述方法的步骤。
可以理解，本申请实施例中的存储器可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。其中，非易失性存储器可以是只读存储器(read-only memory，ROM)、可编程只读存储器(programmable ROM，PROM)、可擦除可编程只读存储器(erasable PROM，EPROM)、电可擦除可编程只读存储器(electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(random access memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(static RAM，SRAM)、动态随机存取存储器(dynamic RAM，DRAM)、同步动态随机存取存储器(synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synchlink DRAM，SLDRAM)和直接内存总线随机存取存储器(direct rambus RAM，DR RAM)。应注意，本文描述的系统和方法的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
根据本申请实施例提供的方法，本申请还提供一种计算机程序产品，该计算机程序产品包括：计算机程序代码，当该计算机程序代码在计算机上运行时，使得该计算机执行图10和图12至图13所示实施例中第一通信装置和第二通信装置分别执行的方法，或者执行图14至图17所示实施例中终端设备和网络设备分别执行的方法。
根据本申请实施例提供的方法，本申请还提供一种计算机可读存储介质，该计算机可读存储介质存储有程序代码，当该程序代码在计算机上运行时，使得该计算机执行图10和图12至图13所示实施例中第一通信装置和第二通信装置分别执行的方法，或者执行图14至图17所示实施例中终端设备和网络设备分别执行的方法。
根据本申请实施例提供的方法，本申请还提供一种系统，其包括前述的一个或多个第一通信装置以及一个或多个第二通信装置。其中，第一通信装置可以是终端设备，第二通信装置可以是网络设备；或者，第一通信装置可以是网络设备，第二通信装置可以是终端设备。
上述各个装置实施例中网络设备与终端设备和方法实施例中的网络设备或终端设备完全对应，由相应的模块或单元执行相应的步骤，例如通信单元(收发器)执行方法实施例中接收或发送的步骤，除发送、接收外的其它步骤可以由处理单元(处理器)执行。具体单元的功能可以参考相应的方法实施例。其中，处理器可以为一个或多个。
在本说明书中使用的术语“部件”、“模块”、“系统”等用于表示计算机相关的实体、硬件、固件、硬件和软件的组合、软件、或执行中的软件。例如，部件可以是但不限于，在处理器上运行的进程、处理器、对象、可执行文件、执行线程、程序和/或计算机。通过图示，在计算设备上运行的应用和计算设备都可以是部件。一个或多个部件可驻留在进程和/或执行线程中，部件可位于一个计算机上和/或分布在2个或更多个计算机之间。此外，这些部件可从在上面存储有各种数据结构的各种计算机可读介质执行。部件可例如根据具有一个或多个数据分组(例如来自与本地系统、分布式系统和/或网络间的另一部件交互的二个部件的数据，例如通过信号与其它系统交互的互联网)的信号通过本地和/或远程进程来通信。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各种说明性逻辑块(illustrative logical block)和步骤(step)，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
在上述实施例中，各功能单元的功能可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机指令(程序)。在计算机上加载和执行所述计算机程序指令(程序)时，全部或部分地产生按照本申请实施例所述的流程或功能。所述 计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(digital subscriber line，DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，高密度数字视频光盘(digital video disc，DVD))、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(read-only memory，ROM)、随机存取存储器(random access memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (39)

  1. 一种通信的方法，其特征在于，包括：
    第一通信装置确定第一信道学习模型是否适用，所述第一信道学习模型用于基于目标信道信息确定第一信道信息，所述第一信道信息的数据量小于所述目标信道信息的数据量；
    在确定所述第一信道学习模型不适用的情况下，所述第一通信装置发送第一消息，所述第一消息用于指示所述第一信道学习模型不适用。
  2. 根据权利要求1所述的方法，其特征在于，所述第一通信装置确定第一信道学习模型是否适用，包括：
    所述第一通信装置在目标信道的长期统计特性的变化量大于或等于第一预设阈值的情况下，确定所述第一信道学习模型不适用；或者，
    所述第一通信装置在目标信道的长期统计特性的变化量小于所述第一预设阈值的情况下，确定所述第一信道学习模型适用。
  3. 根据权利要求1所述的方法，其特征在于，所述第一通信装置确定第一信道学习模型是否适用，包括：
    所述第一通信装置根据接收的第一调度信息确定所述第一信道学习模型是否适用。
  4. 根据权利要求1所述的方法，其特征在于，所述第一通信装置确定第一信道学习模型是否适用，包括：
    所述第一通信装置根据数据传输性能确定所述第一信道学习模型是否适用。
  5. 根据权利要求1所述的方法，其特征在于，所述第一通信装置确定第一信道学习模型是否适用，包括：
    所述第一通信装置在所处的场景发生变化的情况下确定所述第一信道学习模型不适用；或者，
    所述第一通信装置在所处的场景没有发生变化的情况下，确定所述第一信道学习模型适用；
    其中，所述场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、车到其他设备场景、第三代合作伙伴项目协议中定义的场景。
  6. 根据权利要求1所述的方法，其特征在于，所述第一通信装置确定第一信道学习模型是否适用，包括：
    所述第一通信装置在所述第一信道学习模型的性能指标小于第二预设阈值的情况下，确定所述第一信道学习模型不适用，所述性能指标包括连续性和/或真实性；或者，
    所述第一通信装置在所述第一信道学习模型的性能指标大于或等于所述第二预设阈值的情况下，确定所述第一信道学习模型适用。
  7. 根据权利要求1所述的方法，其特征在于，所述第一通信装置确定第一信道学习模型是否适用，包括：
    所述第一通信装置根据所述目标信道信息与所述第二信道信息的误差确定所述第一 信道学习模型是否适用，所述第二信道信息是根据第一信道信息以及第二信道学习模型确定的，所述第二信道学习模型与所述第一信道学习模型对应。
  8. 根据权利要求1至7中任一项所述的方法，其特征在于，所述方法还包括：
    所述第一通信装置接收第一指示信息，所述第一指示信息用于指示所述第一通信装置进行信道学习模型训练。
  9. 根据权利要求8所述的方法，其特征在于，第一指示信息还用于指示以下一项或多项：
    用于传输所述第一消息的资源、所述第一消息的内容、发送所述第一消息的形式、信道学习模型的训练参数，所述训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
  10. 根据权利要求8或9所述的方法，其特征在于，所述方法还包括：
    所述第一通信装置发送第一请求消息，所述第一请求消息用于请求以下一项或多项：进行信道学习模型训练、发送所述第一消息、所述第一指示信息。
  11. 根据权利要求1至10中任一项所述的方法，其特征在于，所述第一消息还用于指示用于更新第二信道学习模型的一个或多个配置参数，所述第二信道学习模型用于根据所述第一信道信息确定第二信道信息。
  12. 根据权利要求1至10中任一项所述的方法，其特征在于，所述方法还包括：
    所述第一通信装置接收第二消息，所述第二消息用于指示用于更新所述第一信道学习模型的一个或多个配置参数。
  13. 根据权利要求11所述的方法，其特征在于，所述方法还包括：
    所述第一通信装置根据第一参数确定用于更新所述第二信道学习模型的一个或多个配置参数；
    其中，所述第一参数包括如下至少一项：所述第一通信装置所在的小区的小区标识、所述第一通信装置所在的场景、所述第一通信装置的类型、所述第一通信装置所在的地理位置。
  14. 一种通信的方法，其特征在于，包括：
    第二通信装置接收第一消息；
    所述第二通信装置根据所述第一消息确定第一信道学习模型不适用，所述第一信道学习模型用于基于目标信道信息确定第一信道信息，所述第一信道信息的维度小于所述目标信道信息的维度；
    所述第二通信装置发送第一指示信息，所述第一指示信息用于指示进行信道学习模型训练。
  15. 根据权利要求14所述的方法，其特征在于，所述第一指示信息还用于指示以下一项或多项：
    用于传输所述第一消息的资源、所述第一消息的内容、所述第一消息的形式、信道学习模型的训练参数，所述训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
  16. 根据权利要求14或15所述的方法，其特征在于，所述方法还包括：
    所述第二通信装置接收第一请求信令，所述第一请求信令用于请求以下一项或多项： 进行信道学习模型训练、发送所述第一消息、所述第一指示信息。
  17. 根据权利要求14至16中任一项所述的方法，其特征在于，所述第一消息还用于指示用于更新第二信道学习模型的一个或多个配置参数，所述第二信道学习模型用于根据所述第一信道信息确定第二信道信息。
  18. 根据权利要求14至17中任一项所述的方法，其特征在于，所述方法还包括：
    所述第二通信装置发送第二消息，所述第二消息用于指示用于更新所述第一信道学习模型的一个或多个配置参数。
  19. 一种通信装置，其特征在于，包括：收发单元和处理单元，
    所述处理单元用于确定第一信道学习模型是否适用，所述第一信道学习模型用于基于目标信道信息确定第一信道信息，所述第一信道信息的数据量小于所述目标信道信息的数据量；
    在确定所述第一信道学习模型不适用的情况下，所述收发单元用于发送第一消息，所述第一消息用于指示所述第一信道学习模型不适用。
  20. 根据权利要求19所述的通信装置，其特征在于，所述处理单元具体用于：在目标信道的长期统计特性的变化量大于或等于第一预设阈值的情况下，确定所述第一信道学习模型不适用；或者，
    在目标信道的长期统计特性的变化量小于所述第一预设阈值的情况下，确定所述第一信道学习模型适用。
  21. 根据权利要求19所述的通信装置，其特征在于，所述处理单元具体用于：根据接收的第一调度信息确定所述第一信道学习模型是否适用。
  22. 根据权利要求19所述的通信装置，其特征在于，所述处理单元具体用于：根据数据传输性能确定所述第一信道学习模型是否适用。
  23. 根据权利要求19所述的通信装置，其特征在于，所述处理单元具有用于：在所处的场景发生变化的情况下确定所述第一信道学习模型不适用；或者，
    在所处的场景没有发生变化的情况下，确定所述第一信道学习模型适用；
    其中，所述场景包括以下至少一种：室内静止、室外静止、低速运动、高速运动、郊区、城镇、宏站、微站、车载场景、车到其他设备场景、第三代合作伙伴项目协议中定义的场景。
  24. 根据权利要求19所述的通信装置，其特征在于，所述处理单元具体用于：在所述第一信道学习模型的性能指标小于第二预设阈值的情况下，确定所述第一信道学习模型不适用，所述性能指标包括连续性和/或真实性；或者，
    在所述第一信道学习模型的性能指标大于或等于所述第二预设阈值的情况下，确定所述第一信道学习模型适用。
  25. 根据权利要求19所述的通信装置法，其特征在于，所述处理单元具体用于：根据所述目标信道信息与所述第二信道信息的误差确定所述第一信道学习模型是否适用，所述第二信道信息是根据第一信道信息以及第二信道学习模型确定的，所述第二信道学习模型与所述第一信道学习模型对应。
  26. 根据权利要求19至25中任一项所述的通信装置，其特征在于，所述收发单元还用于接收第一指示信息，所述第一指示信息用于指示所述第一通信装置进行信道学习模型 训练。
  27. 根据权利要求26所述的通信装置，其特征在于，所述第一指示信息还用于指示以下一项或多项：
    用于传输所述第一消息的资源、所述第一消息的内容、发送所述第一消息的形式、信道学习模型的训练参数，所述训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
  28. 根据权利要求26或27所述的通信装置，其特征在于，所述收发单元还用于发送第一请求消息，所述第一请求消息用于请求以下一项或多项：进行信道学习模型训练、发送所述第一消息、所述第一指示信息。
  29. 根据权利要求19至28中任一项所述的通信装置，其特征在于，所述第一消息还用于指示用于更新第二信道学习模型的一个或多个配置参数，所述第二信道学习模型用于根据所述第一信道信息确定第二信道信息。
  30. 根据权利要求19至28中任一项所述的通信装置，其特征在于，所述收发单元还用于接收第二消息，所述第二消息用于指示用于更新所述第一信道学习模型的一个或多个配置参数。
  31. 根据权利要求29所述的通信装置，其特征在于，所述处理单元还用于根据第一参数确定用于更新所述第二信道学习模型的一个或多个配置参数；
    其中，所述第一参数包括如下至少一项：所述第一通信装置所在的小区的小区标识、所述第一通信装置所在的场景、所述第一通信装置的类型、所述第一通信装置所在的地理位置。
  32. 一种通信装置，其特征在于，包括：收发单元和处理单元，
    所述收发单元用于接收第一消息；
    所述处理单元用于根据所述第一消息确定第一信道学习模型不适用，所述第一信道学习模型用于基于目标信道信息确定第一信道信息，所述第一信道信息的维度小于所述目标信道信息的维度；
    所述收发单元还用于发送第一指示信息，所述第一指示信息用于指示进行信道学习模型训练。
  33. 根据权利要求32所述的通信装置，其特征在于，所述第一指示信息还用于指示以下一项或多项：
    用于传输所述第一消息的资源、所述第一消息的内容、所述第一消息的形式、信道学习模型的训练参数，所述训练参数包括如下至少一项：信道学习模型训练的时间、信道学习模型训练的参考信号的配置信息。
  34. 根据权利要求32或33所述的通信装置，其特征在于，所述收发单元还用于接收第一请求消息，所述第一请求消息用于请求以下一项或多项：进行信道学习模型训练、发送所述第一消息、所述第一指示信息。
  35. 根据权利要求32至34中任一项所述的通信装置法，其特征在于，所述第一消息还用于指示用于更新第二信道学习模型的一个或多个配置参数，所述第二信道学习模型用于根据所述第一信道信息确定第二信道信息。
  36. 根据权利要求32至34中任一项所述的通信装置，其特征在于，所述收发单元还 用于发送第二消息，所述第二消息用于指示用于更新所述第一信道学习模型的一个或多个配置参数。
  37. 一种通信装置，其特征在于，包括至少一个处理器，所述至少一个处理器用于执行存储器中存储的计算机指令，以使得所述通信装置实现如权利要求1至18中任一项所述的方法。
  38. 一种芯片系统，其特征在于，包括：逻辑电路，所述逻辑电路用于与输入/输出接口耦合，通过所述输入/输出接口传输数据，以执行如权利要求1至18中任一项所述的方法。
  39. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机指令，其特征在于，当所述计算机指令被计算设备执行时，使得如权利要求1至18中任一项所述的方法被执行。
PCT/CN2021/100637 2020-07-13 2021-06-17 通信的方法及通信装置 WO2022012257A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/153,876 US20230155702A1 (en) 2020-07-13 2023-01-12 Communication method and communications apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010667447.XA CN113938232A (zh) 2020-07-13 2020-07-13 通信的方法及通信装置
CN202010667447.X 2020-07-13

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/153,876 Continuation US20230155702A1 (en) 2020-07-13 2023-01-12 Communication method and communications apparatus

Publications (1)

Publication Number Publication Date
WO2022012257A1 true WO2022012257A1 (zh) 2022-01-20

Family

ID=79273515

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/100637 WO2022012257A1 (zh) 2020-07-13 2021-06-17 通信的方法及通信装置

Country Status (3)

Country Link
US (1) US20230155702A1 (zh)
CN (1) CN113938232A (zh)
WO (1) WO2022012257A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023179474A1 (zh) * 2022-03-21 2023-09-28 维沃移动通信有限公司 信道特征信息辅助上报及恢复方法、终端和网络侧设备
WO2023211356A1 (en) * 2022-04-29 2023-11-02 Telefonaktiebolaget Lm Ericsson (Publ) User equipment machine learning functionality monitoring
WO2023206207A1 (en) * 2022-04-28 2023-11-02 Qualcomm Incorporated Model management for channel state estimation and feedback
WO2023212059A1 (en) * 2022-04-27 2023-11-02 Interdigital Patent Holdings, Inc. Methods and apparatus for leveraging transfer learning for channel state information enhancement
WO2024033809A1 (en) * 2022-08-09 2024-02-15 Lenovo (Singapore) Pte. Ltd. Performance monitoring of a two-sided model
WO2024065709A1 (zh) * 2022-09-30 2024-04-04 华为技术有限公司 一种通信方法及相关设备
WO2024091533A1 (en) * 2022-10-27 2024-05-02 Apple Inc. Enhanced data collection for csi compression modeling
WO2024197929A1 (en) * 2023-03-31 2024-10-03 Nec Corporation Methods, devices, and computer readable medium for communication

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116233857A (zh) * 2021-12-02 2023-06-06 华为技术有限公司 通信方法和通信装置
CN116801280A (zh) * 2022-03-18 2023-09-22 维沃移动通信有限公司 Ai模型的处理方法及设备
CN116846493A (zh) * 2022-03-24 2023-10-03 维沃移动通信有限公司 信道预测方法、装置及无线通信设备
CN116939650A (zh) * 2022-04-01 2023-10-24 维沃移动通信有限公司 信道特征信息传输方法、装置、终端及网络侧设备
CN114826941A (zh) * 2022-04-27 2022-07-29 中国电子科技集团公司第五十四研究所 一种无线通信网络ai模型配置方法
WO2023206445A1 (zh) * 2022-04-29 2023-11-02 富士通株式会社 Ai监测装置以及方法
WO2024040577A1 (en) * 2022-08-26 2024-02-29 Apple Inc. Technologies for user equipment-trained artificial intelligence models
WO2024092766A1 (zh) * 2022-11-04 2024-05-10 富士通株式会社 模型管理的方法、装置和通信系统
CN117998398A (zh) * 2022-11-04 2024-05-07 维沃移动通信有限公司 确定模型适用性的方法、装置及通信设备
CN115494472B (zh) * 2022-11-16 2023-03-10 中南民族大学 一种基于增强雷达波信号的定位方法、毫米波雷达、装置
CN115882932B (zh) * 2022-11-30 2023-11-17 湖南迈克森伟电子科技有限公司 一种卫星与地面终端的通讯方法和系统
CN118118921A (zh) * 2022-11-30 2024-05-31 华为技术有限公司 一种通信方法及装置
WO2024138435A1 (zh) * 2022-12-28 2024-07-04 Oppo广东移动通信有限公司 用于通信的方法以及通信设备
CN118449662A (zh) * 2023-02-06 2024-08-06 大唐移动通信设备有限公司 一种性能评估方法、设备及可读存储介质
CN116095781B (zh) * 2023-02-10 2023-08-15 惠州市源医科技有限公司 一种信道切换方法及5g智能路由器
CN118509142A (zh) * 2023-02-16 2024-08-16 大唐移动通信设备有限公司 信息传输方法、装置及设备
CN116886475B (zh) * 2023-09-05 2023-12-22 中国信息通信研究院 一种信道估计方法、设备及系统
CN117713885B (zh) * 2024-02-06 2024-06-18 荣耀终端有限公司 一种通信方法、通信装置及通信系统

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108512621A (zh) * 2018-03-02 2018-09-07 东南大学 一种基于神经网络的无线信道建模方法
WO2020080989A1 (en) * 2018-10-19 2020-04-23 Telefonaktiebolaget Lm Ericsson (Publ) Handling of machine learning to improve performance of a wireless communications network

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108512621A (zh) * 2018-03-02 2018-09-07 东南大学 一种基于神经网络的无线信道建模方法
WO2020080989A1 (en) * 2018-10-19 2020-04-23 Telefonaktiebolaget Lm Ericsson (Publ) Handling of machine learning to improve performance of a wireless communications network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
OPPO CMCC CHINA TELECOM CHINA UNICOM: "Discussion on AI/ML Model Transfer in 5GS", 3GPP DRAFT; S1-193039, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. SA WG1, no. Reno, USA; 20191118 - 20191122, 8 November 2019 (2019-11-08), Mobile Competence Centre ; 650, route des Lucioles ; F-06921 Sophia-Antipolis Cedex ; France , XP051822342 *
OPPO, CMCC, CHINA TELECOM, CHINA UNICOM, QUALCOMM: "SID Proposal: Study on AI/ML Model Transfer in 5GS", 3GPP DRAFT; S1-193479, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. SA WG1, no. Reno, USA; 20191118 - 20191122, 25 November 2019 (2019-11-25), Mobile Competence Centre ; 650, route des Lucioles ; F-06921 Sophia-Antipolis Cedex ; France , XP051831222 *
ZTE: "Solution to AI Model sharing architecture", 3GPP DRAFT; S2-2003632, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. SA WG2, no. Online; 20200601 - 20200612, 22 May 2020 (2020-05-22), Mobile Competence Centre ; 650, route des Lucioles ; F-06921 Sophia-Antipolis Cedex ; France , XP051889688 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023179474A1 (zh) * 2022-03-21 2023-09-28 维沃移动通信有限公司 信道特征信息辅助上报及恢复方法、终端和网络侧设备
WO2023212059A1 (en) * 2022-04-27 2023-11-02 Interdigital Patent Holdings, Inc. Methods and apparatus for leveraging transfer learning for channel state information enhancement
WO2023206207A1 (en) * 2022-04-28 2023-11-02 Qualcomm Incorporated Model management for channel state estimation and feedback
WO2023211356A1 (en) * 2022-04-29 2023-11-02 Telefonaktiebolaget Lm Ericsson (Publ) User equipment machine learning functionality monitoring
WO2024033809A1 (en) * 2022-08-09 2024-02-15 Lenovo (Singapore) Pte. Ltd. Performance monitoring of a two-sided model
WO2024065709A1 (zh) * 2022-09-30 2024-04-04 华为技术有限公司 一种通信方法及相关设备
WO2024091533A1 (en) * 2022-10-27 2024-05-02 Apple Inc. Enhanced data collection for csi compression modeling
WO2024197929A1 (en) * 2023-03-31 2024-10-03 Nec Corporation Methods, devices, and computer readable medium for communication

Also Published As

Publication number Publication date
CN113938232A (zh) 2022-01-14
US20230155702A1 (en) 2023-05-18

Similar Documents

Publication Publication Date Title
WO2022012257A1 (zh) 通信的方法及通信装置
WO2022033456A1 (zh) 一种信道状态信息测量反馈方法及相关装置
US11984955B2 (en) Configurable neural network for channel state feedback (CSF) learning
US20220070822A1 (en) Unsupervised learning for simultaneous localization and mapping in deep neural networks using channel state information
US12056609B2 (en) Gradient feedback framework for joint transceiver neural network training
US11863495B2 (en) Signaling for a channel state information reference signal (CSI-RS)
US11949615B2 (en) User equipment (UE) feedback of quantized per-path angle of arrival
WO2022267633A1 (zh) 信息传输的方法和装置
KR20240024082A (ko) 모델 기반 뉴럴 네트워크들을 통한 채널 특징 추출
CN117813913A (zh) 使用神经网络进行信源编码的方法和系统
US11929853B2 (en) Data-driven probabilistic modeling of wireless channels using conditional variational auto-encoders
KR20240115239A (ko) 시맨틱 기반 무선 통신 시스템에서 데이터를 전송하는 방법, 전송 기기, 프로세싱 장치 및 저장 매체, 그리고 데이터를 수신하는 방법, 수신 기기 및 저장 매체
US20240171429A1 (en) Communication method and apparatus
US11792877B2 (en) Indication triggering transmission of known data for training artificial neural networks
US11456834B2 (en) Adaptive demodulation reference signal (DMRS)
WO2023146862A1 (en) Equivariant generative prior for inverse problems with unknown rotation
CN117693021A (zh) 一种波束管理方法
TW202201935A (zh) 用於無線通道估計和追蹤的神經網路增強
US20240348479A1 (en) Channel state feedback for reduced resource consumption reference signals
US20230239179A1 (en) Equivariant generative prior for inverse problems with unknown rotation
US20230412227A1 (en) Method and apparatus for transmitting or receiving information for artificial intelligence based channel state information feedback in wireless communication system
WO2024008004A1 (zh) 一种通信方法及装置
WO2024046288A1 (zh) 通信方法和装置
US20240349227A1 (en) Node in a wireless communication system and method executed by the same
KR20240118769A (ko) 무선 통신 시스템에서 시맨틱 통신을 위한 지식을 갱신하는 방법, 통신 기기, 프로세싱 장치 및 저장 매체

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21842752

Country of ref document: EP

Kind code of ref document: A1