WO2023125699A1 - 一种通信方法及装置 - Google Patents

一种通信方法及装置 Download PDF

Info

Publication number
WO2023125699A1
WO2023125699A1 PCT/CN2022/142946 CN2022142946W WO2023125699A1 WO 2023125699 A1 WO2023125699 A1 WO 2023125699A1 CN 2022142946 W CN2022142946 W CN 2022142946W WO 2023125699 A1 WO2023125699 A1 WO 2023125699A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
downlink channel
information
sub
channel data
Prior art date
Application number
PCT/CN2022/142946
Other languages
English (en)
French (fr)
Inventor
杭海存
张文凯
陈家璇
梁璟
吴艺群
陈志堂
金黄平
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 WO2023125699A1 publication Critical patent/WO2023125699A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition

Definitions

  • the present disclosure relates to the field of communication technologies, and in particular, to a communication method and device.
  • the fifth generation (the 5th generation, 5G) mobile communication system has higher requirements on system capacity and spectrum efficiency.
  • the application of massive multiple-input multiple-output (massive-MIMO) technology plays a vital role in improving the spectral efficiency of the system.
  • the base station can provide high-quality services for more user equipments (user equipment, UE) at the same time.
  • the key link is that the base station precodes the downlink data of multiple UEs.
  • spatial multiplexing spatial multiplexing
  • SINR signal to interference plus noise ratio
  • the base station can obtain the channel state information (channel state information, CSI) of the downlink channel, recover the downlink channel according to the CSI, and use the recovered downlink channel to determine the precoding matrix to perform precoded. Therefore, how to make the CSI fed back by the UE more accurate is a technical issue worth studying.
  • CSI channel state information
  • the present disclosure provides a communication method and device for improving the accuracy of CSI fed back by UE.
  • a first communication method is provided, and the method can be executed on a terminal device side.
  • the method can be implemented by software, hardware, or a combination of software and hardware.
  • the method is performed by a terminal device, or by a circuit system, or by a larger device including the terminal device, and the circuit system can realize the functions of the terminal device.
  • the method includes: obtaining M pieces of first sub-downlink channel data, wherein each piece of first sub-downlink channel data corresponds to one data space in M data spaces, and M is an integer greater than 1; for the M pieces of first sub-downlink channel data For the i-th piece of first sub-downlink channel data in the downlink channel data, determine the i-th piece of first sub-downlink channel data corresponding to the i-th piece of first sub-downlink channel data according to the first dictionary corresponding to the i-th data space in the M data spaces.
  • One piece of information a total of M pieces of first information are determined, i takes an integer from 1 to M, the i-th part of the first sub-downlink channel data corresponds to the i-th data space, and the first dictionary includes a plurality of elements , the first information corresponding to the i-th piece of first sub-downlink channel data corresponds to P elements in the plurality of elements, and P is a positive integer; sending first indication information, the first indication information is used to indicate The M pieces of first information.
  • each piece of first sub-downlink channel data among the M pieces of first sub-downlink channel data obtained by the terminal device may correspond to one of the M data spaces, and the dictionaries corresponding to different data spaces can determine each A share of first information corresponding to the first sub-downlink channel data.
  • Different data spaces can represent different location information, or can represent different channel environment information, and the terminal device can feed back the first information corresponding to different data spaces, which can enable the access network device to clarify the relationship between the first information and the environment information. The corresponding relationship among them, so that the first information fed back by the terminal device can reflect the actual communication environment, and the accuracy of the first information fed back by the terminal device is improved.
  • the access network device can restore and obtain a more accurate downlink channel according to the first information fed back by the terminal device.
  • the first indication information is used to indicate identities of the M pieces of first information
  • sending the first information includes: sending the M pieces of first information in a first order
  • An identifier of information is an arrangement order of the M data spaces.
  • the first order specifies which data space the terminal device first sends the identifier of the first information corresponding to, and which data space corresponds to the identifier of the first information.
  • the access network device can also clarify the correspondence between the identifiers of the first information and the data spaces after receiving the M identifiers of the first information relationship, so as to avoid corresponding errors.
  • the first order is a predefined order; or, receiving second indication information, where the second indication information is used to indicate the first order; or, determining the first order a sequence, and send third indication information, where the third indication information is used to indicate the first sequence.
  • the first sequence is a sequence predefined by the protocol, and the terminal device and the access network device can determine the first sequence according to the protocol.
  • the first order may also be preconfigured in the terminal device and the access network device.
  • the first order may be determined by the access network device, and the access network device may send the second indication information to the terminal device after determining the first order, so that the terminal device can determine the first order according to the second indication information.
  • the first order may be determined by the terminal device, and after determining the first order, the terminal device may send third indication information to the access network device, so that the access network device can determine the first order according to the third indication information. It can be seen that the manner of determining the first order is relatively flexible.
  • the M pieces of first sub-downlink channel data are obtained according to the first downlink channel data, wherein the first downlink channel data is a preprocessing result; or, the The first downlink channel data includes continuous column F data in the preprocessing result; or, the first downlink channel data is compressed information obtained by compressing the preprocessing result.
  • the preprocessing result is obtained by preprocessing the second downlink channel data.
  • the preprocessing result of the second downlink channel data can be directly used as the first downlink channel data, and there is no need to perform excessive processing on the preprocessing result, which is relatively simple.
  • the preprocessing result may also be compressed to obtain the first downlink channel data, thereby reducing the complexity of processing the first downlink channel data.
  • the process of preprocessing a downlink channel data includes, for example, performing space-frequency joint projection on the downlink channel data.
  • the division manner of the M data spaces is predefined; or, receiving fourth indication information, where the fourth indication information is used to indicate the division manner of the M data spaces; Or, determine the division manner of the M data spaces, and send fifth indication information, where the fifth indication information is used to indicate the division manner of the M data spaces.
  • the division method of the M data spaces is predefined by the protocol, and both the terminal device and the access network device can determine the division method of the M data spaces according to the protocol.
  • the division method of the M data spaces is determined by the access network device, and the access network device may send fourth indication information to the terminal device, so that the terminal device can determine the division mode of the M data spaces according to the fourth indication information.
  • the division manner of the M data spaces may be determined by the UE, and the UE may send fifth indication information to the access network device, so that the access network device may determine the division manner of the M data spaces according to the fifth indication information. It can be seen that the division method of the data space is more flexible.
  • a second communication method is provided, and the method can be executed on an access network device side.
  • the method can be implemented by software, hardware, or a combination of software and hardware.
  • the method is performed by an access network device, or by a larger device including the access network device, or by a circuit system capable of implementing the functions of the access network device, or by a device independent of the access network device
  • the access network device is an access network device, such as a base station.
  • the method includes: receiving first indication information, where the first indication information is used to indicate M pieces of first information, where M is an integer greater than 1; for the ith first information among the M pieces of first information, according to The first dictionary corresponding to the i-th data space in the M data spaces restores the i-th second sub-downlink channel data, and obtains M second sub-downlink channel data in total, and the i-th first information corresponds to the i-th second sub-downlink channel data.
  • i takes an integer from 1 to M
  • the first dictionary includes a plurality of elements
  • the first information corresponding to the i-th second sub-downlink channel data corresponds to the plurality of elements P elements of ; according to the M shares of second sub-downlink channel data, recover and obtain downlink channel information.
  • receiving the first indication information includes: receiving the M identifiers of the first information in a first order, where the first order is an arrangement order of the M data spaces.
  • the first order is a predefined order; or, sending second indication information, where the second indication information is used to indicate the first order; or, receiving a third indication information, the third indication information is used to indicate the first order.
  • the M data spaces correspond to M dictionaries, wherein each data space corresponds to a dictionary; or, the M data spaces all correspond to the same dictionary; or, the M The number of dictionaries corresponding to each data space is greater than 1 and less than M.
  • the data space and the dictionary can have a one-to-one correspondence, which can improve the accuracy of the first information determined according to the dictionary; or, all the data spaces can be uniformly corresponding to a dictionary, and the samples used for training to obtain the dictionary can be It is richer, so that the content included in the dictionary is more detailed; or, the number of dictionaries corresponding to the data space can be smaller than the number of data spaces, for example, a dictionary can correspond to multiple data spaces, which can reduce the complexity to a certain extent Spend.
  • recovering and obtaining downlink channel information according to the M pieces of second sub-downlink channel data includes: obtaining compressed information according to the M pieces of second sub-downlink channel data; Obtain the downlink channel information.
  • the division method of the M data spaces is predefined; or, sending fourth indication information, where the fourth indication information is used to indicate the division mode of the M data spaces; Or, receiving fifth indication information, where the fifth indication information is used to indicate the division manner of the M data spaces.
  • a communication device may implement the method described in the first aspect above.
  • the communication device has the functions of the terminal device described above.
  • the device may include a one-to-one corresponding module for executing the method/operation/step/action described in the first aspect, and the module may be a hardware circuit, software, or Hardware circuit combined with software implementation.
  • the communication device includes a baseband device and a radio frequency device.
  • the communication device includes a processing unit (also called a processing module sometimes) and a transceiver unit (also called a transceiver module sometimes). The transceiver unit can realize the sending function and the receiving function.
  • the transceiver unit When the transceiver unit realizes the sending function, it can be called the sending unit (sometimes also called the sending module). When the transceiver unit realizes the receiving function, it can be called the receiving unit (sometimes also called receiving module).
  • the sending unit and the receiving unit can be the same functional module, which is called the transceiver unit, and this functional module can realize the sending function and the receiving function; or, the sending unit and the receiving unit can be different functional modules, and the transceiver unit is for these A general term for functional modules.
  • the processing unit is configured to obtain M pieces of first sub-downlink channel data, wherein each piece of first sub-downlink channel data corresponds to one data space in M data spaces, and M is an integer greater than 1; for the For the i-th piece of first sub-downlink channel data in the M pieces of first sub-downlink channel data, the processing unit is further configured to determine the i-th piece of sub-downlink channel data according to the first dictionary corresponding to the i-th data space in the M data spaces.
  • the first information corresponding to the i-th part of the first sub-downlink channel data a total of M pieces of first information are determined, i takes an integer from 1 to M, and the i-th part of the first sub-downlink channel data corresponds to the i-th data space, the first dictionary includes a plurality of elements, the first information corresponding to the i-th first sub-downlink channel data corresponds to P elements in the plurality of elements, and P is a positive integer; the A transceiving unit, configured to send first indication information, where the first indication information is used to indicate the M pieces of first information.
  • the communication device includes: a processor, coupled to the memory, configured to execute instructions in the memory, so as to implement the method in the first aspect above.
  • the communication device further includes other components, for example, an antenna, an input and output module, an interface, and the like. These components can be hardware, software, or a combination of software and hardware.
  • a communication device may implement the method described in the second aspect above.
  • the communication device has the functions of the above-mentioned access network equipment.
  • the access network equipment is, for example, a base station, or a baseband device in a base station, and the like.
  • the device may include a one-to-one corresponding module for executing the method/operation/step/action described in the second aspect, and the module may be a hardware circuit, or software, or a Hardware circuit combined with software implementation.
  • the communication device includes a baseband device and a radio frequency device.
  • the communication device includes a processing unit (also called a processing module sometimes) and a transceiver unit (also called a transceiver module sometimes).
  • a processing unit also called a processing module sometimes
  • a transceiver unit also called a transceiver module sometimes.
  • the transceiving unit is configured to receive first indication information, the first indication information is used to indicate M pieces of first information, and M is an integer greater than 1; for the i-th piece of the M pieces of first information
  • the first information the processing unit, is used to restore the i-th part of the second sub-downlink channel data according to the first dictionary corresponding to the i-th data space in the M data spaces, and obtain M parts of the second sub-downlink channel data in total , the i-th first information corresponds to the i-th data space, i takes an integer from 1 to M, the first dictionary includes a plurality of elements, and the i-th second sub-downlink channel data corresponds to The first information corresponds to P elements among the plurality of elements; the processing unit is further configured to recover and obtain downlink channel information according to the M pieces of second sub-downlink channel data.
  • the communications device includes: a processor, coupled to the memory, configured to execute instructions in the memory, so as to implement the method in the second aspect above.
  • the communication device further includes other components, for example, an antenna, an input and output module, an interface, and the like. These components can be hardware, software, or a combination of software and hardware.
  • a computer-readable storage medium is provided, the computer-readable storage medium is used for storing computer programs or instructions, and when executed, the method of the first aspect and/or the second aspect is implemented.
  • a computer program product containing instructions is provided, and when it is run on a computer, the method described in the first aspect and/or the second aspect is implemented.
  • a chip system in a seventh aspect, includes a processor and may further include a memory, configured to implement the methods of the first aspect and/or the second aspect above.
  • the system-on-a-chip may consist of chips, or may include chips and other discrete devices.
  • a communication system including the communication device of the third aspect and the communication device of the fourth aspect.
  • Fig. 1 is a schematic diagram of a communication system
  • Fig. 2 is a flowchart of the CSI feedback mechanism
  • FIG. 3 is a schematic diagram of an application scenario
  • 4A to 4E are schematic diagrams of several application frameworks of AI in communication systems
  • Fig. 5 is a flow chart of a communication method
  • Fig. 6 is a kind of schematic diagram of dictionary
  • FIG. 7 is a schematic diagram of a communication method in a case where both the UE and the access network device process compressed information
  • Fig. 8 is a flowchart of another communication method
  • FIG. 9A is a schematic diagram of a network training phase and a network reasoning phase
  • 9B to 9D are several schematic diagrams of the network training phase
  • Fig. 10 is a flowchart of another communication method
  • Fig. 11 is another schematic diagram of the network training phase and the network reasoning phase
  • Fig. 12 is a schematic block diagram of a communication device.
  • the technology provided by this disclosure can be applied to the communication system 10 shown in FIG. 1 .
  • the communication system 10 includes one or more communication devices 30 (eg, terminal equipment).
  • the one or more communication devices 30 are connected to one or more core network (core network, CN) devices via one or more access network (radio access network, RAN) devices 20, so as to realize communication between multiple communication devices communication.
  • core network core network
  • RAN radio access network
  • the communication system 10 is a communication system that supports the fourth generation (the 4th generation, 4G) (including long term evolution (long term evolution, LTE)) access technology, and supports 5G (sometimes also called new radio, NR) access communication systems with advanced technology, wireless fidelity (Wireless Fidelity, Wi-Fi) systems, cellular systems related to the 3rd Generation Partnership Project (3GPP), communication systems that support the integration of multiple wireless technologies, or oriented to Future evolution systems, etc., are not limited.
  • 4G fourth generation
  • LTE long term evolution
  • 5G sometimes also called new radio, NR
  • 5G sometimes also called new radio, NR
  • advanced technology wireless fidelity (Wireless Fidelity, Wi-Fi) systems
  • 3GPP 3rd Generation Partnership Project
  • 3GPP 3rd Generation Partnership Project
  • the terminal equipment and RAN involved in FIG. 1 will be described in detail below respectively.
  • a terminal device may be referred to simply as a terminal.
  • the terminal device may be a device with a wireless transceiver function.
  • Terminal equipment can be mobile or fixed. Terminal equipment can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; it can also be deployed on water (such as ships, etc.); it can also be deployed in the air (such as aircraft, balloons and satellites, etc.).
  • the terminal device may include a mobile phone, a tablet computer (pad), a computer with a wireless transceiver function, a virtual reality (virtual reality, VR) terminal device, an augmented reality (augmented reality, AR) terminal device, an industrial control ( Wireless terminal devices in industrial control, wireless terminal devices in self driving, wireless terminal devices in remote medical, wireless terminal devices in smart grid, transportation security safety), a wireless terminal device in a smart city, and/or a wireless terminal device in a smart home.
  • a virtual reality virtual reality
  • AR augmented reality
  • an industrial control Wireless terminal devices in industrial control, wireless terminal devices in self driving, wireless terminal devices in remote medical, wireless terminal devices in smart grid, transportation security safety
  • a wireless terminal device in a smart city and/or a wireless terminal device in a smart home.
  • the terminal device can also be a cellular phone, a cordless phone, a session initiation protocol (session initiation protocol, SIP) phone, a wireless local loop (wireless local loop, WLL) station, a personal digital assistant (personal digital assistant, PDA), a Functional handheld devices or computing devices, vehicle-mounted devices, wearable devices, terminal devices in the future fifth generation (the 5th generation, 5G) network or in the future evolution of the public land mobile network (PLMN) terminal equipment, etc.
  • the terminal equipment may sometimes also be referred to as user equipment (user equipment, UE).
  • the terminal device can communicate with multiple access network devices of different technologies.
  • the terminal device can communicate with access network devices supporting LTE, or with access network devices supporting 5G, and with Dual connectivity of LTE-capable access network devices and 5G-capable access network devices. This disclosure is not limiting.
  • the device for realizing the function of the terminal device may be a terminal device; it may also be a device capable of supporting the terminal device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. It can be installed in the terminal equipment or can be matched with the terminal equipment.
  • the technical solution provided by the present disclosure is described by taking the terminal device as an example where the apparatus for realizing the function of the terminal device is a terminal device.
  • a system-on-a-chip may be composed of chips, and may also include chips and other discrete devices.
  • the RAN may include one or more RAN devices, such as RAN device 20 .
  • the interface between the RAN device and the terminal device may be a Uu interface (or called an air interface).
  • Uu interface or called an air interface.
  • the names of these interfaces may remain unchanged, or may be replaced by other names, which is not limited in the present disclosure.
  • a RAN device is a node or a device that connects a terminal device to a wireless network, and the RAN device may also be called a network device or a base station.
  • RAN equipment includes, but is not limited to: base stations, next-generation node B (generation nodeB, gNB), 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 (base band unit, BBU), sending and receiving point (transmitting and receiving point, TRP), transmitting point (transmitting point, TP), and/or mobile switching center, etc.
  • generation nodeB generation nodeB, gNB
  • evolved node B evolved node B
  • eNB evolved node B
  • RNC radio network
  • the access network device may also be a centralized unit (centralized unit, CU), a distributed unit (distributed unit, DU), a centralized unit control plane (CU control plane, CU-CP) node, a centralized unit user plane (CU user plane , CU-UP) node, integrated access and backhaul (integrated access and backhaul, IAB), or at least one of wireless controllers in a cloud radio access network (cloud radio access network, CRAN) scenario.
  • the access network device may be a relay station, an access point, a vehicle-mounted device, a terminal device, a wearable device, an access network device in a 5G network, or a device in a public land mobile network (PLMN) that will evolve in the future. Access network equipment, etc.
  • the device for realizing the function of the access network device may be the access network device; it may also be a device capable of supporting the access network device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit Adding a software module, the device can be installed in the access network equipment or can be matched and used with the access network equipment.
  • the technical solution provided by the present disclosure is described by taking the apparatus for realizing the function of the access network device as the access network device and the access network device as a base station as an example.
  • the protocol layer structure may include a control plane protocol layer structure and a user plane protocol layer structure.
  • the control plane protocol layer structure may include at least one of the following: a radio resource control (radio resource control, RRC) layer, a packet data convergence protocol (packet data convergence protocol, PDCP) layer, a radio link control (radio link control, RLC) layer, media access control (media access control, MAC) layer or physical layer (physical, PHY), etc.
  • the user plane protocol layer structure may include at least one of the following: a service data adaptation protocol (service data adaptation protocol, SDAP) layer, a PDCP layer, an RLC layer, a MAC layer, and a physical layer.
  • the above protocol layer structure between the access network device and the terminal device can be regarded as an access stratum (access stratum, AS) structure.
  • AS access stratum
  • NAS non-access stratum
  • the access network device may forward information between the terminal device and the core network device through transparent transmission.
  • the NAS message may be mapped to or included in RRC signaling as an element of RRC signaling.
  • the protocol layer structure between the access network device and the terminal device may further include an artificial intelligence (AI) layer, which is used to transmit data related to the AI function.
  • AI artificial intelligence
  • RAN devices may include CUs and DUs. This design can be called CU and DU separation. Multiple DUs can be centrally controlled by one CU.
  • the interface between the CU and the DU may be referred to as an F1 interface.
  • the control plane (control panel, CP) interface may be F1-C
  • the user plane (user panel, UP) interface may be F1-U.
  • the present disclosure does not limit the specific names of the interfaces.
  • CU and DU can be divided according to the protocol layer of the wireless network: for example, the functions of the PDCP layer and above protocol layers (such as RRC layer and SDAP layer, etc.) etc.) functions are set in the DU; as shown in another example, the functions of the protocol layers above the PDCP layer are set in the CU, and the functions of the PDCP layer and the protocol layers below are set in the DU.
  • the functions of the PDCP layer and above protocol layers such as RRC layer and SDAP layer, etc.
  • the CU or DU may be divided into functions having more protocol layers, and for example, the CU or DU may also be divided into part processing functions having protocol layers.
  • part of the functions of the RLC layer and the functions of the protocol layers above the RLC layer are set in the CU, and the rest of the functions of the RLC layer and the functions of the protocol layers below the RLC layer are set in the DU.
  • the functions of the CU or DU can also be divided according to the business type or other system requirements, for example, according to the delay, and the functions whose processing time needs to meet the delay requirement are set in the DU, which does not need to meet the delay
  • the required feature set is in the CU.
  • the CU may also have one or more functions of the core network.
  • the CU can be set on the network side to facilitate centralized management.
  • the wireless unit (radio unit, RU) of the DU is remotely set.
  • the RU has a radio frequency function.
  • DUs and RUs can be divided at the PHY layer.
  • the DU can implement high-level functions in the PHY layer
  • the RU can implement low-level functions in the PHY layer.
  • the functions of the PHY layer may include at least one of the following: adding a cyclic redundancy check (cyclic redundancy check, CRC) bit, channel coding, rate matching, scrambling, modulation, layer mapping, precoding, Resource mapping, physical antenna mapping, or radio frequency transmission functions.
  • CRC cyclic redundancy check
  • the functions of the PHY layer may include at least one of the following: CRC check, channel decoding, de-rate matching, descrambling, demodulation, de-layer mapping, channel detection, resource de-mapping, physical antenna de-mapping, or RF receiving function.
  • the high-level functions in the PHY layer may include part of the functions of the PHY layer, which are closer to the MAC layer; the lower-level functions in the PHY layer may include another part of the functions of the PHY layer, for example, this part of functions is closer to the radio frequency function.
  • high-level functions in the PHY layer may include adding CRC bits, channel coding, rate matching, scrambling, modulation, and layer mapping
  • low-level functions in the PHY layer may include precoding, resource mapping, physical antenna mapping, and radio transmission functions
  • high-level functions in the PHY layer can include adding CRC bits, channel coding, rate matching, scrambling, modulation, layer mapping, and precoding
  • low-level functions in the PHY layer can include resource mapping, physical antenna mapping, and radio frequency send function.
  • the high-level functions in the PHY layer may include CRC check, channel decoding, de-rate matching, decoding, demodulation, and de-layer mapping
  • the low-level functions in the PHY layer may include channel detection, resource de-mapping, physical antenna de-mapping, and RF receiving functions
  • the high-level functions in the PHY layer may include CRC check, channel decoding, de-rate matching, decoding, demodulation, de-layer mapping, and channel detection
  • the low-level functions in the PHY layer may include resource de-mapping , physical antenna demapping, and RF receiving functions.
  • the functions of the CU can be further divided, and the control plane and the user plane are separated and realized by different entities, namely, the control plane CU entity (ie, the CU-CP entity) and the user plane CU entity (ie, the CU-UP entity). entity).
  • the CU-CP entity and the CU-UP entity can be coupled or connected to the DU respectively to jointly complete the functions of the RAN device.
  • the signaling generated by the CU can be sent to the terminal device through the DU, or the signaling generated by the terminal device can be sent to the CU through the DU.
  • signaling at the RRC or PDCP layer can be finally processed as signaling at the physical layer and sent to the terminal device, or converted from received signaling at the physical layer.
  • the signaling at the RRC or PDCP layer can be considered to be sent through the DU, or sent through the DU and the RU.
  • any one of the foregoing DU, CU, CU-CP, CU-UP, and RU may be a software module, a hardware structure, or a software module+hardware structure, without limitation.
  • the existence forms of different entities may be different, which is not limited.
  • DU, CU, CU-CP, and CU-UP are software modules
  • RU is a hardware structure.
  • modules and the methods performed by them are also within the protection scope of the present disclosure.
  • the method of the present disclosure when executed by an access network device, it may specifically be executed by at least one of CU, CU-CP, CU-UP, DU, RU, or near real-time RIC described below.
  • the methods performed by each module are also within the protection scope of the present disclosure.
  • network equipment since the network equipment involved in this disclosure is mainly access network equipment, in the following, unless otherwise specified, the “network equipment” may refer to "access network equipment”.
  • the number of devices in the communication system shown in FIG. 1 is only for illustration, and the present disclosure is not limited thereto.
  • the communication system may include more terminal devices, more RAN devices, or It includes other devices, for example, may include core network devices, and/or nodes for implementing artificial intelligence functions.
  • the network architecture shown in Figure 1 above can be applied to various radio access technology (radio access technology, RAT) communication systems, such as 4G communication systems, or 5G (or new radio (NR))
  • RAT radio access technology
  • the communication system may also be a transition system between the LTE communication system and the 5G communication system, and the transition system may also be called a 4.5G communication system, or may also be a future communication system, such as a 6G communication system.
  • RAT radio access technology
  • the network architecture and business scenarios described in this disclosure are for more clearly illustrating the technical solution of this disclosure, and do not constitute a limitation to the technical solution provided by this disclosure. Those of ordinary skill in the art know that with the evolution and new For the emergence of business scenarios, the technical solutions provided in this disclosure are also applicable to similar technical problems.
  • the method provided in this disclosure can be used not only for communication between access network equipment and terminal equipment, but also for communication between other communication equipment, such as communication between a macro base station and a micro base station in a wireless backhaul link,
  • communication between the first terminal device and the second terminal device in a sidelink is not limited.
  • the present disclosure uses communication between a network device and a terminal device as an example for description.
  • precoding may be performed based on channel state information (channel state information, CSI) fed back by the terminal device.
  • channel state information channel state information, CSI
  • the access network device can process the signal to be transmitted with the help of the precoding matrix that matches the channel condition when the channel state information is known.
  • the precoded signal to be sent can be adapted to the channel, so that the quality of the signal received by the terminal device (such as signal to interference plus noise ratio (SINR), etc.) can be improved. , which can improve the system throughput.
  • SINR signal to interference plus noise ratio
  • Using precoding technology it is possible to realize effective transmission on the same time-frequency resources between a sending device (such as an access network device) and multiple receiving devices (such as a terminal device), that is, to effectively realize multiple user multiple input multiple output (multiple user multiple input multiple output, MU-MIMO).
  • the sending device (such as an access network device) and the receiving device (such as a terminal device) can effectively transmit multiple data streams on the same time-frequency resources, that is, effectively realize single-user multiple-input multiple-output ( single user multiple input multiple output, SU-MIMO).
  • the relevant description about the precoding technology is only an example for easy understanding, and is not intended to limit the disclosure scope of the present disclosure.
  • the sending device may also perform precoding in other ways. For example, in a case where channel information (such as but not limited to a channel matrix) cannot be obtained, a pre-set precoding matrix or a weighting processing manner is used to perform precoding and the like. For the sake of brevity, its specific content will not be repeated here.
  • the CSI feedback may also be referred to as a CSI report (CSI report).
  • CSI feedback is in the wireless communication system, from the receiving end (such as the terminal device) of the data (such as but not limited to the data carried on the physical downlink shared channel (PDSCH)) to the sending end (such as the access network device) ) to report the information used to describe the channel properties of the communication link.
  • the CSI report includes, for example, one or more items of downlink channel matrix, precoding matrix indicator (precoding matrix indicator, PMI), rank indicator (rank indicator, RI), or channel quality indicator (channel quality indicator, CQI) and other information.
  • precoding matrix indicator precoding matrix indicator
  • rank indicator rank indicator
  • CQI channel quality indicator
  • the contents included in the CSI listed above are only illustrative descriptions, and should not constitute any limitation to the present disclosure.
  • the CSI may include one or more of the above items, and may also include other information used to characterize the CSI other than those listed above, which is not limited in the present disclosure
  • a neural network is a concrete implementation of machine learning techniques. According to the general approximation theorem, the neural network can theoretically approximate any continuous function, so that the neural network has the ability to learn any mapping.
  • Traditional communication systems need to rely on rich expert knowledge to design communication modules, while deep learning communication systems based on neural networks can automatically discover hidden pattern structures from a large number of data sets, establish mapping relationships between data, and achieve better results than traditional communication systems. The performance of the modeling method.
  • DNN deep neural network
  • MLP multi-layer perceptron
  • CNN convolutional neural networks
  • RNN recurrent neural network
  • the AE network may include an encoder (encoder) and a corresponding decoder (decoder), for example, the encoder and/or the decoder are implemented through a neural network (such as DNN).
  • the encoder may also be called an encoder network
  • the decoder may also be called a decoder network.
  • the encoder and the corresponding decoder can be jointly trained. The trained encoder and decoder can be used to encode and decode information.
  • the number of nouns means “singular noun or plural noun", that is, “one or more”, unless otherwise specified.
  • At least one means one or more
  • plural means two or more.
  • “And/or” describes the association relationship of associated objects, indicating that there can be three types of relationships, for example, A and/or B, which can mean: A exists alone, A and B exist at the same time, and B exists alone, where A, B can be singular or plural.
  • the character “/" can indicate that the associated objects are an "or” relationship.
  • A/B means: A or B.
  • the symbol “/” can also represent a division operation.
  • the symbol “ ⁇ ” can also be replaced with the symbol “*”.
  • first and second are used to distinguish multiple objects, and are not used to limit the size, content, order, timing, application scenarios, priority or importance of multiple objects, etc. .
  • first indication information and the second indication information may be the same indication information, or different indication information, and this name does not mean the size, transmission mode, or content of the two indication information. , priority, application scenarios or importance.
  • the CSI feedback mechanism adopts the flow shown in FIG. 2 .
  • the base station sends signaling, and correspondingly, the UE receives the signaling from the base station.
  • the signaling is used to configure channel measurement information, for example, the signaling notifies the UE of at least one of the following: time information for channel measurement, type of reference signal (reference signal, RS) for channel measurement, time domain resource of the reference signal, Frequency domain resources of reference signals, reporting conditions of measurement quantities, etc.
  • time information for channel measurement for example, the signaling notifies the UE of at least one of the following: time information for channel measurement, type of reference signal (reference signal, RS) for channel measurement, time domain resource of the reference signal, Frequency domain resources of reference signals, reporting conditions of measurement quantities, etc.
  • RS reference signal
  • the base station sends the reference signal to the UE, and correspondingly, the UE receives the reference signal from the base station.
  • the UE measures the reference signal to obtain the CSI.
  • the UE sends the CSI to the base station, and correspondingly, the base station receives the CSI from the UE.
  • the base station sends data to the UE according to the CSI, and correspondingly, the UE receives the data from the base station.
  • the base station determines a precoding matrix according to the CSI, and uses the precoding matrix to precode data to be sent to the UE.
  • the data sent by the base station to the UE is carried in a downlink channel, for example, carried in a PDSCH.
  • the number of antenna ports that can be supported also increases. Since the size of the complete downlink channel matrix is proportional to the number of antenna ports, in a massive MIMO system, it means a huge feedback overhead to make the CSI fed back by the UE more accurate. Due to the huge feedback overhead, the available resources for data transmission will be reduced, and thus the system capacity will be reduced. Therefore, in order to improve the system capacity, it is necessary to study how to reduce the feedback overhead of CSI. Feedback of CSI based on a dual-domain compressed codebook is an effective way to reduce feedback overhead.
  • the dual-domain compression codebook is generally designed according to factors such as the shape of the assumed antenna panel and the number of subbands.
  • factors such as the shape of the assumed antenna panel and the number of subbands.
  • the codebook determined for the fixed antenna panel shape and the number of subbands may not be able to meet the actual communication requirements.
  • the environment reduces the accuracy of the CSI fed back by the UE. Therefore, how to make the CSI fed back by the UE more accurate is a technical issue worth studying.
  • each piece of first sub-downlink channel data among the M pieces of first sub-downlink channel data obtained by the UE may correspond to one of the M data spaces, and each piece of first sub-downlink channel data may be determined according to dictionaries corresponding to different data spaces.
  • Different data spaces can represent different location information, or can represent different channel environment information, and the UE feeds back the first information corresponding to different data spaces, which can enable the access network device to clarify the relationship between the first information and the environment information.
  • Corresponding relationship so that the first information fed back by the UE can reflect the actual communication environment, improving the accuracy of the first information fed back by the UE.
  • the access network device can restore and obtain a relatively accurate downlink channel according to the first information fed back by the UE.
  • FIG. 3 shows a communication network architecture in the communication system 10 provided in the present disclosure, and any embodiment provided later may be applicable to this architecture.
  • the network equipment included in FIG. 3 is, for example, the access network equipment 20 included in the communication system 10
  • the terminal equipment included in FIG. 3 is, for example, the communication device 30 included in the communication system 10 .
  • Network devices and terminal devices can communicate.
  • the present disclosure may involve machine learning technology, which is a specific implementation of AI technology.
  • AI technology will be introduced below. It should be understood that this introduction is not intended to limit the present disclosure.
  • AI is a technology that performs complex calculations by simulating the human brain. With the increase of data storage and capacity, AI has been applied more and more.
  • an independent network element such as an AI network element, AI node, or AI device, etc.
  • the AI network element can be directly connected to the access network device, or can be indirectly connected through a third-party network element and the access network device.
  • the third-party network element may be a core network element.
  • AI entities may be configured or set in other network elements in the communication system to implement AI-related operations.
  • the AI entity may also be called an AI module, an AI unit or other names, and is mainly used to realize some or all AI functions, and the disclosure does not limit its specific name.
  • the other network element may be an access network device, a core network device, or a network management (operation, administration and maintenance, OAM), etc.
  • the network element that performs the AI function is a network element with a built-in AI function.
  • the AI function may include at least one of the following: data collection, model training (or model learning), model information release, model inference (or called model reasoning, reasoning, or prediction, etc.), model monitoring, or model verification , or release of inference results, etc.
  • AI functions may also be referred to as AI (related) operations, or AI-related functions.
  • OAM is used to operate, manage and/or maintain core network equipment (network management of core network equipment), and/or is used to operate, manage and/or maintain access network equipment (network management of access network equipment) .
  • the present disclosure includes a first OAM and a second OAM, the first OAM is the network management of the core network equipment, and the second OAM is the network management of the access network equipment.
  • the first OAM and/or the second OAM includes an AI entity.
  • the present disclosure includes a third OAM, and the third OAM is the network manager of the core network device and the access network device at the same time.
  • the AI entity is included in the third OAM.
  • FIG. 4A it is a schematic diagram of the first application framework of AI in a communication system.
  • Data source data source
  • the model training host obtains the AI model by training or updating the training data provided by the data source, and deploys the AI model in the model inference host.
  • the AI model represents the mapping relationship between the input and output of the model. Learning the AI model through the model training node is equivalent to using the training data to learn the mapping relationship between the input and output of the model.
  • the model inference node uses the AI model to perform inference based on the inference data provided by the data source, and obtains the inference result.
  • the model inference node inputs the inference data into the AI model, and obtains an output through the AI model, and the output is the inference result.
  • the inference result may indicate: configuration parameters used (executed) by the execution object, and/or operations performed by the execution object.
  • the reasoning result can be uniformly planned by the execution (actor) entity, and sent to one or more execution objects (for example, a network element of the core network, a base station, or a UE, etc.) for execution.
  • the model inference node can feed back its inference result to the model training node. This process can be called model feedback.
  • the fed back inference result is used for the model training node to update the AI model, and the updated AI model is deployed on the model Inference node.
  • the execution object can feed back the collected network parameters to the data source. This process can be called performance feedback, and the fed back network parameters can be used as training data or inference data.
  • the aforementioned AI model includes a decoder network in an AE network.
  • the decoder network is deployed on the access network device side.
  • the inference results of the decoder network are used, for example, to reconstruct the downlink channel matrix.
  • the above AI model includes the encoder network in the AE network. Wherein, the encoder network is deployed on the UE side.
  • the inference results of the encoder network are used, for example, to encode the downlink channel matrix.
  • the application framework shown in FIG. 4A can be deployed on the network elements shown in FIG. 1 .
  • the application framework in FIG. 4A may be deployed on at least one of the terminal device, access network device, core network device (not shown) or independently deployed AI network element (not shown) in FIG. 1 .
  • the AI network element (which can be regarded as a model training node) can analyze or train the training data (training data) provided by the terminal device and/or the access network device to obtain a model.
  • At least one of the terminal device, the access network device, or the core network device (which can be regarded as a model reasoning node) can use the model and reasoning data to perform reasoning and obtain the output of the model.
  • the reasoning data may be provided by the terminal device and/or the access network device.
  • the input of the model includes inference data
  • the output of the model is the inference result corresponding to the model.
  • At least one of the terminal device, the access network device, or the core network device (which can be regarded as an execution object) can perform a corresponding operation according to the reasoning result.
  • the model inference node and the execution object may be the same or different, without limitation.
  • the network architecture to which the method provided in the present disclosure can be applied is introduced as an example below with reference to FIG. 4B to FIG. 4E .
  • the access network device includes a near real-time access network intelligent controller (RAN intelligent controller, RIC) module for model training and reasoning.
  • RAN intelligent controller RIC
  • near real-time RIC can be used to train an AI model and use that AI model for inference.
  • the near real-time RIC can obtain network-side and/or terminal-side information from at least one of CU, DU, or RU, and the information can be used as training data or inference data.
  • the near real-time RIC may submit the reasoning result to at least one of CU, DU, RU or terminal device.
  • the inference results can be exchanged between the CU and the DU.
  • the reasoning results can be exchanged between the DU and the RU, for example, the near real-time RIC submits the reasoning result to the DU, and the DU forwards it to the RU.
  • a non-real-time RIC is included outside the access network (optionally, the non-real-time RIC can be located in the OAM or in the core network device) for model training and reasoning.
  • non-real-time RIC is used to train an AI model and use that model for inference.
  • the non-real-time RIC can obtain network-side and/or terminal-side information from at least one of CU, DU, or RU, which can be used as training data or inference data, and the inference results can be submitted to CU, DU, or RU or at least one of the terminal devices.
  • the inference results can be exchanged between the CU and the DU.
  • the reasoning results can be exchanged between the DU and the RU, for example, the non-real-time RIC submits the reasoning result to the DU, and the DU forwards it to the RU.
  • the access network equipment includes a near-real-time RIC, and the access network equipment includes a non-real-time RIC (optionally, the non-real-time RIC can be located in the OAM or the core network equipment middle).
  • non-real-time RIC can be used for model training and inference.
  • the near real-time RIC can be used for model training and reasoning.
  • the non-real-time RIC performs model training, and the near-real-time RIC can obtain AI model information from the non-real-time RIC, and obtain network-side and/or terminal-side information from at least one of CU, DU or RU, and use the information And the AI model information to get the reasoning result.
  • the near real-time RIC may submit the reasoning result to at least one of CU, DU, RU or terminal device.
  • the inference results can be exchanged between the CU and the DU.
  • the reasoning results can be exchanged between the DU and the RU, for example, the near real-time RIC submits the reasoning result to the DU, and the DU forwards it to the RU.
  • near real-time RIC is used to train model A and use model A for inference.
  • non-real-time RIC is used to train Model B and utilize Model B for inference.
  • the non-real-time RIC is used to train the model C, and the information of the model C is sent to the near-real-time RIC, and the near-real-time RIC uses the model C for inference.
  • FIG. 4C is an example diagram of a network architecture to which the method provided by the present disclosure can be applied. Compared with FIG. 4B , in FIG. 4B CU is separated into CU-CP and CU-UP.
  • FIG. 4D is an example diagram of a network architecture to which the method provided by the present disclosure can be applied.
  • the access network device includes one or more AI entities, and the functions of the AI entities are similar to the near real-time RIC described above.
  • the OAM includes one or more AI entities, and the functions of the AI entities are similar to the non-real-time RIC described above.
  • the core network device includes one or more AI entities, and the functions of the AI entities are similar to the above-mentioned non-real-time RIC.
  • both the OAM and the core network equipment include AI entities, the models trained by their respective AI entities are different, and/or the models used for reasoning are different.
  • the different models include at least one of the following differences: the structural parameters of the model (such as the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the at least one of the bias), the input parameters of the model (such as the type of the input parameter and/or the dimension of the input parameter), or the output parameters of the model (such as the type of the output parameter and/or the dimension of the output parameter).
  • the structural parameters of the model such as the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the at least one of the bias
  • the input parameters of the model such as the type of the input parameter and/or the dimension of the input parameter
  • the output parameters of the model such as the type of the output parameter and/or the dimension of the output parameter.
  • FIG. 4E is an example diagram of a network architecture to which the method provided in the present disclosure can be applied.
  • the access network devices in Fig. 4E are separated into CU and DU.
  • the CU may include an AI entity, and the function of the AI entity is similar to the above-mentioned near real-time RIC.
  • the DU may include an AI entity, and the function of the AI entity is similar to the above-mentioned near real-time RIC.
  • both the CU and the DU include AI entities, the models trained by their respective AI entities are different, and/or the models used for reasoning are different.
  • the CU in FIG. 4E may be further split into CU-CP and CU-UP.
  • one or more AI models may be deployed in the CU-CP.
  • one or more AI models can be deployed in CU-UP.
  • the OAM of the access network device and the OAM of the core network device are shown as unified deployment.
  • the OAM of the access network device and the OAM of the core network device may be deployed separately and independently.
  • a model can be inferred to obtain an output, and the output includes one parameter or multiple parameters.
  • the learning process or training process of different models can be deployed in different devices or nodes, or can be deployed in the same device or node.
  • Inference processes of different models can be deployed in different devices or nodes, or can be deployed in the same device or node.
  • the aforementioned AI model includes a decoder network in the AE network, and on the network side, the inference result of the decoder network is used, for example, to reconstruct the downlink channel matrix.
  • the aforementioned AI model includes an encoder network in the AE network, and model information of the encoder network can be sent to the UE for inference by the UE.
  • the AI model can be referred to simply as a model or a network model, etc., which can be regarded as the parameters from the input (such as the input matrix) to the output parameters (such as the output matrix) mapping between.
  • the input matrix may be a matrix determined according to the received CSI.
  • the training data may include a known input matrix, or a known input matrix and a corresponding output matrix, for training the AI model.
  • the training data may be data from the access network device, CU, CU-CP, CU-UP, DU, RU, UE and/or other entities, and/or data inferred by AI technology, without limitation.
  • Inference data includes input matrices for inferring output matrices using the model.
  • Inference data may be data from access network devices, CUs, CU-CPs, CU-UPs, DUs, RUs, UEs and/or other entities.
  • the inferred matrix can be regarded as policy information and sent to the execution object.
  • the deduced matrix may be sent to the access network device, CU, CU-CP, CU-UP, DU, RU, or UE, etc. for further processing, such as reconstruction of the downlink channel matrix.
  • the decoder network in the AE network can be deployed in the access network device (such as a base station) or outside the access network device, such as deployed in OAM, AI network
  • the access network device such as a base station
  • OAM optical authentication
  • AI network a network that uses OAM to determine whether the access network device is a base station.
  • the inference result of the decoder network can be obtained by inference by the access network device, or can be sent to the access network device after inference by the non-real-time RIC.
  • the present disclosure takes the decoder network deployed in the access network device as an example for description.
  • the encoder network in the AE network if the encoder network in the AE network is deployed on the terminal side, the encoder network can be deployed in the UE, and the UE can use the encoder network to perform inference.
  • FIG. 5 is a flowchart of a communication method provided by the present disclosure.
  • the UE obtains M pieces of first sub-downlink channel data, where each piece of first sub-downlink channel data corresponds to one data space in M data spaces.
  • M is an integer greater than 1.
  • the M pieces of first sub-downlink channel data are obtained, for example, according to the first downlink channel data.
  • the UE may divide the first downlink channel data into M data spaces, or it may be understood that the UE may divide the first downlink channel data into M parts, so as to obtain M parts of the first sub-downlink channel data.
  • Each piece of first sub-downlink channel data corresponds to a data space. It can also be understood that there is a one-to-one correspondence between data spaces and first sub-downlink channel data.
  • the first downlink channel data is, for example, the original downlink channel data (or called the original downlink channel matrix or downlink channel response), that is, after the UE obtains the original downlink channel data, it can directly divide it into M parts, without Perform other processing on the original downlink channel data, thereby reducing processing steps; or, the first downlink channel data may also be data obtained by preprocessing the second downlink channel data, and the second downlink channel data is obtained according to The original downlink channel matrix is obtained, and the original downlink channel data can be simplified through the preprocessing process, thereby simplifying the processing process of the UE on the first downlink channel data; or, the first downlink channel data may also be the output data of the neural network , such as the original downlink channel matrix and other content are invisible to the UE, and the UE can directly obtain the first downlink channel data output by the neural network.
  • the original downlink channel data or called the original downlink channel matrix or downlink channel response
  • the second downlink channel data is obtained according to the original downlink channel matrix, for example, the second downlink channel data is the original downlink channel matrix itself, or the second downlink channel data is obtained after processing the original downlink channel matrix Feature vector.
  • the preprocessing process may be different, which will be introduced below.
  • the second downlink channel data is the original downlink channel matrix.
  • the original downlink channel matrix is called the first downlink channel matrix.
  • the dimension of the first downlink channel matrix is [N tx , N rx , N RB ], where N tx represents the number of antennas or ports of the sending end of the downlink signal (such as the access network device), and N rx represents the number of downlink signals The number of antennas or ports of the receiving end (such as UE), and N RB represents the number of frequency domain units, such as the number of resource blocks (resource block, RB) or the number of subbands.
  • the UE may perform dimension transformation processing on the first downlink channel matrix to obtain transformed data, or in other words, obtain a transformed first downlink channel matrix.
  • the dimension of the transformed first downlink channel matrix is [N tx *N rx , N RB ] or [N tx N rx , N RB ], for example, the matrix is represented by H, and H is a complex matrix,
  • the first downlink channel data is matrix H, for example.
  • two sets of DFT bases can be generated through discrete Fourier transform (DFT), which are spatial bases and frequency-domain basis
  • DFT discrete Fourier transform
  • the spatial domain basis is N tx N rx N tx N rx *1 DFT column vectors
  • the frequency domain basis is N rb N rb *1 DFT column vectors.
  • the UE can perform space-frequency joint projection on the reduced-dimensional first downlink channel matrix H according to the space-domain base and the frequency-domain base.
  • One way of space-frequency joint projection can refer to the following formula:
  • SH is the Hermitian matrix of S, also known as a self-conjugate matrix, which can be obtained by performing conjugate transposition of the matrix S.
  • a common frequency-domain sub-band granularity is 1RB, 2RB, 4RB, or 8RB, etc., which is not limited here.
  • N sb N rb /4.
  • S represents the airspace base, and its specific shape is related to the antenna panel. Assuming that the antenna panel is dual-polarized, the horizontal element is Nh, and the vertical element is Nv, then the expression form of S is:
  • F represents the frequency domain base, and its expression is related to the subband N sb .
  • F can satisfy the following formula:
  • an oversampling factor can also be added.
  • oversampling can be used to generate multiple sets of orthogonal spatial domain bases ⁇ S 1 , S 2 , S 3 ... ⁇ and multiple sets of orthogonal frequency domain bases ⁇ F 1 , F 2 , F 3 ... ⁇ , select a group of S i and F j as the spatial domain basis and frequency domain basis of the present disclosure, for example, select a group with a more accurate projection direction.
  • the oversampling factor of both the air domain and the frequency domain is 4.
  • the first downlink channel data is, for example, a complex matrix obtained after preprocessing the second downlink channel data, for example, a complex matrix C complex .
  • the second downlink channel data is an eigenvector obtained after processing the first downlink channel matrix.
  • the process of processing the first downlink channel matrix to obtain the eigenvector and the process of preprocessing the eigenvector to obtain the first downlink channel data are all regarded as the first downlink channel matrix preprocessing process.
  • the dimension of the first downlink channel matrix is [N tx , N rx , N RB ], and the first downlink channel of dimension [N tx , N rx , N RB ] is decomposed by singular value decomposition (singular value decomposition, SVD)
  • the matrix is dimensionally reduced to obtain the characteristic subspace matrix of the downlink channel, or simply called the characteristic subspace, and the dimension of the characteristic subspace is [N tx , N sb ].
  • the UE reduces the dimensionality of the first downlink channel matrix through SVD, it can process different ranks (ranks) of the first downlink channel matrix respectively, where different ranks can also be understood as different streams, or Different layers.
  • a piece of channel information (or a channel estimation result) may correspond to one or more layers. The following describes the UE's processing process for the L-th layer of the first downlink channel matrix. There may be various methods without limitation.
  • Each subband of the L-th layer may contain a number of RBs, and the UE may synthesize downlink channels of a number of RBs to calculate an equivalent downlink channel in a subband. Assuming that the downlink channel corresponding to the kth RB in the subband c of the L-th layer is expressed as H k , then the equivalent downlink channel in the subband c can be expressed as:
  • the dimension of H k is [N tx *N rx ], The dimension of is [N tx *N tx ].
  • matrix The k-th column of the sub-band c can be used as the L-th layer feature vector corresponding to the sub-band c (in order to avoid confusion, the feature vector corresponding to the sub-band is called the sub-feature vector), and its dimension is [N tx *1], that is, the first The sub-eigenvector of the c-th subband of the L layer
  • the sub-feature vectors of the L-th layer on each sub-band can be obtained, and these sub-feature vectors are concatenated to obtain a feature vector, which can be used as input data in the present disclosure.
  • the first downlink channel data is, for example, the feature vector V, whose dimension is [N rx , N
  • the eigenvector is a complex matrix
  • two sets of DFT bases can be generated by DFT, which are the spatial bases and frequency-domain basis
  • the spatial domain basis is N tx N tx *1 DFT column vectors
  • the frequency domain basis is N sb N sb *1 DFT column vectors.
  • the UE can perform space-frequency joint projection on the dimensionality-reduced downlink channel matrix H according to the space-domain base and the frequency-domain base.
  • One way of space-frequency joint projection can refer to the following formula:
  • the obtained complex matrix C complex is a sparse representation of the characteristic subspace of the original downlink channel, and its dimension is consistent with the dimension of the characteristic vector before space-frequency joint projection, which is N tx *N sb .
  • the preprocessing process of the second downlink channel data is completed.
  • parameters such as SH , N sb , and the airspace base S please refer to the above.
  • an oversampling factor can also be added.
  • oversampling can be used to generate multiple sets of orthogonal spatial domain bases ⁇ S 1 , S 2 , S 3 ... ⁇ and multiple sets of orthogonal frequency domain bases ⁇ F 1 , F 2 , F 3 ... ⁇ , select a group of S i and F j as the spatial domain basis and frequency domain basis of the present disclosure, for example, select a group with a more accurate projection direction.
  • the oversampling factor of both the air domain and the frequency domain is 4.
  • one way for the UE to obtain the first downlink channel data according to the complex matrix C complex is that the UE directly uses the complex matrix C complex as the first downlink channel data, that is, the first downlink channel data is for the second The result of preprocessing the downlink channel data.
  • F is a positive integer.
  • the value of F can be predefined by the protocol, or different F can be determined according to different overheads.
  • the protocol can provide a mapping relationship between overhead and F, so that the UE and the access network device can Overhead requirements can determine the same F.
  • the value of F may also be indicated by the access network device, for example, the access network device sends information indicating the value of F to the UE, and the UE can determine the value of F after receiving the information.
  • the value of F may also be determined by the UE. For example, the UE determines the value of F according to factors such as channel state and/or network configuration, so as to reduce the impact on air interface transmission. After determining the value of F, the UE can send information indicating the value of F to the access network device, and the access network device can determine the value of F after receiving the information.
  • the UE may also perform compression processing on the complex matrix C complex to obtain compressed information, which may be used as the first downlink channel data.
  • the UE may input the complex matrix C complex into the encoder network, and the encoder network performs compression processing on the complex matrix C complex , and the output of the encoder network is compressed information.
  • the first downlink channel data is obtained through compression, which can reduce the complexity when the UE processes the first downlink channel data.
  • the foregoing process is to obtain the first downlink channel data.
  • the UE can divide the first downlink channel data into M data spaces, thereby obtaining M parts of the first sub-downlink channel data.
  • the first sub-downlink channel data has a one-to-one relationship with the data space, for example, the i-th part of the first sub-downlink channel data in the M parts of the first sub-downlink channel data corresponds to the i-th data in the M data spaces Space, i can take an integer from 1 to M.
  • the present disclosure involves M data spaces, and the M data spaces may correspond to dictionaries, for example, the M data spaces may correspond to N dictionaries, and N is an integer greater than or equal to 1 and less than or equal to M.
  • N M/2, where every 2 data spaces correspond to a dictionary. Other possible situations will not be cited one by one.
  • the dictionaries corresponding to different data spaces can be the same or different without limitation. The use of the dictionary will be introduced in S502 below.
  • the M data spaces (or in other words, the division method of the M data spaces) will also be involved in the process of training the dictionary, and the training process of the dictionary will be introduced in subsequent embodiments, so the division of the M data spaces Ways and the like will also be introduced in subsequent embodiments.
  • the variables saved in a dictionary include at least one of ⁇ data space index, element index, element ⁇ , that is, the variables saved in a dictionary can include data space index, element index, or one of the elements
  • other information may be included in the dictionary, or no other information may be included, and there is no limitation on this.
  • the index of the data space included in a dictionary is the index of the data space corresponding to the dictionary. For example, there is a one-to-one correspondence between dictionaries and data spaces, then a dictionary corresponds to a data space, and a dictionary includes the index of the data space corresponding to the dictionary; or, for example, M data spaces correspond to the same dictionary, then the dictionary corresponds to M data space, the dictionary may not include the data space index.
  • Elements are, for example, vectors, and a dictionary can include multiple elements. Each element may have a corresponding index, that is, there may be a one-to-one correspondence between elements and element indexes. If N is greater than 1, the indexes of elements included in different dictionaries can be reused, for example, the index of elements in each dictionary can start from 1 or from 0, that is, the elements included in different dictionaries are numbered independently; or, The indexes of the elements included in different dictionaries can also be different, that is, the elements included in different dictionaries are jointly numbered. For example, the index of the elements of the first dictionary is from 0 to d-1, and the index of the elements of the second dictionary is from d start. Refer to FIG.
  • FIG. 6 which is a schematic diagram of N dictionaries.
  • N M as an example, that is, it includes M dictionaries in total.
  • 0 to 3 in each dictionary in FIG. 6 represent the index of the element.
  • the number of indexes of the elements included in each dictionary is 4, but it is not limited to this in fact.
  • the number of elements included in different dictionaries may be the same or different.
  • the access network device can also indicate the dictionary index to the UE The corresponding relationship between the dictionary index and the index of the data space; or, if M>N>1, the UE may also report the corresponding relationship between the dictionary index and the index of the data space to the access network device.
  • the UE determines first information corresponding to the i-th piece of first sub-downlink channel data among the M pieces of first sub-downlink channel data according to the first dictionary corresponding to the i-th data space among the M data spaces.
  • i takes an integer from 1 to M, so the UE determines M pieces of first information in total.
  • the i-th part of the first sub-downlink channel data corresponds to the i-th data space in the M data spaces, for example, dividing the first downlink channel data into M data spaces will obtain M parts of the first sub-downlink channel data
  • the i-th piece of first sub-downlink channel data is the part of the first downlink channel data divided into the i-th data space.
  • N M
  • each data space has its own corresponding dictionary
  • the first dictionary is, for example, the dictionary corresponding to the i-th data space among the M data spaces, that is, the UE can
  • the first dictionary of determines the first information corresponding to the i-th piece of first sub-downlink channel data.
  • dictionaries corresponding to different data spaces can be called first dictionaries, but the first dictionaries corresponding to different data spaces may be the same or different.
  • N 1
  • a dictionary corresponds to M data spaces
  • the first dictionary is the dictionary
  • the UE can determine the first dictionary according to the first dictionary
  • the first information corresponding to i pieces of first sub-downlink channel data.
  • the first dictionary may include multiple elements, from which the UE may determine P elements corresponding to the i-th piece of first sub-downlink channel data, where P is a positive integer.
  • the P elements most relevant to the i-th sub-data are the P elements corresponding to the i-th first sub-downlink channel data, and these P elements can be used as the i-th sub-data First information corresponding to the first sub-downlink channel data.
  • the P elements can form the first information according to the first combination method, for example, the first combination method is to multiply the P elements, or the first combination method is to carry out weighted summation of the P elements ( For example, averaging, or using other possible weights to perform weighted summation), or connecting P elements in series, etc., there is no limitation on the first combination.
  • the first combination manner is, for example, predefined by a protocol, or determined by the access network device and notified to the UE, or determined by the UE and notified to the access network device. For M pieces of first sub-downlink channel data, the UE can determine its corresponding first information, and then the UE can determine M pieces of first information in total, and the M pieces of first information are M elements.
  • the UE sends first indication information.
  • the UE sends the first indication information to the access network device, and correspondingly, the access network device may receive the first indication information from the UE.
  • the first indication information may indicate M pieces of first information, and the access network device can determine the M pieces of first information according to the first indication information.
  • the first indication information includes identifiers of M pieces of first information, so as to indicate M pieces of first information.
  • An identifier of the first information is, for example, an index of the first information in the corresponding dictionary.
  • the M pieces of first information include the first information corresponding to the i-th piece of first sub-downlink channel data, and the identifier of the first information is The index of the first information in the first dictionary.
  • the UE can determine the identities of the M pieces of first information. For example, the UE can determine M identities in total, and the UE can send the M identities to the access network device.
  • the UE When the UE sends M identifiers of the first information, it can be regarded as that the UE has sent CSI, that is, the identifiers of the M first information can be used as CSI; or, the identifiers of the M first information can also be used as PMI; or, M identifiers of the first information can also be used as PMI;
  • the identification of the first information can realize a function similar to that of PMI or CSI.
  • the first indication information may not include the identifiers of the M pieces of first information, but indicate the first information in other ways.
  • each combination relationship may include one element in each of the N dictionaries.
  • Each combination relationship may correspond to one indication information, and if the UE sends certain indication information, it indicates the combination relationship corresponding to the indication information.
  • the first indication information corresponds to a combination relationship of M pieces of first information, then the UE may indicate M pieces of first information by sending the first indication information.
  • the UE when the UE sends M identifiers of the first information, it may send them in the first order, and the first order is the arrangement order of the M data spaces, that is, the first order specifies which data space the UE sends first.
  • the first sequence is 2-1-4-3
  • the UE sends M identifiers of the first information, it first sends data space 2
  • the identifier of the corresponding first information and then send the identifier of the first information corresponding to data space 1
  • the access network device can also clarify the correspondence between the identifiers of the first information and the data spaces after receiving the M identifiers of the first information , so as to avoid corresponding errors.
  • the first sequence is a sequence predefined by the protocol, and the UE and the access network device can determine the first sequence according to the protocol.
  • the first order may also be preconfigured in the UE and the access network device.
  • the first order may be determined by the access network device, and the access network device may send second indication information to the UE after determining the first order, the second indication information is used to indicate the first order, and the UE can determine the order according to the second indication information first order.
  • the first order may be determined by the UE.
  • the UE may send third indication information to the access network device. The third indication information is used to indicate the first order, and the access network device can determine the first order according to the third indication information.
  • first order is a sequence predefined by the protocol, and the UE and the access network device can determine the first sequence according to the protocol.
  • the first order may also be preconfigured in the UE and the access network device.
  • the first order may be determined by the access network device, and the access network device may send second indication information to
  • the access network device restores and obtains the i-th second sub-downlink channel data according to the first dictionary corresponding to the i-th data space among the M data spaces .
  • i takes an integer from 1 to M, then the access network device can obtain M pieces of second sub-downlink channel data in total.
  • the access network device can clarify the correspondence between the identifiers of the first information and the data space, so that the access network device can
  • the corresponding dictionary determines the first information corresponding to the identifier of the first information, and the first information determined by the access network device is regarded as the second sub-downlink channel data recovered by the access network device.
  • N M
  • data spaces correspond to dictionaries one-to-one
  • the dictionary corresponding to the i-th data space is, for example, the first dictionary
  • the access network device can determine the first dictionary in the first dictionary The identification of the i first information, thereby determining the first information corresponding to the identification of the i first information in the first dictionary, that is, recovering the second sub-downlink channel data corresponding to the i first information (that is, i-th second sub-downlink channel data).
  • the access network device may determine the identifier of the i-th first information in the first dictionary, thereby determining The identifier of the i-th first information corresponds to the first information in the first dictionary, that is, recovers the second sub-downlink channel data corresponding to the i-th first information (ie, the i-th second sub-downlink channel data) .
  • M>N>1 then for the i-th first information identifier, the access network device can specify the data space corresponding to the i-th first information identifier according to the first order, for example, the i-th data space .
  • the access network device may further determine a dictionary corresponding to the i-th data space, such as the first dictionary, according to the correspondence between the index of the data space and the dictionary index. Then the access network device can determine the first information corresponding to the identifier of the i-th first information in the first dictionary, that is, recover the second sub-downlink channel data corresponding to the i-th first information (that is, the i-th second sub-downlink channel data).
  • the M pieces of second sub-downlink channel data obtained by the access network device and the M pieces of first sub-downlink channel data obtained by the UE may be the same data.
  • the i-th piece of first sub-downlink channel data and the i-th piece of second sub-downlink channel data are the same data.
  • the process for the UE to obtain the first information according to the dictionary is equivalent to quantizing M pieces of first sub-downlink channel data, that is, what the UE sends to the access network device is quantized information, and the access network device recovers the
  • the M pieces of second sub-downlink channel data may suffer some losses during the process of quantization and recovery, which may result in a certain deviation between the M pieces of second sub-downlink channel data and the M pieces of first sub-downlink channel data.
  • the i-th piece of first sub-downlink channel data and the i-th piece of second sub-downlink channel data may be different data.
  • the deviation between the M parts of the second sub-downlink channel data and the M parts of the first sub-downlink channel data tends to decrease.
  • the access network device restores and obtains downlink channel information according to the M shares of second sub-downlink channel data.
  • the access network device reconstructs the downlink channel matrix according to the M shares of second sub-downlink channel data, for example, reconstructs the first downlink channel matrix.
  • the access network device obtains M After the second sub-downlink channel data, M parts of the second sub-downlink channel data can be spliced, and the obtained information is called the angle delay domain coefficient, for example, and the angle delay domain coefficient is a matrix, which can be expressed as
  • the access network device can restore the M pieces of sub-compressed information to obtain K pieces of restored information, where K is a positive integer, and K may or may not be equal to M.
  • the access network device side can set up a decoder network corresponding to the encoder network, and the access network device can input M copies of the second sub-downlink channel data into the decoder network, the decoder network can output K recovery information.
  • the access network equipment can splicing K pieces of recovery information to obtain a matrix of angular delay domain coefficients, which can be expressed as
  • FIG. 7 is a schematic diagram of the UE using the compressed information as the first downlink channel data and the access network device needing to restore the compressed information.
  • the UE inputs the second downlink channel data into the encoder network, the encoder network compresses the second downlink channel data, and the encoder network outputs compressed information, which can be used as the first downlink channel data.
  • the UE may also input the complex matrix C complex into the encoder network, the encoder network performs compression on the complex matrix C complex , and the encoder network outputs compressed information, which may be used as the first downlink channel data.
  • the UE obtains 4 pieces of first sub-downlink channel data.
  • the UE processes 4 copies of the first sub-downlink channel data through 4 dictionaries, and obtains 4 identifiers of the first information.
  • the circles in Figure 7 represent dictionaries, C M represents the number of elements included in the Mth dictionary, log 2 C M Indicates the number of transmission bits corresponding to the Mth dictionary.
  • the number of transmission bits corresponding to the dictionary can be obtained by rounding up, for example, log 2 C M in Figure 7 can also be replaced by log 2 C 1 can also be replaced by wait.
  • the number of transmission bits corresponding to the dictionary can also be obtained by rounding down, which is not specifically limited.
  • a piece of first information is an element corresponding to a piece of first sub-downlink channel data in a corresponding dictionary.
  • the UE sends 4 identifiers of the first information to the access network device, and after receiving the 4 identifiers of the first information, the access network device can restore 4 sub-compressed information according to the 4 dictionaries, and the access network device converts the 4 sub-compressed information Perform splicing and other processing, and then input the obtained information into the decoder network to obtain recovery information, and the access network equipment can recover downlink channel information according to the recovery information.
  • the encoder network may need to use the codebook when compressing the complex matrix C complex , and correspondingly, the decoder network may also need to use the codebook when restoring the compressed information.
  • the codebook may also be called a dictionary, but the dictionary is different from the N dictionaries described in this disclosure.
  • the access network device can use the angle delay domain coefficient Recover the downlink channel information.
  • the restored downlink channel information and the first downlink channel matrix may be the same information.
  • Access network equipment will Perform inverse transformation to obtain the recovered downlink channel (or in other words, the reconstructed downlink channel). For example, access network equipment will One way to do the inverse transform is as follows:
  • the dimension of is N tx N rx *N RB .
  • the It can be directly used as the recovered downlink channel information, or it can also be converted into The dimension of is converted to the same dimension as that of the first downlink channel matrix, and the recovered downlink channel information is obtained after the dimension conversion.
  • Access network equipment will The inverse transformation is performed to obtain the characteristic subspace of the restored downlink channel. For example, access network equipment will One way to do the inverse transform is as follows:
  • the UE may divide the first downlink channel data into M data spaces, and determine the first information corresponding to each piece of first sub-downlink channel data according to dictionaries corresponding to different data spaces. Different data spaces Different location information can be represented, or different channel environment information can be represented, and the UE feeds back the first information corresponding to different data spaces, which can enable the access network device to clarify the correspondence between the first information and the environment information, thereby
  • the first information fed back by the UE can reflect the actual communication environment, and the accuracy of the first information fed back by the UE is improved.
  • the access network device can restore and obtain a relatively accurate downlink channel according to the first information fed back by the UE.
  • the embodiment shown in FIG. 5 introduces the process of network reasoning.
  • the dictionary is involved in the process of network reasoning, and the dictionary can be obtained through network training.
  • there may be multiple ways for training to obtain a dictionary For example, if the UE side does not set the encoder network, the access network device side does not set the decoder network, or, even if the UE side sets the encoder network, the access network device side sets the decoder network, the encoder network and the decoder network
  • the network may or may not be co-trained with dictionaries. If you only need to obtain a dictionary through training, but do not need to obtain a codec network, you can refer to another communication method of the present disclosure described next, which introduces a network training process through which a dictionary can be obtained. Please refer to FIG. 8 , which is a flowchart of the method.
  • the first node obtains M shares of third sub-downlink channel data.
  • each third sub-downlink channel data corresponds to one data space in the M data spaces.
  • the M data spaces in the present disclosure may have the same characteristics as the M data spaces described in the embodiment shown in FIG. 5 .
  • the M pieces of third sub-downlink channel data are obtained, for example, according to the third downlink channel data.
  • the UE or the first node may divide the third downlink channel data into M data spaces, or it may be understood that the UE or the first node may divide the third downlink channel data into M parts, so as to obtain M parts of the third downlink channel channel data.
  • the third downlink channel data is, for example, the original downlink channel data, for example, the original downlink channel data in this embodiment is called the third downlink channel matrix; or, the third downlink channel data may also be the fourth downlink channel data.
  • the fourth downlink channel data is obtained according to the third downlink channel matrix; or, the third downlink channel data may also be the data output by the neural network.
  • the third downlink channel matrix may be regarded as training data, or referred to as training samples.
  • the third downlink channel matrix includes one or more training data, for example, the third downlink channel matrix actually includes one or more sub-downlink channel matrices, and one of the sub-downlink channel matrices may be regarded as a piece of training data.
  • the third sub-downlink channel matrix here may be independent of each other, and is not included in a large matrix, that is, the third downlink channel matrix is not regarded as a large matrix, and the third downlink channel matrix can be understood as A general term for one or more third sub-downlink channel matrices.
  • the third downlink channel data is obtained by preprocessing the fourth downlink channel data
  • a preprocessing process is involved.
  • preprocessing process of the fourth downlink channel data reference may be made to the introduction of the preprocessing process of the second downlink channel data in S501 of the embodiment shown in FIG. 5 .
  • the first node is, for example, a UE, or an access network device, or may also be a third-party device (such as an AI node, etc.).
  • the training process can be online training or offline training.
  • the first node may divide the third downlink channel data into M data spaces to obtain M pieces of third sub-downlink channel data.
  • the third sub-downlink channel data has a one-to-one relationship with the data space, for example, the i-th part of the third sub-downlink channel data in the M parts of the third sub-downlink channel data corresponds to the i-th data in the M data spaces Space, i can take an integer from 1 to M.
  • the first node To divide the third downlink channel data into M data spaces, the first node first needs to determine the M data spaces, or determine the division method of the M data spaces. Take for example that the first node is a UE or an access network device.
  • the division method of the M data spaces is predefined by the protocol, and both the UE and the access network device can determine the division method of the M data spaces according to the protocol.
  • the division method of the M data spaces is determined by the access network device, and the access network device may send fourth indication information to the UE.
  • the fourth indication information may indicate the division mode of the M data spaces, and the UE determines the division mode of the M data spaces according to the fourth indication information.
  • a division manner of the M data spaces can be determined.
  • the division method of the M data spaces may be determined by the UE, and the UE may send fifth indication information to the access network device, where the fifth indication information may indicate the division mode of the M data spaces, and the access network device may, according to the fifth indication information, A division manner of the M data spaces is determined.
  • the divided 4 third sub-downlink channel data respectively include the polarization 1+real part included in the third downlink channel data, the third downlink The part of polarization 1+imaginary part included in the channel data, the part of polarization 2+real part included in the third downlink channel data, and the part of polarization 2+imaginary part included in the third downlink channel data.
  • the antenna element is dual-polarized, and polarization 1 and polarization 2 represent two polarization directions, which can be considered independent of each other; from the nature of complex numbers, the data includes real Part and imaginary part, the processing process of real part and imaginary part is also relatively independent.
  • the data space can be divided according to the antenna polarization direction and the real and imaginary parts of the complex number, so that each data space can be processed independently.
  • the size of each data space is 1/M of the original data, and different data spaces can represent different environmental information.
  • the manner of dividing the data space may also be a manner of unequal division, which is not limited.
  • the UE and the access network device may also determine the division method of the M data spaces, the determination method is similar to the present disclosure, or the first node may also determine the division method of the M data spaces Indicate to UE and/or access network equipment. That is, in the embodiment shown in FIG. 5 , if the UE wants to divide the first downlink channel data into M data spaces, it also needs to first determine the division method of the M data spaces, and then the M data spaces provided by this disclosure can be used How the data space is divided. Wherein, in the network reasoning phase and the network training phase, the M data spaces used are divided in the same way.
  • the first node performs clustering (clustering) training to obtain N dictionaries.
  • N may be equal to M, may also be equal to 1, and may also be M>N>1.
  • the training process of the first node may be different, and they will be introduced separately below .
  • N M, that is, there is a one-to-one correspondence between the data space and the dictionary.
  • the first node can perform clustering training according to the ith part of the third sub-downlink channel data in the M parts of the third sub-downlink channel data, so as to obtain a dictionary (such as the first dictionary) corresponding to the i-th data space, and the i-th part
  • the third sub-downlink channel data corresponds to the i-th data space. That is to say, the first node can perform training in each data space separately, so as to obtain dictionaries corresponding to each data space, and M dictionaries in total can be obtained.
  • Clustering is to divide a data set into different classes or clusters according to a certain standard (such as distance), so that the similarity of data objects in the same cluster is as large as possible, and the data objects that are not in the same cluster The variance is also as large as possible. That is, after clustering, the data of the same class can be gathered together as much as possible, and the data of different classes can be separated as much as possible.
  • Each class of data has a class center value. If the present disclosure adopts a clustering method to train the network model, the elements included in the dictionary may also be called cluster center values.
  • Training in a data space is to obtain elements corresponding to the data space, and these elements can be used as elements included in the dictionary corresponding to the data space.
  • the number of elements included in the dictionary corresponding to a data space may be related to the bit overhead corresponding to the data space.
  • one type of bit overhead is 48 bits (bit), which is the total transmission overhead corresponding to M data spaces.
  • bit overhead corresponding to each data space is equal, assuming that there are 4 data spaces in total, and the transmission overhead corresponding to each data space is 1/M of the total bit overhead, then the bit overhead corresponding to one data space is 12 bits.
  • 12bits can carry 2 12 identifiers at most, so the number of elements included in the first dictionary needs to be less than or equal to 2 12 . It can be seen that the number of elements corresponding to different bit overheads is different.
  • the first node may respectively train different dictionaries according to different bit overheads, and the bit overheads may correspond to the dictionaries one by one. That is to say, the first node can train one or more dictionaries for one data space. If multiple dictionaries are trained, the multiple dictionaries can correspond to different bit overheads, so that the UE can choose according to the current bit overhead when performing network inference. proper dictionary.
  • the bit overhead corresponding to the data space is generally determined by the network side, for example, by the access network device, which may be determined according to the real-time channel conditions with the UE. Then, when the UE is performing the network inference process described in the embodiment shown in FIG. 5, the access network device may first send information to the UE to determine the bit overhead of this transmission, and the information may indicate the M data spaces corresponding to The total bit overhead, or the bit overhead corresponding to a data space can also be determined, and the UE can select an appropriate dictionary according to the bit overhead indicated by the access network device.
  • the first node may also determine the dimensions of the elements included in the first dictionary according to the dimensions of the ith third sub-downlink channel data.
  • the dimensions of the elements included in the first dictionary can also be regarded as the depth of the first dictionary, which is related to the dimensions of the sub-downlink channel data used to train the first dictionary. Therefore, the first node can determine the dimensions of the elements included in the first dictionary according to the dimensions of the i-th third sub-downlink channel data. For example, the first node may convert the i-th piece of third sub-downlink channel data into a vector, and the length of the vector is the dimension of the elements included in the first dictionary.
  • the i-th part of the third sub-downlink channel data is the part of the polarization 1+real part included in the third downlink channel data.
  • This part is, for example, a matrix with a dimension of [16,13].
  • the first node can use the matrix Converted to a vector with a length of 16*13, then the dimension of the elements included in the first dictionary is 16*13.
  • the first node needs to convert a matrix into a vector, which can be converted by row or by column.
  • the UE also needs to perform the conversion process.
  • the switching sequence of the UE needs to be known by both the UE and the access network equipment.
  • the switching sequence of the UE may be predefined by a protocol, or the switching sequence of the UE may be determined by the access network device and notified to the UE, or the switching sequence of the UE may be determined by the UE and notified to the access network device.
  • the conversion sequence in the network reasoning process can be consistent with the conversion sequence in the network training process.
  • the first node After determining the number of elements included in the first dictionary and the dimensions of the elements included in the first dictionary, the first node can perform clustering training according to the i-th piece of third sub-downlink channel data, thereby obtaining the first dictionary. For each data space, the first node can be trained in a similar manner, so that M dictionaries can be obtained.
  • N 1, that is, M data spaces correspond to one dictionary, for example, the dictionary is called the first dictionary.
  • the first node may perform clustering training according to the M shares of third sub-downlink channel data, so as to obtain dictionaries corresponding to M data spaces (for example, the first dictionary). That is to say, after the first node obtains M shares of third sub-downlink channel data, it can perform unified training to obtain a dictionary, and the dictionary corresponds to each data space in the M data spaces. Utilize M parts of the third sub-downlink channel data to train a dictionary. For this dictionary, the sample data (or training data) is equivalent to an increase of M times. The training data is more abundant, making the elements included in the dictionary more abundant. Meticulous, it is beneficial for the access network equipment to recover and obtain more accurate downlink channel information.
  • the purpose of using M pieces of third sub-downlink channel data for training is to obtain elements corresponding to M data spaces, and these elements can be used as elements included in a dictionary obtained through training.
  • the first node may respectively train different dictionaries according to different bit overheads, and the bit overheads may correspond to the dictionaries one by one. That is to say, the first node can train one or more dictionaries, and if multiple dictionaries are trained, the multiple dictionaries can correspond to different bit overheads, so that the UE can select an appropriate dictionary according to the current bit overhead when performing network inference.
  • the first node may also determine the dimensions of the elements included in the first dictionary according to the dimensions of the third downlink channel data.
  • the first dictionary is, for example, one of the dictionaries obtained through training of the first node.
  • the first node After determining the number of elements included in the first dictionary and the dimensions of the elements included in the first dictionary, the first node can perform clustering training according to the M pieces of third sub-downlink channel data, so as to obtain the first dictionary.
  • the first node may perform cluster training according to at least one piece of third sub-downlink channel data among the M pieces of third sub-downlink channel data, so as to obtain one dictionary (for example, the first dictionary) among the N dictionaries.
  • the third sub-downlink channel data has a one-to-one correspondence with data spaces, and at least one piece of third sub-downlink channel data corresponds to at least one data space.
  • a dictionary can correspond to one or more data spaces, for example, there is a corresponding relationship between the index of the dictionary and the index of the data space, then when the first node trains a dictionary, it uses the third sub-downlink channel in the data space corresponding to the dictionary data for training.
  • the first node may respectively train different dictionaries according to different bit overheads, and the bit overheads may correspond to the dictionaries one by one. That is to say, one dictionary corresponds to one or more data spaces, then for this one or more data spaces, the first node can train one or more dictionaries according to different bit overheads, if multiple dictionaries are trained, then more Each dictionary can correspond to different bit overheads, so that the UE can select an appropriate dictionary according to the current bit overheads when performing network inference.
  • the first node may also determine the dimensions of the elements included in the first dictionary according to the dimensions of the third downlink channel data.
  • the first dictionary is, for example, one of the dictionaries obtained through training of the first node.
  • the first node may perform cluster training according to at least one piece of third sub-downlink channel data, so as to obtain the first dictionary.
  • a loss function can be defined, which describes the gap or difference between the output value of the neural network and the ideal target value.
  • a loss function may not be used during cluster training, or a loss function may be used.
  • one kind of loss function is to take the minimum value among the distances between multiple training samples and the cluster center as the target, or take the most relevant training sample among the multiple training samples and the cluster center as the target.
  • the loss function may also be other functions, and the present disclosure does not limit the implementation manner of the loss function.
  • the training process of the dictionary is the process of adjusting the parameters of the dictionary so that the value of the loss function is less than the threshold, or the value of the loss function meets the target requirements. Adjusting the parameters of the dictionary includes, for example, adjusting the elements of the dictionary.
  • the first node is trained to obtain N dictionaries, so that the UE can use N dictionaries when performing the network reasoning process described in the embodiment shown in Figure 5, and the access network device is recovering downlink channel information It is also possible to use N dictionaries.
  • N dictionaries By dividing the data space and N dictionaries, the environment information corresponding to the downlink channel can be reflected, which helps the access network equipment recover more accurate downlink channel information.
  • FIG. 9A is a schematic diagram of a training process and a network reasoning process provided in the present disclosure.
  • the UE executes the training process as an example.
  • the whole process in Figure 9A can be regarded as a network reasoning process, and the network reasoning process can also It can be considered as the processing process of a data, of course, the data is not actually the training data used for training, but the processing process of the data is similar to the training data.
  • the training process includes obtaining a dictionary by performing cluster training on multiple training data, and the processing process can be regarded as representing the downlink channel data through the obtained dictionary.
  • the training data is original downlink channel data
  • the original downlink channel data actually includes a plurality of training data (or referred to as training samples).
  • the UE preprocesses the feature vector to obtain sparse coefficients of the feature vector, and the sparse coefficients of feature vectors corresponding to multiple training data can be used as third downlink channel data. Since the network training uses real numbers, the data input is divided into two parts, the real part and the imaginary part.
  • the dimensions of the third downlink channel data are, for example, [E, 2, 32, 13].
  • E in the dimension of the third downlink channel data is regarded as the quantity of training data, that is, at this time, the third downlink channel data may be regarded as including E training data, and E is a positive integer.
  • "2" in the dimension of the third downlink channel data represents a real part and an imaginary part, "32” represents N tx , and "13" represents N sb .
  • the four pieces of third sub-downlink channel data are respectively y 1 , y 2 , y 3 , and y 4 .
  • the dimensions of y 1 , y 2 , y 3 , and y 4 are all [S,16*13], and S represents the number of training data corresponding to the third sub-downlink channel data, and 16*13 is, for example, the training data to be trained Dimensions of the dictionary.
  • UE trains these 4 dictionaries in a clustering manner.
  • the information of these four dictionaries can be specified in the protocol, or sent by the UE to the access network device. If it is an online training method, the information of the four dictionaries can be sent by the UE to the access network device.
  • the UE can obtain four pieces of first information based on the four training dictionaries and four pieces of first sub-downlink channel data, and one piece of first information is a piece of first sub-downlink channel data in the corresponding The corresponding element in the dictionary.
  • the UE sends four identifiers of the first information to the access network device, where each identifier of the first information may occupy X bits.
  • the access network device can recover the four first messages according to the four dictionaries, and then the access network device can process the four first messages by splicing and other processing to recover the downlink channel information. In other words, the downlink channel matrix can be reconstructed.
  • y 1 may be regarded as including 4 pieces of third sub-downlink channel data.
  • the dimension of y 1 is [4*S, 16*13], where S represents the number of training data corresponding to the third sub-downlink channel data, and the number of training data corresponding to the 4 third sub-downlink channel data is 4*S, in addition, 16*13 is, for example, the dimension of the dictionary to be trained.
  • the UE trains the dictionary in a clustering manner.
  • the UE can independently find the corresponding first information in the corresponding dictionary. If according to FIG. 9B , the M data spaces uniformly correspond to a dictionary, then during the network inference process, the UE can find the corresponding first information for each data space in the dictionary trained in FIG. 9B .
  • the UE may determine four pieces of first information. For example, the UE sends the identifiers of the four first pieces of information to the access network device. After receiving the identifiers of the four first pieces of information, the access network device can restore the four first pieces of information according to the four dictionaries. The four pieces of first information are processed by splicing, etc., to recover the downlink channel information, or in other words, to reconstruct the downlink channel matrix.
  • the third downlink channel data is the preprocessing result of the original downlink channel data (or feature vector).
  • the third downlink channel data may also be continuous F column data extracted from the preprocessing result. If this is the case, the dimensions of the elements included in the dictionary may vary.
  • the i-th part of the third sub-downlink channel data is one of M parts of the F-column data taken from the preprocessing result
  • the i-th part of the third sub-downlink channel data is, for example, the dimension is [16, F] matrix
  • the UE can convert the matrix into a vector with a length of 16*F
  • the dimension of the elements included in the first dictionary is 16*F.
  • F is generally smaller than the number of subbands, which can reduce the storage space occupied by the dictionary.
  • E represents the number of training data, and E is a positive integer.
  • the continuous F-column data is taken out from the third downlink channel data, and the continuous F-column data is divided into M data spaces to obtain M pieces of third sub-downlink channel data.
  • M 4 as an example
  • the four pieces of third sub-downlink channel data obtained by division are respectively y 1 , y 2 , y 3 , and y 4 .
  • the dimensions of y 1 , y 2 , y 3 , and y 4 are all [S,16*F].
  • UE trains these 4 dictionaries in a clustering manner.
  • the UE can independently find the corresponding first information in the corresponding dictionary.
  • the UE may determine 4 pieces of first information according to the 4 dictionaries trained in FIG. 9C .
  • the UE sends the identifiers of the four first pieces of information to the access network device.
  • the access network device can restore the four first pieces of information according to the four dictionaries.
  • the four pieces of first information are processed by splicing, etc., to recover the downlink channel information, or in other words, to reconstruct the downlink channel matrix.
  • the continuous F-column data is taken out from the third downlink channel data, and the continuous F-column data is divided into M data spaces to obtain M pieces of third sub-downlink channel data.
  • the four pieces of third sub-downlink channel data can be collectively expressed as y 1 , where the dimension of y 1 is [4*S, 16*13].
  • the UE trains the dictionary in a clustering manner.
  • the UE can independently find the corresponding first information in the corresponding dictionary. If according to FIG. 9D , the M data spaces uniformly correspond to a dictionary, then during the network reasoning process, the UE can find the corresponding first information for each data space in the dictionary trained in FIG. 9D .
  • the UE may determine four pieces of first information. For example, the UE sends the identifiers of the four first pieces of information to the access network device. After receiving the identifiers of the four first pieces of information, the access network device can restore the four first pieces of information according to the four dictionaries. The four pieces of first information are processed by splicing, etc., to recover the downlink channel information, or in other words, to reconstruct the downlink channel matrix.
  • the network training process introduced in the embodiment shown in Figure 8 is the process of training to obtain a dictionary, and it is also introduced in the previous section that an encoder network may be set on the UE side, and a decoder corresponding to the encoder network may be set on the access network device side network, another network training process is the process of jointly training the encoder network, decoder network, and dictionary.
  • another communication method of the present disclosure is introduced, through which a joint training process is introduced. Please refer to FIG. 10 , which is a flowchart of the method.
  • the second node obtains fifth downlink channel data.
  • the fifth downlink channel data is, for example, original downlink channel data; or, the fifth downlink channel data may also be data obtained by preprocessing the original downlink channel data; or, the fifth downlink channel data may also be neural network output The data.
  • the original downlink channel data may be regarded as training data, or referred to as training samples. In the process of training the dictionary, the second node needs to train the training samples.
  • the original downlink channel data may include one or more training data.
  • the fifth downlink channel data is obtained by preprocessing the original downlink channel data, a preprocessing process is involved.
  • preprocessing process of the original downlink channel data reference may be made to the introduction of the preprocessing process of the second downlink channel data in S501 of the embodiment shown in FIG. 5 .
  • the second node is, for example, a UE, or an access network device, or may also be a third-party device (such as an AI node, etc.).
  • the training process can be online training or offline training.
  • the second node and the first node described in the embodiment shown in FIG. 8 may be the same node, or may be different nodes.
  • the second node may use the fifth downlink channel data to jointly train the encoder network, the dictionary, and the decoder network.
  • the training process is introduced below through S1002-S1006.
  • the second node inputs the fifth downlink channel data into the encoder network, and obtains sixth downlink channel data output by the encoder network.
  • the encoder network is an encoder network that needs to be trained, and the second node inputs the fifth downlink channel data into the encoder network, and the encoder network can perform compression and other processing on the fifth downlink channel data.
  • the encoder network will output sixth downlink channel data after processing.
  • the second node obtains M shares of sixth sub-downlink channel data.
  • each piece of sixth sub-downlink channel data corresponds to one data space in the M data spaces.
  • the M data spaces in the present disclosure may have the same characteristics as the M data spaces described in the embodiment shown in FIG. 5 .
  • the M pieces of sixth sub-downlink channel data are obtained according to the sixth downlink channel data.
  • the sixth downlink channel data is divided into M data spaces to obtain M pieces of sixth sub-downlink channel data.
  • S1001 refer to S801 in the embodiment shown in FIG. 8 .
  • the second node obtains M pieces of third information according to M pieces of sixth sub-downlink channel data and N dictionaries to be trained.
  • the second node trains the dictionary to be trained according to the i-th data space in the M data spaces, and i takes an integer from 1 to M, then the second node can train M data spaces to be trained dictionary.
  • the second node trains the dictionary to be trained corresponding to the data space according to the i-th data space in the M data spaces.
  • Six sub-downlink channel data the second node obtains the third information corresponding to the i-th sixth sub-downlink channel data according to the dictionary to be trained corresponding to the i-th data space, then the second node can obtain M pieces of third information in total .
  • the third information corresponding to the i-th piece of sixth sub-downlink channel data is an element corresponding to the i-th piece of sixth sub-downlink channel data in the dictionary to be trained corresponding to the i-th data space.
  • an initial model can be set as a dictionary to be trained, and multiple rounds of training are performed on the initial model through the original downlink channel data (wherein, the process of using one training data for training can be regarded as a round of training process ), the dictionary used in the network inference phase can be obtained after the training is completed. Therefore, the dictionary to be trained corresponding to the above i-th data space may be an initial model, or an intermediate model obtained after at least one round of training on the initial model.
  • the second node can also train the dictionaries to be trained according to M data spaces, and then the second node can train M identical dictionaries or 1 dictionary.
  • the second node trains the dictionary to be trained according to the M data spaces.
  • a training method is, for the i-th sixth sub-downlink channel data in the M sixth sub-downlink channel data, the second node according to the to-be-trained dictionary to obtain the third information corresponding to the i-th piece of sixth sub-downlink channel data, then the second node can obtain M pieces of third information in total.
  • the third information corresponding to the i-th piece of sixth sub-downlink channel data is an element corresponding to the i-th piece of sixth sub-downlink channel data in the dictionary to be trained.
  • the second node restores and obtains the i-th fifth sub-downlink channel data according to the dictionary to be trained corresponding to the i-th data space among the M data spaces .
  • i takes an integer from 1 to M, then the second node can obtain M pieces of fifth sub-downlink channel data in total.
  • the M pieces of fifth sub-downlink channel data obtained by the second node and the M pieces of sixth sub-downlink channel data obtained by the second node may be the same data.
  • the i-th piece of sixth sub-downlink channel data and the i-th piece of fifth sub-downlink channel data are the same data.
  • the second node inputs M pieces of fifth sub-downlink channel data into the decoder network, and obtains L recovery information output by the decoder network, where L is a positive integer.
  • the second node splices the M pieces of fifth sub-downlink channel data, and inputs the spliced sub-downlink channel data into the decoder network to obtain the first recovery information output by the decoder network.
  • the decoder network is a decoder network that needs to be trained, and is also a decoder network corresponding to the encoder network in S1002.
  • the original downlink channel data includes a plurality of training data
  • one of the training data may include sub-training data and labels.
  • the second node can input the sub-training data into the encoder network, and after passing through the decoder network, the decoder network will output an inference result (such as the L recovery information or the first recovery information described in this disclosure).
  • the second node can calculate the error between the inference result and the label according to the loss function. Based on the error, the second node can use the backpropagation optimization algorithm (or model optimization algorithm, etc.) to optimize the encoder network and/or decoding parameters of the server network.
  • a large amount of training data is used to train the encoder network and the decoder network, so that the difference between the output of the decoder network and the label is less than the preset value, and then the training of the neural network is completed.
  • the training process of the encoder network and the decoder network described above adopts the training method of supervised learning, that is, based on the training data and labels, the training of the encoder network and the decoder network is realized by using the loss function.
  • the training process of the intelligent model can also use unsupervised learning, using the algorithm to learn the internal mode of the training data, and realize the training of the intelligent model based on the training data.
  • the training process of the intelligent model can also use reinforcement learning to obtain the excitation signal of environmental feedback through interaction with the environment, so as to learn the strategy to solve the problem and realize the optimization of the model.
  • the disclosure does not limit the training method of the model, the type of the model, and the like.
  • the second node trains the encoder network and the decoder network, it can perform training according to a certain loss function.
  • the same loss function can be set for the M data spaces, that is, for any one of the M data spaces, the second node can perform joint training according to the loss function.
  • the output of the decoder network is L pieces of recovery information
  • the mean square error (mean square error, MSE) between the data obtained by splicing the L pieces of recovery information recovered by the decoder network and the fifth downlink channel data can be used as The loss function, or the correlation between the data obtained by splicing the L recovery information recovered by the decoder network and the third downlink channel data is used as the loss function; or, the output of the decoder network is the first recovery information, the MSE between the first recovery information and the fifth downlink channel data may be used as the loss function, or the correlation between the first recovery information and the third downlink channel data may be used as the loss function, etc.
  • the output of the decoder network is L recovery information
  • the MSE between the recovery information recovered by the decoder network and the data input to the encoder network can be used as a loss function corresponding to the data space.
  • the recovery information corresponding to the loss function obtained through the recovery of the decoder network is the recovery information corresponding to the data space obtained through the recovery of the decoder network
  • the input data of the encoder network corresponding to the loss function refers to Data corresponding to the data space in the fifth downlink channel data input to the encoder network.
  • the second node jointly trains the encoder network, the decoder network, and the dictionary, and can obtain N dictionaries, as well as the encoder network and the corresponding decoder network, so that the UE is performing the process shown in Figure 5.
  • N dictionaries and encoder networks may be used in the network reasoning process described in the embodiment, and the access network device may also use N dictionaries and decoder networks when restoring downlink channel information. By dividing the data space and N dictionaries, the environment information corresponding to the downlink channel can be reflected, which helps the access network equipment recover more accurate downlink channel information.
  • the network training method provided by the embodiment shown in Figure 8 can be used to train the dictionary separately; if the embodiment shown in Figure 5 needs to use the encoding If the encoder network and decoder network are used, the network training method provided by the embodiment shown in FIG. 10 can be used to jointly train the encoder network, decoder network and dictionary.
  • FIG. 11 is a schematic diagram of the training process and network reasoning process provided by the present disclosure.
  • the training process such as joint training, obtains the encoder network, decoder network and dictionary.
  • the network reasoning process will use the encoder network , decoder network and dictionary.
  • From the third downlink channel data to q 1 ⁇ q 4 that is, before sending information to the access network equipment, it can be regarded as a training process; and the whole process in Figure 11 can be regarded as a network reasoning process, and the network reasoning process can also be regarded as a training process.
  • It can be regarded as the processing process of a training data.
  • the data is not actually the training data used for training, but the processing process of the data is consistent with the training data.
  • the original downlink channel data may include multiple training data.
  • the UE preprocesses the feature vector to obtain sparse coefficients of the feature vector.
  • the UE uses the encoder network to compress the sparse coefficients of the feature vector to obtain compressed information.
  • the compressed information corresponding to the multiple training data can be used as the third downlink channel data.
  • the UE divides the third downlink channel data into 4 data spaces to obtain 4 pieces of third sub-downlink channel data, and the 4 pieces of third sub-downlink channel data are respectively y 1 , y 2 , y 3 , and y 4 .
  • the dimensions of y 1 , y 2 , y 3 , and y 4 are all [S,16*13]
  • S represents the number of training data corresponding to the third sub-downlink channel data
  • 16*13 is, for example, the dictionary to be trained dimension.
  • UE trains these 4 dictionaries in a clustering manner.
  • the UE can obtain 4 pieces of first information according to the 4 dictionaries to be trained and the 4 pieces of first sub-downlink channel data.
  • One piece of first information is an element corresponding to a piece of first sub-downlink channel data in a corresponding dictionary.
  • the UE sends four identifiers of the first information to the access network device, where each identifier of the first information may occupy X bits.
  • the access network device After the access network device receives the 4 first information identifiers, it can recover 4 sub-compressed information according to the 4 dictionaries, and the access network device splices the 4 sub-compressed information, and then inputs the processed result into the decoder network , to get the recovery information output by the decoder network. After obtaining the recovery information output by the decoder network, the access network device can recover the downlink channel information according to the recovery information.
  • the loss function described in the embodiment shown in FIG. 10 may be applied during the training process, so that the performance of the trained codec network is better.
  • the communication device provided by the present disclosure is introduced.
  • the access network device and the UE include hardware structures and/or software modules corresponding to each function.
  • the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed by hardware or computer software drives the hardware depends on the specific application scenario and design constraints of the technical solution.
  • the present disclosure provides a communication device.
  • the communication device includes, for example, a processing unit and a transceiver unit (or called a communication unit).
  • the processing unit can be used to implement the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 or FIG. 10
  • the processing function of the UE described in the illustrated embodiment, the transceiver unit may be used to implement all the transmitting and receiving functions of the UE described in the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 , or the embodiment shown in FIG. 10 or Some send and receive functions.
  • the processing unit may be used to implement the processing functions implemented by the access network device described in the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 , or the embodiment shown in FIG. 10
  • the transceiver unit may be used to implement the All or part of the sending and receiving functions of the access network device described in the shown embodiment, the embodiment shown in FIG. 8 , or the embodiment shown in FIG. 10 .
  • the processing unit and/or the transceiver unit may be realized by a virtual module, for example, the processing unit may be realized by a software function unit or a virtual device, and the transceiver unit may be realized by a software function unit or a virtual device.
  • the processing unit and/or the transceiver unit may also be implemented by a physical device (such as a circuit system and/or a processor, etc.). The following describes the case where the processing unit and the transceiver unit are implemented by a physical device.
  • Fig. 12 shows a schematic structural diagram of a communication device provided by the present disclosure.
  • the communication device 1200 may be the UE described in the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 , or the embodiment shown in FIG. 10 , a circuit system of the UE, or a circuit system applicable to the UE, etc. , for implementing the method corresponding to the UE in the foregoing method embodiments.
  • the communication device 1200 may be the access network device described in the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 , or the embodiment shown in FIG.
  • a circuit system is a chip system.
  • the communication device 1200 includes one or more processors 1201 .
  • the processor 1201 may implement certain control functions.
  • the processor 1201 may be a general-purpose processor or a special-purpose processor. For example, including: baseband processor, central processing unit, etc.
  • the baseband processor can be used to process communication protocols and communication data.
  • the central processing unit can be used to control the communication device 1200, execute software programs and/or process data.
  • Different processors may be independent devices, or may be arranged in one or more processing circuits, for example, integrated on one or more application-specific integrated circuits.
  • the communication device 1200 includes one or more memories 1202 for storing instructions 1204, and the instructions 1204 can be executed on the processor, so that the communication device 1200 executes the methods described in the foregoing method embodiments.
  • data may also be stored in the memory 1202 .
  • the processor and memory can be set separately or integrated together.
  • the memory can be a non-volatile memory, such as a hard disk (hard disk drive, HDD) or a solid-state drive (solid-state drive, SSD), etc., or a volatile memory (volatile memory), such as a random access memory (random -access memory, RAM).
  • a memory is, but is not limited to, any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer.
  • the memory in the present disclosure may also be a circuit or any other device capable of implementing a storage function for storing program instructions and/or data.
  • the communication device 1200 may store instructions 1203 (sometimes also referred to as codes or programs), and the instructions 1203 may be executed on the processor, so that the communication device 1200 executes the methods described in the above-mentioned embodiments .
  • Data may be stored in the processor 1201 .
  • the processing unit is realized by one or more processors 1201, or, the processing unit is realized by one or more processors 1201 and one or more memories 1202, or, the processing unit is realized by one or more Processor 1201, one or more memories 1202, and instructions 1203 implement.
  • the communication device 1200 may further include a transceiver 1205 and an antenna 1206 .
  • the transceiver 1205 may be called a transceiver unit, a transceiver, a transceiver circuit, a transceiver, an input/output interface, etc., and is used to realize the transceiver function of the communication device 1200 through the antenna 1206 .
  • the transceiver unit is implemented by a transceiver 1205
  • the transceiver unit is implemented by a transceiver 1205 and an antenna 1206 .
  • the communication device 1200 may further include one or more of the following components: a wireless communication module, an audio module, an external memory interface, an internal memory, a universal serial bus (universal serial bus, USB) interface, a power management module, an antenna, Speakers, microphones, I/O modules, sensor modules, motors, cameras, or displays, etc. It can be understood that, in some embodiments, the communication device 1200 may include more or fewer components, or some components may be integrated, or some components may be separated. These components may be realized by hardware, software, or a combination of software and hardware.
  • the processor 1201 and transceiver 1205 described in this disclosure can be implemented in integrated circuit (integrated circuit, IC), analog IC, radio frequency integrated circuit (radio frequency identification, RFID), mixed signal IC, application specific integrated circuit (application specific integrated circuit) , ASIC), printed circuit board (printed circuit board, PCB), or electronic equipment, etc.
  • the communication device described in this article may be an independent device (for example, an independent integrated circuit, a mobile phone, etc.), or it may be a part of a larger device (for example, a module that can be embedded in other devices).
  • a module for example, a module that can be embedded in other devices.
  • the present disclosure provides a terminal device, which can be used in the foregoing embodiments.
  • the terminal device includes corresponding means, units and/or circuits for realizing the UE functions described in the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 or the embodiment shown in FIG. 10 .
  • a terminal device includes a transceiver module (or called a transceiver unit), configured to support the terminal device to implement a transceiver function, and a processing module (or called a processing unit), configured to support the terminal device to process signals.
  • the present disclosure also provides an access network device, which can be used in the foregoing embodiments.
  • the access network device includes corresponding means and units for realizing the functions of the access network device described in the embodiment shown in FIG. 5 , the embodiment shown in FIG. 8 , or the embodiment shown in FIG. 10 and/or circuits.
  • the access network device includes a transceiver module (or called a transceiver unit), used to support the access network device to implement the transceiver function, and a processing module (or called a processing unit), used to support the access network device Process the signal.
  • the technical solution provided by the present disclosure may be fully or partially realized by software, hardware, firmware or any combination thereof.
  • software When implemented using software, it may be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the computer program instructions When the computer program instructions are loaded and executed on a computer, the processes or functions according to the present disclosure are produced in whole or in part.
  • the computer may be a general computer, a special computer, a computer network, an access network device, a terminal device, an AI node or other programmable devices.
  • the computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website, computer, server or data center Transmission to another website site, computer, server or data center by wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.).
  • the computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center integrated with one or more available media.
  • the available medium may be a magnetic medium (for example, a floppy disk, a hard disk, or a magnetic tape), an optical medium (for example, a digital video disc (digital video disc, DVD)), or a semiconductor medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Communication Control (AREA)

Abstract

本公开涉及一种通信方法及装置。终端设备获得M份第一子下行信道数据，其中每份第一子下行信道数据对应于M个数据空间中的一个数据空间。对于M份第一子下行信道数据中的第i份第一子下行信道数据，终端设备根据与第i个数据空间对应的第一字典确定第i份第一子下行信道数据对应的第一信息，共确定M个第一信息。终端设备发送第一指示信息，以指示M个第一信息。不同的数据空间能够表征不同的信道环境信息，终端设备反馈不同的数据空间对应的第一信息，可以使得接入网设备能够明确第一信息与环境信息之间的对应关系，从而使得终端设备所反馈的第一信息能够反映实际的通信环境，提高了终端设备所反馈的第一信息的准确性。

Description

一种通信方法及装置
相关申请的交叉引用
本申请要求在2021年12月31日提交中国国家知识产权局、申请号为202111663303.8、申请名称为“一种通信方法及装置”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
本公开涉及通信技术领域，尤其涉及一种通信方法及装置。
背景技术
第五代(the 5th generation，5G)移动通信系统对系统容量、频谱效率等方面有了更高的要求。在5G通信系统中，大规模多输入多输出(massive multiple-input multiple-output，massive-MIMO)技术的应用对提高系统的频谱效率起到了至关重要的作用。利用massive-MIMO技术，基站可以同时为更多的用户设备(user equipment，UE)提供高质量的服务。这其中较为关键的环节就是基站对多UE的下行数据进行预编码，通过预编码可以实现空分复用(spatial multiplexing)，降低UE间的干扰，提升接收端的信干噪比(signal to interference plus noise ratio，SINR)，由此提升系统吞吐率。基站为了更加准确地对UE的下行数据进行预编码，可以获得下行信道的信道状态信息(channel state information，CSI)，根据CSI恢复出下行信道，并利用恢复的下行信道确定预编码矩阵，以进行预编码。因此，如何使得UE所反馈的CSI更为准确，是一个值得研究的技术问题。
发明内容
本公开提供一种通信方法及装置，用于提高UE所反馈的CSI的准确性。
第一方面，提供第一种通信方法，该方法可在终端设备侧执行。该方法可通过软件、硬件、或软硬件结合的方式执行。例如，该方法由终端设备执行，或由电路系统执行，或者由包括终端设备的较大设备执行，该电路系统能够实现终端设备的功能。该方法包括：获得M份第一子下行信道数据，其中每份第一子下行信道数据对应于M个数据空间中的一个数据空间，M为大于1的整数；对于所述M份第一子下行信道数据中的第i份第一子下行信道数据，根据与所述M个数据空间中的第i个数据空间对应的第一字典确定所述第i份第一子下行信道数据对应的第一信息，共确定M个第一信息，i取从1至M的整数，所述第i份第一子下行信道数据对应于所述第i个数据空间，所述第一字典包括多个元素，所述第i份第一子下行信道数据对应的第一信息对应于所述多个元素中的P个元素，P为正整数；发送第一指示信息，所述第一指示信息用于指示所述M个第一信息。
在本公开中，终端设备获得的M份第一子下行信道数据中的每份第一子下行信道数据可以对应M个数据空间中的一个数据空间，根据不同的数据空间对应的字典可以确定各份第一子下行信道数据对应的第一信息。而不同的数据空间能够表征不同的位置信息，或者说能够表征不同的信道环境信息，终端设备反馈不同的数据空间对应的第一信息，可以使 得接入网设备能够明确第一信息与环境信息之间的对应关系，从而使得终端设备所反馈的第一信息能够反映实际的通信环境，提高了终端设备所反馈的第一信息的准确性。接入网设备根据终端设备所反馈的第一信息，能够恢复得到较为准确的下行信道。
在一种可选的实施方式中，所述第一指示信息用于指示所述M个第一信息的标识，其中，发送所述第一信息，包括：按照第一顺序发送所述M个第一信息的标识，所述第一顺序为所述M个数据空间的排列顺序。第一顺序规定了终端设备先发送哪个数据空间对应的第一信息的标识，后发送哪个数据空间对应的第一信息的标识。对于终端设备和接入网设备来说，第一顺序都是已知的，因此接入网设备接收M个第一信息的标识后，也能明确第一信息的标识与数据空间之间的对应关系，从而避免对应出错。
在一种可选的实施方式中，所述第一顺序为预定义的顺序；或，接收第二指示信息，所述第二指示信息用于指示所述第一顺序；或，确定所述第一顺序，并发送第三指示信息，所述第三指示信息用于指示所述第一顺序。例如，第一顺序为协议预定义的顺序，终端设备和接入网设备根据协议就能确定第一顺序。或者，第一顺序也可以预配置在终端设备和接入网设备中。或者，第一顺序可由接入网设备确定，接入网设备确定第一顺序后可以向终端设备发送第二指示信息，使得终端设备根据第二指示信息能够确定第一顺序。或者，第一顺序可由终端设备确定，终端设备确定第一顺序后可以向接入网设备发送第三指示信息，使得接入网设备根据第三指示信息能够确定第一顺序。可见，确定第一顺序的方式是较为灵活的。
在一种可选的实施方式中，所述M份第一子下行信道数据是根据第一下行信道数据得到的，其中，所述第一下行信道数据为预处理结果；或，所述第一下行信道数据包括预处理结果中连续的F列数据；或，所述第一下行信道数据为对预处理结果进行压缩所得到的压缩信息。其中，所述预处理结果是对第二下行信道数据进行预处理得到的。可以直接将第二下行信道数据的预处理结果作为第一下行信道数据，无需再对该预处理结果进行过多处理，较为简单。或者，考虑在频域方向(时延域)上，能量一般都主要集中在时延＝0的附近，除此之外其他区域的能量基本可以忽略不计。因此，终端设备可以选择时延＝0的两侧的连续的F列作为第一下行信道数据即可，剩余的部分可以默认系数为0，由此可以减小处理第一下行信道数据时的复杂度。或者，也可以对预处理结果进行压缩后得到第一下行信道数据，由此能够减小处理第一下行信道数据时的复杂度。对一个下行信道数据进行预处理的过程，例如包括对该下行信道数据进行空频联合投影等。
在一种可选的实施方式中，所述M个数据空间的划分方式为预定义；或，接收第四指示信息，所述第四指示信息用于指示所述M个数据空间的划分方式；或，确定所述M个数据空间的划分方式，并发送第五指示信息，所述第五指示信息用于指示所述M个数据空间的划分方式。例如，M个数据空间的划分方式通过协议预定义，则终端设备和接入网设备都可以根据协议确定M个数据空间的划分方式。或者，M个数据空间的划分方式由接入网设备确定，接入网设备可向终端设备发送第四指示信息，使得终端设备根据第四指示信息能够确定M个数据空间的划分方式。或者，M个数据空间的划分方式可由UE确定，UE可向接入网设备发送第五指示信息，使得接入网设备根据第五指示信息可确定M个数据空间的划分方式。可见，数据空间的划分方式较为灵活。
第二方面，提供第二种通信方法，该方法可在接入网设备侧执行。该方法可通过软件、硬件、或软硬件结合的方式执行。例如，该方法由接入网设备执行，或由包括接入网设备 的较大设备执行，或由电路系统执行，该电路系统能够实现接入网设备的功能，或者由独立于接入网设备的AI模块辅助接入网设备或者辅助接入网设备的网元执行，不予限制。示例性地，所述接入网设备为接入网设备，例如基站。该方法包括：接收第一指示信息，所述第一指示信息用于指示M个第一信息，M为大于1的整数；对于所述M个第一信息中的第i个第一信息，根据M个数据空间中的第i个数据空间对应的第一字典，恢复第i份第二子下行信道数据，共得到M份第二子下行信道数据，所述第i个第一信息对应于所述第i个数据空间，i取从1至M的整数，所述第一字典包括多个元素，所述第i份第二子下行信道数据对应的第一信息对应于所述多个元素中的P个元素；根据所述M份第二子下行信道数据，恢复得到下行信道信息。
在一种可选的实施方式中，接收第一指示信息，包括：按照第一顺序接收所述M个第一信息的标识，所述第一顺序为所述M个数据空间的排列顺序。
在一种可选的实施方式中，所述第一顺序为预定义的顺序；或，发送第二指示信息，所述第二指示信息用于指示所述第一顺序；或，接收第三指示信息，所述第三指示信息用于指示所述第一顺序。
在一种可选的实施方式中，所述M个数据空间对应M个字典，其中每个数据空间对应一个字典；或，所述M个数据空间均对应于同一个字典；或，所述M个数据空间对应的字典的数量大于1且小于M。即，数据空间与字典可以是一一对应的关系，这样可以提高根据字典所确定的第一信息的准确性；或者，所有的数据空间可以统一对应一个字典，用于训练得到该字典的样本可以更为丰富，使得该字典所包括的内容更为细致；或者，数据空间对应的字典的数量可以小于数据空间的数量，例如一个字典可对应多个数据空间，这样能够在一定程度上减小复杂度。
在一种可选的实施方式中，根据所述M份第二子下行信道数据，恢复得到下行信道信息，包括：根据所述M份第二子下行信道数据得到压缩信息；根据所述压缩信息得到所述下行信道信息。
在一种可选的实施方式中，所述M个数据空间的划分方式为预定义；或，发送第四指示信息，所述第四指示信息用于指示所述M个数据空间的划分方式；或，接收第五指示信息，所述第五指示信息用于指示所述M个数据空间的划分方式。
关于第二方面或第二方面的各种可选的实施方式所带来的技术效果，可参考对于第一方面或相应的实施方式的技术效果的介绍。
第三方面，提供一种通信装置。所述通信装置可以实现上述第一方面所述的方法。所述通信装置具备上述终端设备的功能。在一种可选的实现方式中，该装置可以包括执行第一方面中所描述的方法/操作/步骤/动作所一一对应的模块，该模块可以是硬件电路，也可是软件，也可以是硬件电路结合软件实现。在一种可选的实现方式中，所述通信装置包括基带装置和射频装置。在另一种可选的实现方式中，所述通信装置包括处理单元(有时也称为处理模块)和收发单元(有时也称为收发模块)。收发单元能够实现发送功能和接收功能，在收发单元实现发送功能时，可称为发送单元(有时也称为发送模块)，在收发单元实现接收功能时，可称为接收单元(有时也称为接收模块)。发送单元和接收单元可以是同一个功能模块，该功能模块称为收发单元，该功能模块能实现发送功能和接收功能；或者，发送单元和接收单元可以是不同的功能模块，收发单元是对这些功能模块的统称。
其中，所述处理单元，用于获得M份第一子下行信道数据，其中每份第一子下行信道 数据对应于M个数据空间中的一个数据空间，M为大于1的整数；对于所述M份第一子下行信道数据中的第i份第一子下行信道数据，所述处理单元，还用于根据与所述M个数据空间中的第i个数据空间对应的第一字典确定所述第i份第一子下行信道数据对应的第一信息，共确定M个第一信息，i取从1至M的整数，所述第i份第一子下行信道数据对应于所述第i个数据空间，所述第一字典包括多个元素，所述第i份第一子下行信道数据对应的第一信息对应于所述多个元素中的P个元素，P为正整数；所述收发单元，用于发送第一指示信息，所述第一指示信息用于指示所述M个第一信息。
再例如，所述通信装置包括：处理器，与存储器耦合，用于执行存储器中的指令，以实现上述第一方面的方法。可选的，该通信装置还包括其他部件，例如，天线，输入输出模块，接口等等。这些部件可以是硬件，软件，或者软件和硬件的结合。
第四方面，提供一种通信装置。所述通信装置可以实现上述第二方面所述的方法。所述通信装置具备上述接入网设备的功能。所述接入网设备例如为基站，或为基站中的基带装置等。在一种可选的实现方式中，该装置可以包括执行第二方面中所描述的方法/操作/步骤/动作所一一对应的模块，该模块可以是硬件电路，也可是软件，也可以是硬件电路结合软件实现。在一种可选的实现方式中，所述通信装置包括基带装置和射频装置。在另一种可选的实现方式中，所述通信装置包括处理单元(有时也称为处理模块)和收发单元(有时也称为收发模块)。关于收发单元的实现方式，可参考第三方面的相关介绍。
其中，所述收发单元，用于接收第一指示信息，所述第一指示信息用于指示M个第一信息，M为大于1的整数；对于所述M个第一信息中的第i个第一信息，所述处理单元，用于根据M个数据空间中的第i个数据空间对应的第一字典，恢复第i份第二子下行信道数据，共得到M份第二子下行信道数据，所述第i个第一信息对应于所述第i个数据空间，i取从1至M的整数，所述第一字典包括多个元素，所述第i份第二子下行信道数据对应的第一信息对应于所述多个元素中的P个元素；所述处理单元，还用于根据所述M份第二子下行信道数据，恢复得到下行信道信息。
再例如，所述通信装置包括：处理器，与存储器耦合，用于执行存储器中的指令，以实现上述第二方面的方法。可选的，该通信装置还包括其他部件，例如，天线，输入输出模块，接口等等。这些部件可以是硬件，软件，或者软件和硬件的结合。
第五方面，提供一种计算机可读存储介质，所述计算机可读存储介质用于存储计算机程序或指令，当其被运行时，使得第一方面和/或第二方面的方法被实现。
第六方面，提供一种包含指令的计算机程序产品，当其在计算机上运行时，使得第一方面和/或第二方面所述的方法被实现。
第七方面，提供一种芯片系统，该芯片系统包括处理器，还可以包括存储器，用于实现上述第一方面和/或第二方面的方法。该芯片系统可以由芯片构成，也可以包含芯片和其他分立器件。
第八方面，提供一种通信系统，包括第三方面的通信装置和第四方面的通信装置。
附图说明
图1为一种通信系统的示意图；
图2为CSI反馈机制的流程图；
图3为一种应用场景的示意图；
图4A～图4E为AI在通信系统中的几种应用框架的示意图；
图5为一种通信方法的流程图；
图6为字典的一种示意图；
图7为UE和接入网设备均处理压缩信息的情况下的通信方法的示意图；
图8为另一种通信方法的流程图；
图9A为网络训练阶段和网络推理阶段的一种示意图；
图9B～图9D为网络训练阶段的几种示意图；
图10为又一种通信方法的流程图；
图11为网络训练阶段和网络推理阶段的又一种示意图；
图12为通信装置的一种示意性框图。
具体实施方式
为了使本公开的目的、技术方案和优点更加清楚，下面将结合附图对本公开作进一步地详细描述。
本公开提供的技术可以应用于图1所示的通信系统10中。通信系统10包括一个或多个通信装置30(例如，终端设备)。该一个或多个通信装置30经由一个或多个接入网(radio access network，RAN)设备20连接到一个或多个核心网(core network，CN)设备，以实现多个通信设备之间的通信。例如，通信系统10是支持第四代(the 4th generation，4G)(包括长期演进(long term evolution，LTE))接入技术的通信系统，支持5G(有时也称为new radio，NR)接入技术的通信系统，无线保真(wireless fidelity，Wi-Fi)系统，第三代合作伙伴计划(3rd generation partnership project，3GPP)相关的蜂窝系统，支持多种无线技术融合的通信系统，或者是面向未来的演进系统等，不予限制。
下面分别对图1所涉及的终端设备和RAN进行详细说明。
1、终端设备。
终端设备可以简称为终端。终端设备可以是一种具有无线收发功能的设备。终端设备可以是移动的，或固定的。终端设备可以部署在陆地上，包括室内或室外，手持或车载；也可以部署在水面上(如轮船等)；还可以部署在空中(例如飞机、气球和卫星上等)。所述终端设备可以包括手机(mobile phone)、平板电脑(pad)、带无线收发功能的电脑、虚拟现实(virtual reality，VR)终端设备、增强现实(augmented reality，AR)终端设备、工业控制(industrial control)中的无线终端设备、无人驾驶(self driving)中的无线终端设备、远程医疗(remote medical)中的无线终端设备、智能电网(smart grid)中的无线终端设备、运输安全(transportation safety)中的无线终端设备、智慧城市(smart city)中的无线终端设备、和/或智慧家庭(smart home)中的无线终端设备。终端设备还可以是蜂窝电话、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或计算设备、车载设备、可穿戴设备，未来第五代(the 5th generation，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。终端设备有时也可以称为用户设备(user equipment，UE)。可选的，终端设备可以与不同技术的多个接入网设备进行通信，例如，终端设备可以与支持LTE的接入网设备通信，也可以与支持5G的接入网设备通信，又可以与支持LTE的接入网设备以 及支持5G的接入网设备的双连接。本公开并不限定。
本公开中，用于实现终端设备的功能的装置可以是终端设备；也可以是能够支持终端设备实现该功能的装置，例如芯片系统、硬件电路、软件模块、或硬件电路加软件模块，该装置可以被安装在终端设备中或可以与终端设备匹配使用。本公开提供的技术方案中，以用于实现终端设备的功能的装置是终端设备，终端设备是UE为例，描述本公开提供的技术方案。
本公开中，芯片系统可以由芯片构成，也可以包括芯片和其他分立器件。
2、RAN。
RAN可以包括一个或多个RAN设备，比如RAN设备20。RAN设备与终端设备之间的接口可以为Uu接口(或称为空口)。在未来通信中，这些接口的名称可以不变，或者也可以用其它名称代替，本公开对此不作限定。
RAN设备为将终端设备接入到无线网络的节点或设备，RAN设备又可以称为网络设备或基站。RAN设备例如包括但不限于：基站、5G中的下一代节点B(generation nodeB，gNB)、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(base band unit，BBU)、收发点(transmitting and receiving point，TRP)、发射点(transmitting point，TP)、和/或移动交换中心等。或者，接入网设备还可以是集中单元(centralized unit，CU)、分布单元(distributed unit,DU)、集中单元控制面(CU control plane，CU-CP)节点、集中单元用户面(CU user plane，CU-UP)节点、接入回传一体化(integrated access and backhaul，IAB)、或云无线接入网络(cloud radio access network，CRAN)场景下的无线控制器等中的至少一个。或者，接入网设备可以为中继站、接入点、车载设备、终端设备、可穿戴设备、5G网络中的接入网设备或者未来演进的公共陆地移动网络(public land mobile network，PLMN)中的接入网设备等。
本公开中，用于实现接入网设备的功能的装置可以是接入网设备；也可以是能够支持接入网设备实现该功能的装置，例如芯片系统、硬件电路、软件模块、或硬件电路加软件模块，该装置可以被安装在接入网设备中或可以与接入网设备匹配使用。在本公开提供的技术方案中，以用于实现接入网设备的功能的装置是接入网设备，接入网设备是基站为例，描述本公开提供的技术方案。
(1)协议层结构。
接入网设备和终端设备之间的通信遵循一定的协议层结构。该协议层结构可以包括控制面协议层结构和用户面协议层结构。例如，控制面协议层结构可以包括以下至少一项：无线资源控制(radio resource control，RRC)层、分组数据汇聚层协议(packet data convergence protocol，PDCP)层、无线链路控制(radio link control，RLC)层、媒体接入控制(media access control，MAC)层或物理层(physical，PHY)等。例如，用户面协议层结构可以包括以下至少一项：业务数据适配协议(service data adaptation protocol，SDAP)层、PDCP层、RLC层、MAC层和物理层等。
上述接入网设备和终端设备之间的协议层结构可以看作接入层(access stratum，AS)结构。可选的，在AS之上，还可以存在非接入层(non-access stratum，NAS)，用于接入网设备向终端设备转发来自核心网设备的信息，或者用于接入网设备向核心网设备转发来 自终端设备的信息。此时，可以认为终端设备和核心网设备之间存在逻辑接口。可选的，接入网设备可以通过透传的方式转发终端设备和核心网设备之间的信息。例如，NAS消息可以映射到或者包含于RRC信令中，作为RRC信令的元素。
可选的，接入网设备和终端设备之间的协议层结构还可以包括人工智能(artificial intelligence，AI)层，用于传输AI功能相关的数据。
(2)集中单元(central unit，CU)和分布单元(distributed unit，DU)。
RAN设备可以包括CU和DU。该设计可以称为CU和DU分离。多个DU可以由一个CU集中控制。作为示例，CU和DU之间的接口可以称为F1接口。其中，控制面(control panel，CP)接口可以为F1-C，用户面(user panel，UP)接口可以为F1-U。本公开不限制各接口的具体名称。CU和DU可以根据无线网络的协议层划分：比如，PDCP层及以上协议层(例如RRC层和SDAP层等)的功能设置在CU，PDCP层以下协议层(例如RLC层、MAC层和PHY层等)的功能设置在DU；又比如所示，PDCP层以上协议层的功能设置在CU，PDCP层及以下协议层的功能设置在DU。
上述对CU和DU的处理功能按照协议层的划分仅仅是一种举例，也可以按照其他的方式进行划分。例如可以将CU或者DU划分为具有更多协议层的功能，又例如将CU或DU还可以划分为具有协议层的部分处理功能。在一种设计中，将RLC层的部分功能和RLC层以上的协议层的功能设置在CU，将RLC层的剩余功能和RLC层以下的协议层的功能设置在DU。在另一种设计中，还可以按照业务类型或者其他系统需求对CU或者DU的功能进行划分，例如按时延划分，将处理时间需要满足时延要求的功能设置在DU，不需要满足该时延要求的功能设置在CU。
可选地，CU也可以具有核心网的一个或多个功能。示例性的，CU可以设置在网络侧方便集中管理。
可选的，将DU的无线单元(radio unit，RU)拉远设置。其中，RU具有射频功能。示例性地，DU和RU可以在PHY层进行划分。例如，DU可以实现PHY层中的高层功能，RU可以实现PHY层中的低层功能。其中，用于发送时，PHY层的功能可以包括以下至少一项：添加循环冗余校验(cyclic redundancy check，CRC)位、信道编码、速率匹配、加扰、调制、层映射、预编码、资源映射、物理天线映射、或射频发送功能。用于接收时，PHY层的功能可以包括以下至少一项：CRC校验、信道解码、解速率匹配、解扰、解调、解层映射、信道检测、资源解映射、物理天线解映射、或射频接收功能。其中，PHY层中的高层功能可以包括PHY层的一部分功能，该部分功能更加靠近MAC层；PHY层中的低层功能可以包括PHY层的另一部分功能，例如该部分功能更加靠近射频功能。例如，PHY层中的高层功能可以包括添加CRC位、信道编码、速率匹配、加扰、调制、和层映射，PHY层中的低层功能可以包括预编码、资源映射、物理天线映射、和射频发送功能；或者，PHY层中的高层功能可以包括添加CRC位、信道编码、速率匹配、加扰、调制、层映射和预编码，PHY层中的低层功能可以包括资源映射、物理天线映射、和射频发送功能。例如，PHY层中的高层功能可以包括CRC校验、信道解码、解速率匹配、解码、解调、和解层映射，PHY层中的低层功能可以包括信道检测、资源解映射、物理天线解映射、和射频接收功能；或者，PHY层中的高层功能可以包括CRC校验、信道解码、解速率匹配、解码、解调、解层映射、和信道检测，PHY层中的低层功能可以包括资源解映射、物理天线解映射、和射频接收功能。
可选地，可以对CU的功能进行进一步划分，将控制面和用户面分离并通过不同实体来实现，分别为控制面CU实体(即CU-CP实体)和用户面CU实体(即CU-UP实体)。该CU-CP实体和CU-UP实体可以分别与DU相耦合或者相连接，共同完成RAN设备的功能。
在上述描述的架构中，CU产生的信令可以通过DU发送给终端设备，或者终端设备产生的信令可以通过DU发送给CU。例如，RRC或PDCP层的信令最终可以被处理为物理层的信令发送给终端设备，或者，由接收到的物理层的信令转变而来。在这种架构下，该RRC或PDCP层的信令，即可以认为是通过DU发送的，或者，通过DU和RU发送的。
可选的，上述DU、CU、CU-CP、CU-UP和RU中的任一个可以是软件模块、硬件结构、或者软件模块+硬件结构，不予限制。其中，不同实体的存在形式可以是不同的，不予限制。例如DU、CU、CU-CP、CU-UP是软件模块，RU是硬件结构。这些模块及其执行的方法也在本公开的保护范围内。例如，本公开的方法由接入网设备执行时，具体可以由CU、CU-CP、CU-UP、DU、RU或下文介绍的近实时RIC中的至少一项执行。各模块所执行的方法也在本公开的保护范围内。
需要说明的是，因本公开所涉及的网络设备主要是接入网设备，因此在后文中，如无特殊说明，则所述的“网络设备”可以指“接入网设备”。
应理解，图1所示的通信系统中各个设备的数量仅作为示意，本公开并不限于此，实际应用中在通信系统中还可以包括更多的终端设备、更多的RAN设备，还可以包括其它设备，例如可以包括核心网设备，和/或用于实现人工智能功能的节点。
上述图1所示的网络架构可以适用于各种无线接入技术(radio access technology，RAT)的通信系统，例如4G通信系统，也可以是5G(或者称为新无线(new radio，NR))通信系统，也可以是LTE通信系统与5G通信系统之间的过渡系统，该过渡系统也可以称为4.5G通信系统，或者也可以是未来的通信系统，例如6G通信系统。本公开描述的网络架构以及业务场景是为了更加清楚的说明本公开的技术方案，并不构成对于本公开提供的技术方案的限定，本领域普通技术人员可知，随着通信网络架构的演变和新业务场景的出现，本公开提供的技术方案对于类似的技术问题，同样适用。
本公开提供的方法除了可以用于接入网设备和终端设备之间的通信，也可以用于其他通信设备之间的通信，例如无线回传链路中宏基站和微基站之间的通信，例如侧行链路(sidelink，SL)中第一终端设备和第二终端设备之间的通信，不予限制。本公开以网络设备和终端设备之间的通信为例进行描述。
接入网设备向终端设备发送数据时，可以基于终端设备反馈的信道状态信息(channel state information，CSI)进行预编码。为了便于理解本公开，下面对本公开中涉及的一些技术术语做简单说明。
1、预编码技术。
接入网设备可以在已知信道状态信息的情况下，借助与信道条件相匹配的预编码矩阵对待发送的信号进行处理。通过该技术，可以使得经过预编码的待发送的信号与信道相适配，从而使得终端设备接收到的信号的质量(例如信干噪比(signal to interference plus noise ratio，SINR)等)得以提升，从而可以提升系统吞吐率。采用预编码技术，可以实现发送设备(如接入网设备)与多个接收设备(如终端设备)在相同的时频资源上有效地传输，即有效地实现多用户多输入多输出(multiple user multiple input multiple output，MU-MIMO)。 采用预编码技术，可以实现发送设备(如接入网设备)与接收设备(如终端设备)在相同的时频资源上有效地进行多数据流传输，即有效地实现单用户多输入多输出(single user multiple input multiple output，SU-MIMO)。应注意，有关预编码技术的相关描述仅为便于理解而示例，并非用于限制本公开的公开范围。在具体实现过程中，发送设备还可以通过其他方式进行预编码。例如，在无法获知信道信息(例如但不限于信道矩阵)的情况下，采用预先设置的预编码矩阵或者加权处理方式进行预编码等。为了简洁，其具体内容本文不再赘述。
2、CSI反馈(CSI feedback)。
CSI反馈还可以称为CSI报告(CSI report)。CSI反馈是在无线通信系统中，由数据(例如但不限于物理下行共享信道(physical downlink shared channel，PDSCH)上承载的数据)的接收端(如终端设备)向发送端(如接入网设备)上报用于描述通信链路的信道属性的信息。CSI报告例如包括下行信道矩阵、预编码矩阵指示(precoding matrix indicator，PMI)、秩指示(rank indicator，RI)、或信道质量指示(channel quality indicator，CQI)等信息中的一项或多项。以上列举的CSI包括的内容仅为示例性说明，不应对本公开构成任何限定。CSI可以包括如上一项或多项，也可以包括除上述列举之外的其他用于表征CSI的信息，本公开对此不作限定。
3、神经网络(neural network，NN)。
神经网络是机器学习技术的一种具体实现形式。根据通用近似定理，神经网络在理论上可以逼近任意连续函数，从而使得神经网络具备学习任意映射的能力。传统的通信系统需要借助丰富的专家知识来设计通信模块，而基于神经网络的深度学习通信系统可以从大量的数据集中自动发现隐含的模式结构，建立数据之间的映射关系，获得优于传统建模方法的性能。
例如，深度神经网络(deep neural network，DNN)是层数较多的一种神经网络。按照网络结构和/或使用场景的不同，DNN可以包括多层感知机(multi-layer perceptron，MLP)、卷积神经网络(convolutional neural networks，CNN)和递归神经网络(recurrent neural network，RNN)等。本公开不限制DNN的具体形式。
4、自编码器(auto-encoder，AE)网络，或者简称为AE。
AE网络可包括编码器(encoder)和对应的解码器(decoder)，例如编码器和/或解码器通过神经网络(如DNN)实现。此时，编码器还可以称为编码器网络，解码器还可以称为解码器网络。例如，在AE网络中，编码器和对应的解码器可联合训练得到。训练得到的编码器和解码器可以用于进行信息的编解码。
本公开中，对于名词的数目，除非特别说明，表示“单数名词或复数名词”，即"一个或多个”。“至少一个”是指一个或者多个，“多个”是指两个或两个以上。“和/或”，描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B的情况，其中A,B可以是单数或者复数。在表示特征时，字符“/”可以表示前后关联对象是一种“或”的关系。例如，A/B，表示：A或B。在表示运算时，符号“/”还可以表示除法运算。另外本公开中，符号“×”也可用符号“*”替换。
本公开提及“第一”、“第二”等序数词是用于对多个对象进行区分，不用于限定多个对象的大小、内容、顺序、时序、应用场景、优先级或者重要程度等。例如，第一指示信息和第二指示信息，可以是同一个指示信息，也可以是不同的指示信息，且，这种名称也并 不是表示这两个指示信息的大小、传输方式、指示的内容、优先级、应用场景或者重要程度等的不同。
一种可能的实现中，CSI反馈机制采用如图2所示的流程。
S21、基站发送信令，相应的，UE从基站接收该信令。
该信令用于配置信道测量信息，例如该信令通知UE以下至少一项：进行信道测量的时间信息、进行信道测量的参考信号(reference signal，RS)的类型、参考信号的时域资源、参考信号的频域资源、和测量量的上报条件等。
S22、基站向UE发送参考信号，相应的，UE从基站接收参考信号。
UE对该参考信号进行测量，可得到CSI。
S23、UE向基站发送CSI，相应的，基站从UE接收CSI。
S24、基站根据CSI向UE发送数据，相应的，UE从基站接收该数据。
基站根据CSI确定预编码矩阵，利用预编码矩阵对待发送给UE的数据进行预编码。基站向UE发送的数据承载于下行信道中，例如承载于PDSCH中。
UE所反馈的CSI精度越高，信息越丰富，基站根据CSI所恢复的下行信道就越准确，从而基站确定的预编码矩阵越准确，使得下行空分复用性能越好，UE的接收信干噪比越高，系统吞吐率也越高。然而，随着MIMO系统天线阵列规模不断增大，可支持的天线端口数也增多。由于完整的下行信道矩阵的大小与天线端口数成正比，在大规模MIMO系统中，要使得UE反馈的CSI的精度较高，就意味着巨大的反馈开销。由于巨大的反馈开销会减少数据传输的可用资源，并因此会降低系统容量。因此为了提高系统容量，需要研究如何降低CSI的反馈开销。基于双域压缩码本来反馈CSI，是一种能够有效降低反馈开销的方式。
双域压缩码本一般是根据假设的天线面板形态和子带的个数等因素设计的。而在实际的通信环境中，由于信道环境的复杂多变，以及实际的天线面板形态的多样性，针对固定的天线面板形态和子带个数所确定的码本，并不一定能满足实际的通信环境，降低了UE所反馈的CSI的准确性。因此，如何使得UE所反馈的CSI更为准确，是一个值得研究的技术问题。
鉴于此，提供本公开的技术方案。本公开中，UE获得的M份第一子下行信道数据中的每份第一子下行信道数据可以对应M个数据空间中的一个数据空间，根据不同的数据空间对应的字典可以确定各份第一子下行信道数据对应的第一信息。而不同的数据空间能够表征不同的位置信息，或者说能够表征不同的信道环境信息，UE反馈不同的数据空间对应的第一信息，可以使得接入网设备能够明确第一信息与环境信息之间的对应关系，从而使得UE所反馈的第一信息能够反映实际的通信环境，提高了UE所反馈的第一信息的准确性。接入网设备根据UE所反馈的第一信息，能够恢复得到较为准确的下行信道。
图3示出了本公开提供的通信系统10中的一种通信网络架构，后续提供的任一个实施例均可适用于该架构。图3所包括的网络设备，例如为通信系统10所包括的接入网设备20，图3所包括的终端设备，例如为通信系统10所包括的通信装置30。网络设备与终端设备能够进行通信。
本公开可能涉及机器学习技术，该机器学习技术是AI技术的具体实现，为了便于理解，下面对AI技术进行介绍。可以理解的是，该介绍并不作为对本公开的限定。
AI，是一种通过模拟人脑进行复杂计算的技术。随着数据存储和能力的能升，AI 得到了越来越多的应用。
本公开中，可以在前述图1所示的通信系统中引入独立的网元(如称为AI网元、AI节点、或AI设备等)来实现AI功能。AI网元可以和接入网设备直接连接，或者可以通过第三方网元和接入网设备实现间接连接。可选的，第三方网元可以是核心网网元。或者，可以在通信系统中的其他网元内配置或设置AI实体，用于实现AI相关的操作。其中，AI实体还可以称为AI模块、AI单元或其他名称，主要用于实现部分或全部AI功能，本公开不限制其具体名称。可选的，该其他网元可以是接入网设备、核心网设备、或网管(operation,administration and maintenance，OAM)等。在这种情况下，执行AI功能的网元为内置AI功能的网元。
本公开中，AI功能可以包括以下至少一项：数据收集、模型训练(或模型学习)、模型信息发布、模型推断(或称为模型推理、推理、或预测等)、模型监控或模型校验、或推理结果发布等。AI功能还可以称为AI(相关的)操作、或AI相关的功能。
本公开中，OAM用于操作、管理和/或维护核心网设备(核心网设备的网管)，和/或，用于操作、管理和/或维护接入网设备(接入网设备的网管)。例如，本公开中包括第一OAM和第二OAM，第一OAM是核心网设备的网管，第二OAM是接入网设备的网管。可选的，第一OAM和/或第二OAM中包括AI实体。再例如，本公开中包括第三OAM，第三OAM同时是核心网设备和接入网设备的网管。可选的，第三OAM中包括AI实体。
如图4A所示，为AI在通信系统中的第一种应用框架的示意图。数据源(data source)用于存储训练数据和推理数据。模型训练节点(model training host)通过对数据源提供的训练数据(training data)进行训练或者更新训练，得到AI模型，且将AI模型部署在模型推理节点(model inference host)中。其中，AI模型表征了模型的输入和输出之间的映射关系。通过模型训练节点学习得到AI模型，相当于由模型训练节点利用训练数据学习得到模型的输入和输出之间的映射关系。模型推理节点使用AI模型，基于数据源提供的推理数据进行推理，得到推理结果。该方法还可以描述为：模型推理节点将推理数据输入到AI模型，通过AI模型得到输出，该输出即为推理结果。该推理结果可以指示：由执行对象使用(执行)的配置参数、和/或由执行对象执行的操作。推理结果可以由执行(actor)实体统一规划，并发送给一个或多个执行对象(例如，核心网网元、基站或UE等)去执行。可选的，模型推理节点可以将其推理结果反馈给模型训练节点，该过程可以称为模型反馈，所反馈的推理结果用于模型训练节点更新AI模型，并将更新后的AI模型部署在模型推理节点中。可选的，执行对象可以将其收集到的网络参数反馈给数据源，该过程可以称为表现反馈，所反馈的网络参数可以作为训练数据或推理数据。
例如，上述AI模型包括AE网络中的解码器网络。解码器网络被部署在接入网设备侧。解码器网络的推理结果例如用于下行信道矩阵的重构。上述AI模型包括AE网络中的编码器网络。其中，编码器网络被部署在UE侧。编码器网络的推理结果例如用于下行信道矩阵的编码。
图4A所示的应用框架可以部署在图1中所示的网元。例如，图4A的应用框架可以部署在图1的终端设备、接入网设备、核心网设备(未示出)或独立部署的AI网元(未示出)中的至少一项。例如，AI网元(可看做模型训练节点)可对终端设备和/ 或接入网设备提供的训练数据(training data)进行分析或训练，得到一个模型。终端设备、接入网设备、或核心网设备中的至少一项(可看做模型推理节点)可以使用该模型和推理数据进行推理，得到模型的输出。其中，推理数据可以是由终端设备和/或接入网设备提供的。该模型的输入包括推理数据，该模型的输出即为该模型所对应的推理结果。终端设备、接入网设备、或核心网设备中的至少一项(可看做执行对象)可以根据推理结果进行相应的操作。其中，模型推理节点和执行对象可以相同，也可以不同，不予限制。
下面结合图4B～图4E对本公开提供的方法能够应用的网络架构进行举例介绍。
如图4B所示，第一种可能的实现中，接入网设备中包括近实时接入网智能控制(RAN intelligent controller，RIC)模块，用于进行模型训练和推理。例如，近实时RIC可以用于训练AI模型，利用该AI模型进行推理。例如，近实时RIC可以从CU、DU或RU中的至少一项获得网络侧和/或终端侧的信息，该信息可以作为训练数据或者推理数据。可选的，近实时RIC可以将推理结果递交至CU、DU、RU或终端设备中的至少一项。可选的，CU和DU之间可以交互推理结果。可选的，DU和RU之间可以交互推理结果，例如近实时RIC将推理结果递交至DU，由DU转发给RU。
如图4B所示，第二种可能的实现中，接入网之外包括非实时RIC(可选的，非实时RIC可以位于OAM中或者核心网设备中)，用于进行模型训练和推理。例如，非实时RIC用于训练AI模型，利用该模型进行推理。例如，非实时RIC可以从CU、DU或RU中的至少一项获得网络侧和/或终端侧的信息，该信息可以作为训练数据或者推理数据，该推理结果可以被递交至CU、DU、RU或终端设备中的至少一项。可选的，CU和DU之间可以交互推理结果。可选的，DU和RU之间可以交互推理结果，例如非实时RIC将推理结果递交至DU，由DU转发给RU。
如图4B所示，第三种可能的实现中，接入网设备中包括近实时RIC，接入网设备之外包括非实时RIC(可选的，非实时RIC可以位于OAM中或者核心网设备中)。同上述第二种可能的实现，非实时RIC可以用于进行模型训练和推理。和/或，同上述第一种可能的实现，近实时RIC可以用于进行模型训练和推理。和/或，非实时RIC进行模型训练，近实时RIC可以从非实时RIC获得AI模型信息，并从CU、DU或RU中的至少一项获得网络侧和/或终端侧的信息，利用该信息和该AI模型信息得到推理结果。可选的，近实时RIC可以将推理结果递交至CU、DU、RU或终端设备中的至少一项。可选的，CU和DU之间可以交互推理结果。可选的，DU和RU之间可以交互推理结果，例如近实时RIC将推理结果递交至DU，由DU转发给RU。例如，近实时RIC用于训练模型A，利用模型A进行推理。例如，非实时RIC用于训练模型B，利用模型B进行推理。例如，非实时RIC用于训练模型C，将模型C的信息发送给近实时RIC，近实时RIC利用模型C进行推理。
图4C所示为本公开提供的方法能够应用的一种网络架构的示例图。相对图4B，图4B中将CU分离为了CU-CP和CU-UP。
图4D所示为本公开提供的方法能够应用的一种网络架构的示例图。如图4D所示，可选的，接入网设备中包括一个或多个AI实体，该AI实体的功能类似上述近实时RIC。可选的，OAM中包括一个或多个AI实体，该AI实体的功能类似上述非实时RIC。可选的，核心网设备中包括一个或多个AI实体，该AI实体的功能类似上述非实时RIC。 当OAM和核心网设备中都包括AI实体时，他们各自的AI实体所训练得到的模型不同，和/或用于进行推理的模型不同。
本公开中，模型不同包括以下至少一项不同：模型的结构参数(例如神经网络层数、神经网络宽度、层间的连接关系、神经元的权值、神经元的激活函数、或激活函数中的偏置中的至少一项)、模型的输入参数(例如输入参数的类型和/或输入参数的维度)、或模型的输出参数(例如输出参数的类型和/或输出参数的维度)。
图4E所示为本公开提供的方法能够应用的一种网络架构的示例图。相对图4D，图4E中的接入网设备分离为CU和DU。可选的，CU中可以包括AI实体，该AI实体的功能类似上述近实时RIC。可选的，DU中可以包括AI实体，该AI实体的功能类似上述近实时RIC。当CU和DU中都包括AI实体时，他们各自的AI实体所训练得到的模型不同，和/或，用于进行推理的模型不同。可选的，还可以进一步将图4E中的CU拆分为CU-CP和CU-UP。可选的，CU-CP中可以部署有一个或多个AI模型。可选的，CU-UP中可以部署有一个或多个AI模型。
图4D或图4E中，接入网设备的OAM和核心网设备的OAM示出为统一部署。可替代地，如前文所述，图4D或图4E中，接入网设备的OAM和核心网设备的OAM可以分开独立部署。
本公开中，一个模型可以推理得到一个输出，该输出包括一个参数或者多个参数。不同模型的学习过程或训练过程可以部署在不同的设备或节点中，或者可以部署在相同的设备或节点中。不同模型的推理过程可以部署在不同的设备或节点中，或者可以部署在相同的设备或节点中。
可选的，上述AI模型包括AE网络中的解码器网络，在网络侧，解码器网络的推理结果例如用于下行信道矩阵的重构。可选的，上述AI模型包括AE网络中的编码器网络，该编码器网络的模型信息可以被发送给UE，用于UE进行推理。
需要说明的是，在上述图4A至图4E的框架中，AI模型可以简称为模型或网络模型等，其可以看做是从输入的参数(例如输入矩阵)到输出的参数(例如输出矩阵)之间的映射。例如，对于网络侧的解码器网络，输入矩阵可以是根据接收的CSI确定的矩阵。训练数据可以包括已知的输入矩阵，或包括已知的输入矩阵和对应的输出矩阵，用于训练AI模型。训练数据可以是来自接入网设备、CU、CU-CP、CU-UP、DU、RU、UE和/或其它实体的数据，和/或是通过AI技术推理出的数据，不予限制。推理数据包括输入矩阵，用于利用模型推理出输出矩阵。推理数据可以是来自接入网设备、CU、CU-CP、CU-UP、DU、RU、UE和/或其它实体的数据。推理出的矩阵可以看做策略信息，发送给执行对象。推理出的矩阵可以被发送给接入网设备、CU、CU-CP、CU-UP、DU、RU、或UE等，用于进行进一步处理，例如用于下行信道矩阵的重构。
本公开中，在网络侧如果部署了AE网络中的解码器网络，则该解码器网络可以部署于接入网设备(如基站)中或者接入网设备之外，例如部署于OAM、AI网元或者核心网设备中，或者部署于RU、DU或近实时RIC中，不予限制。该解码器网络的推理结果可以由接入网设备进行推理得到，或者可以由非实时RIC进行推理后发送至接入网设备。为了简化描述，本公开以解码器网络部署于接入网设备中为例进行描述。
本公开中，在终端侧如果部署了AE网络中的编码器网络，则该编码器网络可部署于UE中，UE可以利用该编码器网络进行推理。
下面结合附图介绍本公开提供的方法。在这些方法中，所包括的步骤或操作仅是示例，本公开还可以执行其它操作或者各种操作的变形。此外，各个步骤可以按照本公开呈现的不同的顺序来执行，并且有可能并非要执行全部操作。
请参考图5，为本公开提供的一种通信方法的流程图。
S501、UE获得M份第一子下行信道数据，其中，每份第一子下行信道数据对应于M个数据空间中的一个数据空间。M为大于1的整数。
M份第一子下行信道数据例如是根据第一下行信道数据获得的。例如，UE可将第一下行信道数据划分到M个数据空间，或者理解为，UE可将第一下行信道数据划分为M份，从而得到M份第一子下行信道数据。每份第一子下行信道数据对应于一个数据空间，也可以理解为，数据空间与第一子下行信道数据是一一对应的关系。第一下行信道数据例如为原始的下行信道数据(或者称为，原始的下行信道矩阵或下行信道响应)，即，UE得到原始的下行信道数据后就可以直接将其划分为M份，无需再对原始的下行信道数据进行其他处理，由此能够减少处理步骤；或者，第一下行信道数据也可以是对第二下行信道数据进行预处理所得到的数据，第二下行信道数据是根据原始的下行信道矩阵得到的，通过预处理过程能够简化原始的下行信道数据，从而简化UE对第一下行信道数据的处理过程；或者，第一下行信道数据还可能是神经网络输出的数据，例如原始的下行信道矩阵等内容对于UE来说不可见，UE直接获得神经网络输出的第一下行信道数据即可。
如果第一下行信道数据是对第二下行信道数据进行预处理得到的，那么就涉及到预处理过程。而第二下行信道数据是根据原始的下行信道矩阵得到的，例如第二下行信道数据是原始的下行信道矩阵本身，或者，第二下行信道数据是对原始的下行信道矩阵进行处理后所得到的特征向量。针对第二下行信道数据的不同的实现方式，预处理过程可能有所不同，下面进行介绍。
1、第二下行信道数据是原始的下行信道矩阵。例如将原始的下行信道矩阵称为第一下行信道矩阵。
例如第一下行信道矩阵的维度为[N tx,N rx,N RB]，其中N tx表示下行信号的发送端(例如接入网设备)的天线数或端口数，N rx表示下行信号的接收端(例如UE)的天线数或端口数，N RB表示频域单元数，例如资源块(resource block，RB)的数量，或者子带的数量。
可选的，进一步，UE可将第一下行信道矩阵进行维度变换处理，得到变换后的数据，或者说得到变换的第一下行信道矩阵。变换后的第一下行信道矩阵的维度为[N tx*N rx,N RB]或[N txN rx,N RB]，例如将该矩阵用H表示，H为一个复数矩阵，
Figure PCTCN2022142946-appb-000001
可选的，第一下行信道数据例如为矩阵H。
可选的，进一步，通过离散傅里叶变换(discrete fourier transform，DFT)可产生两组DFT基底，分别是空域基底
Figure PCTCN2022142946-appb-000002
和频域基底
Figure PCTCN2022142946-appb-000003
其中，空域基底为N txN rx个N txN rx*1的DFT列向量，频域基底为N rb个N rb*1的DFT列向量。UE可根据空域基底和频域基底对降维的第一下行信道矩阵H进行空频联合投影，空频联合投影的一种方式可参考如下公式：
Figure PCTCN2022142946-appb-000004
得到了复数矩阵C complex，也就完成了对第二下行信道数据的预处理过程。其中，S H是S的埃尔米特(hermitian)阵，也称为自共轭矩阵，可通过将矩阵S进行共轭转置得到。N sb表示 频域子带个数，例如N sb＝N rb/a，a表示频域子带颗粒度或子带宽带，即每个子带包括的RB个数。常见的频域子带颗粒度为1RB、2RB、4RB或8RB等，这里不加限制。以频域子带颗粒度是4RB为例，则N sb＝N rb/4。S表示空域基底，其具体形态与天线面板有关，假设天线面板是双极化的，水平阵子为Nh，垂直阵子为Nv，则得到S的表现形式为：
Figure PCTCN2022142946-appb-000005
F表示频域基底，其表现形式与子带N sb有关，例如F可满足如下公式：
F＝DFT(N sb)      (公式3)
可选的，在DFT的过程中，还可以加入过采样因子，例如，可利用过采样方式产生多组正交空域基底{S 1,S 2,S 3…}和多组正交频域基底{F 1,F 2,F 3…}，从中挑选一组S i和F j作为本公开的空域基底和频域基底，例如可从中挑选投影方向较为准确的一组。例如空域和频域的过采样因子均为4。
可选的，第一下行信道数据例如为第二下行信道数据经预处理后得到的复数矩阵，例如复数矩阵C complex。
2、第二下行信道数据是对第一下行信道矩阵进行处理后得到的特征向量。
在这种情况下，需要先对第一下行信道矩阵进行处理，得到特征向量，再对特征向量进行预处理，得到第一下行信道数据。或者也可以理解为，对第一下行信道矩阵进行处理以得到特征向量的过程，以及对特征向量进行预处理以得到第一下行信道数据的过程，都视为对第一下行信道矩阵的预处理过程。
例如第一下行信道矩阵的维度为[N tx,N rx,N RB]，通过奇异值分解(singular value decomposition，SVD)将[N tx,N rx,N RB]维的第一下行信道矩阵进行降维，得到下行信道的特征子空间矩阵，或简称为特征子空间，该特征子空间的维度是[N tx,N sb]。UE在通过SVD将第一下行信道矩阵进行降维时，可对第一下行信道矩阵的不同的秩(rank)分别进行处理，其中，不同的rank，也可理解为不同的流，或不同的层(layer)。一个信道信息(或者说，一个信道估计结果)可对应一个或多个层，下面介绍UE对第一下行信道矩阵的第L层的处理过程，该方法可以有多种，不加以限制。
第L层的每个子带内可包含a个RB，则UE可以综合a个RB的下行信道来计算一个子带内的等效下行信道。假设第L层的子带c内的第k个RB对应的下行信道表示为H k，则该子带c内的等效下行信道可表示为：
Figure PCTCN2022142946-appb-000006
UE对
Figure PCTCN2022142946-appb-000007
进行SVD分解，可得到：
Figure PCTCN2022142946-appb-000008
即，
Figure PCTCN2022142946-appb-000009
其中，H k的维度为[N tx*N rx]，
Figure PCTCN2022142946-appb-000010
的维度为[N tx*N tx]。矩阵
Figure PCTCN2022142946-appb-000011
的第k列即可作为该子带c对应的第L层特征向量(为了避免混淆，将子带对应的特征向量称为子特征向量)，其维度为[N tx*1]，即，第L层的第c个子带的子特征向量
Figure PCTCN2022142946-appb-000012
按照类似方式可得到第L层在每个子带上的子特征向量，将这些子特征向量进行拼接，就得到了特征向量，该特征向量就可作为本公开中的输入数据。例如该特征向量可表示为V＝[V 1V 2…V a]。可选的，第一下行信道数据例如为该特征向量V，其维度为[N rx,N RB]。
可选的，进一步，假设该特征向量
Figure PCTCN2022142946-appb-000013
是一个复数矩阵，通过DFT可产生两组 DFT基底，分别是空域基底
Figure PCTCN2022142946-appb-000014
和频域基底
Figure PCTCN2022142946-appb-000015
其中，空域基底为N tx个N tx*1的DFT列向量，频域基底为N sb个N sb*1的DFT列向量。UE可根据空域基底和频域基底对降维的下行信道矩阵H进行空频联合投影，空频联合投影的一种方式可参考如下公式：
Figure PCTCN2022142946-appb-000016
得到的复数矩阵C complex是原始的下行信道的特征子空间的稀疏化表征，其维度与空频联合投影前的特征向量的维度保持一致，是N tx*N sb。得到了复数矩阵C complex，也就完成了对第二下行信道数据的预处理过程。关于S H、N sb、以及空域基底S等参数的介绍可参考上文。
可选的，在DFT的过程中，也可以加入过采样因子，例如，可利用过采样方式产生多组正交空域基底{S 1,S 2,S 3…}和多组正交频域基底{F 1,F 2,F 3…}，从中挑选一组S i和F j作为本公开的空域基底和频域基底，例如可从中挑选投影方向较为准确的一组。例如空域和频域的过采样因子均为4。UE通过如上两种方式中的任一种得到复数矩阵C complex后，可根据复数矩阵C complex得到第一下行信道数据。可选的，UE根据复数矩阵C complex得到第一下行信道数据的一种方式为，UE直接将复数矩阵C complex作为第一下行信道数据，即，第一下行信道数据为对第二下行信道数据进行预处理的结果。
或者，UE根据复数矩阵C complex得到第一下行信道数据的另一种方式为，UE可以从复数矩阵C complex中选取部分数据作为第一下行信道数据。例如，在频域方向(时延域)上，能量一般都主要集中在时延＝0的附近，除此之外其他区域的能量基本可以忽略不计。因此，UE选择时延＝0的两侧的连续的F列作为第一下行信道数据即可，剩余的部分可以默认系数为0。例如，UE可以从复数矩阵C complex中选择连续的F列作为第一下行信道数据，对于复数矩阵C complex中未被选择的列可以不做处理。由此既考虑到了能量分布的情况，又能够减小处理开销。
F例如为正整数，F的取值可通过协议预定义，或者根据不同的开销可确定不同的F，例如协议可提供开销与F之间的映射关系，从而UE和接入网设备根据当前的开销需求就能确定同样的F。或者，F的取值也可由接入网设备指示，例如接入网设备向UE发送用于指示F的取值的信息，UE接收该信息后就能确定F的取值。或者，F的取值也可由UE确定，例如UE根据信道状态和/或网络形态等因素来确定F的取值，以减小对空口传输的影响。UE确定F的取值后可向接入网设备发送用于指示F的取值的信息，接入网设备接收该信息后就能确定F的取值。
或者，UE根据复数矩阵C complex得到第一下行信道数据的再一种方式为，UE也可将复数矩阵C complex进行压缩处理，得到压缩信息，该压缩信息可作为第一下行信道数据。例如，UE可将复数矩阵C complex输入编码器网络，编码器网络对该复数矩阵C complex进行压缩处理，编码器网络输出的就是压缩信息。在这种方式下，第一下行信道数据是经过压 缩得到的，能够减小UE处理第一下行信道数据时的复杂度。
前述的过程是为了获得第一下行信道数据，在得到第一下行信道数据后，UE可将第一下行信道数据划分到M个数据空间中，从而得到M份第一子下行信道数据。其中，第一子下行信道数据与数据空间是一一对应的关系，例如，M份第一子下行信道数据中的第i份第一子下行信道数据对应M个数据空间中的第i个数据空间，i可以取从1至M的整数。
本公开涉及M个数据空间，M个数据空间可以与字典对应，例如，M个数据空间可对应N个字典，N为大于或等于1且小于或等于M的整数。可选的，N＝M，即数据空间与字典一一对应，每个数据空间对应一个字典；或者，N＝1，即，M个数据空间均对应同一个字典，可认为该字典与每个数据空间均对应。可选的，N＝M/2，其中，每2个数据空间对应一个字典。其他可能的情况不再一一举例。不同的数据空间对应的字典可以相同，也可以不同，不予限制。在下文的S502中将介绍字典的用途。另外，M个数据空间(或者说，M个数据空间的划分方式)在训练字典的过程中也会涉及，关于字典的训练过程将在后续的实施例中介绍，因此关于M个数据空间的划分方式等也将在后续实施例中一并介绍。
一个字典中保存的变量至少包括{数据空间的索引，元素的索引，元素}中的一个，也就是说，一个字典中保存的变量可包括数据空间的索引、元素的索引、或元素中的一项或多项，除此之外，字典中还可包括其他信息，或者也可以不再包括其他信息，对此不做限制。其中，一个字典所包括的数据空间的索引，是该字典所对应的数据空间的索引。例如字典与数据空间一一对应，则一个字典对应一个数据空间，一个字典中就包括该字典所对应的数据空间的索引；或者，例如M个数据空间均对应同一个字典，则该字典对应M个数据空间，该字典中可不包括数据空间索引。元素例如为矢量，一个字典可包括多个元素。每个元素可以有对应的索引，即，元素与元素的索引可以是一一对应的关系。如果N大于1，则不同的字典包括的元素的索引可以复用，例如每个字典中的元素的索引都可以从1开始或从0开始，即不同的字典包括的元素独立进行编号；或者，不同的字典包括的元素的索引也可以不同，即不同的字典包括的元素进行联合编号，例如第一个字典的元素的索引为从0～d-1，第二个字典的元素的索引就从d开始。可参考图6，为N个字典的一种示意图，图6以N＝M为例，即，共包括M个字典。图6中每个字典内的0～3，表示的是元素的索引，这里以每个字典包括的元素的索引数量均是4为例，实际上不限于此。另外，不同的字典所包括的元素的个数可以相同，也可以不同。
可选的，关于字典的表达方式还包括，如果N位于1和M之间，则一个字典中可包含{字典索引，元素的索引，元素}，接入网设备和终端设备之间可互通字典索引与数据空间的索引之间的对应关系。如果N＝1，则字典索引可以省略，默认所有的数据空间的索引都对应该字典。如果M>N>1，则字典索引与数据空间的索引之间的对应关系可以是协议预定义的一种默认的规则，例如M＝4，N＝2，在该规则中，数据空间的索引0和数据空间的索引2对应于字典索引0，数据空间的索引1和数据空间的索引3对应于字典索引1；或者，如果M>N>1，接入网设备也可以向UE指示字典索引与数据空间的索引之间的对应关系；或者，如果M>N>1，UE也可以向接入网设备上报字典索引与数据空间的索引之间的对应关系。如果M＝N，则字典索引与数据空间的索引可以一一对应，或者，字典中可包含{数据空间的索引，元素的索引，元素}。
S502、UE根据M个数据空间中的第i个数据空间对应的第一字典，确定M份第一子下行信道数据中的第i份第一子下行信道数据对应的第一信息。i取从1至M的整数，因此UE共确定M个第一信息。
第i份第一子下行信道数据对应于M个数据空间中的第i个数据空间，例如将第一下行信道数据划分到M个数据空间中就会得到M份第一子下行信道数据，其中的第i份第一子下行信道数据就是第一下行信道数据划分到第i个数据空间中的部分。例如N＝M，每个数据空间分别有各自对应的字典，则第一字典例如为M个数据空间中的第i个数据空间所对应的字典，即，UE可根据第i个数据空间所对应的第一字典确定第i份第一子下行信道数据对应的第一信息。其中，如果i取从1至M的整数，则不同的数据空间对应的字典都可以称为第一字典，但不同的数据空间对应的第一字典可能相同，也可能不同。或者，N＝1，一个字典对应M个数据空间，则第一字典即为该字典，对于M个数据空间中的任一个数据空间，均使用第一字典，则UE可根据第一字典确定第i份第一子下行信道数据对应的第一信息。
其中，如果N＝1，则UE可确定该字典为第i个数据空间对应的第一字典。或者，如果M＝N，字典与数据空间一一对应，则UE能够确定第i个数据空间对应的第一字典。或者，如果M>N>1，则UE可以根据字典索引与数据空间的索引之间的对应关系确定第i个数据空间对应的第一字典。例如M＝4，N＝2，该对应关系规定，数据空间的索引0和数据空间的索引2对应于字典索引0，数据空间的索引1和数据空间的索引3对应于字典索引1，i就相当于数据空间的索引，那么UE可以根据i的取值以及该对应关系，确定第i个数据空间对应的第一字典。例如i＝1，则UE可以确定第1个数据空间对应的第一字典为字典索引1所指示的字典。
根据前文对于字典的介绍可知，第一字典可包括多个元素，UE可以从中确定第i份第一子下行信道数据所对应的P个元素，P为正整数。例如在第一字典包括的多个元素中，与第i份子数据最相关的P个元素即为第i份第一子下行信道数据对应的P个元素，这P个元素就可作为第i份第一子下行信道数据对应的第一信息。其中，如果P大于1，则P个元素可以按照第一组合方式构成第一信息，例如第一组合方式为将P个元素相乘，或者第一组合方式为将P个元素进行加权求和(例如求平均，或者利用其它可能的权值进行加权求和)，或者将P个元素串联等，对于第一组合方式不做限制。第一组合方式例如通过协议预定义，或者由接入网设备确定并告知UE，或者由UE确定并告知接入网设备等。对于M份第一子下行信道数据，UE均可以确定其对应的第一信息，则UE共可以确定M个第一信息，M个第一信息即为M个元素。
S503、UE发送第一指示信息。例如，UE向接入网设备发送第一指示信息，那么相应的，接入网设备可从UE接收第一指示信息。第一指示信息可指示M个第一信息，接入网设备根据第一指示信息就能确定M个第一信息。
可选的，第一指示信息包括M个第一信息的标识，从而可以指示M个第一信息。一个第一信息的标识例如为该第一信息在对应字典中的索引，例如，M个第一信息包括第i份第一子下行信道数据对应的第一信息，该第一信息的标识即为该第一信息在第一字典中的索引。UE在确定M个第一信息后就可以确定M个第一信息的标识，例如UE共可以确定M个标识，UE可将这M个标识发送给接入网设备。UE发送M个第一信息的标识，就可以视为UE发送了CSI，即，M个第一信息的标识可以作为CSI；或者，M个第一信息 的标识也可以作为PMI；或者，M个第一信息的标识能够实现与PMI或CSI类似的功能。
或者，第一指示信息也可以不包括M个第一信息的标识，而是通过其他方式来指示第一信息。例如，字典元素之间有不同的组合关系，每种组合关系中可以包括N个字典中的每个字典里的一个元素。每种组合关系可对应一个指示信息，如果UE发送了某个指示信息，就表明指示该指示信息所对应的组合关系。例如第一指示信息与M个第一信息的组合关系对应，那么UE发送第一指示信息就可以指示M个第一信息。
可选的，UE在发送M个第一信息的标识时，可以按照第一顺序发送，第一顺序为M个数据空间的排列顺序，即，第一顺序规定了UE先发送哪个数据空间对应的第一信息的标识，后发送哪个数据空间对应的第一信息的标识。例如M＝4，M个数据空间分别为数据空间1～数据空间4，第一顺序为2-1-4-3，则UE在发送M个第一信息的标识时，最先发送数据空间2对应的第一信息的标识，接着发送数据空间1对应的第一信息的标识，然后发送数据空间4对应的第一信息的标识，之后再发送数据空间3对应的第一信息的标识。对于UE和接入网设备来说，第一顺序都是已知的，因此接入网设备接收M个第一信息的标识后，也能明确第一信息的标识与数据空间之间的对应关系，从而避免对应出错。
例如，第一顺序为协议预定义的顺序，UE和接入网设备根据协议就能确定第一顺序。或者，第一顺序也可以预配置在UE和接入网设备中。或者，第一顺序可由接入网设备确定，接入网设备确定第一顺序后可以向UE发送第二指示信息，第二指示信息用于指示第一顺序，UE根据第二指示信息就能确定第一顺序。或者，第一顺序可由UE确定，UE确定第一顺序后可以向接入网设备发送第三指示信息，第三指示信息用于指示第一顺序，接入网设备根据第三指示信息就能确定第一顺序。
S504、对于M个第一信息中的第i个第一信息，接入网设备根据M个数据空间中的第i个数据空间对应的第一字典，恢复得到第i份第二子下行信道数据。i取从1至M的整数，则接入网设备共可以得到M份第二子下行信道数据。
例如接入网设备按照第一顺序接收了M个第一信息的标识，则接入网设备能够明确第一信息的标识与数据空间之间的对应关系，从而接入网设备可以根据数据空间所对应的字典确定第一信息的标识所对应的第一信息，而接入网设备所确定的第一信息，就视为接入网设备恢复出的第二子下行信道数据。例如N＝M，数据空间与字典一一对应，第i个数据空间对应的字典例如为第一字典，则对于第i个第一信息的标识，接入网设备可以在第一字典中确定第i个第一信息的标识，从而确定第i个第一信息的标识在第一字典中对应的第一信息，即，恢复出第i个第一信息对应的第二子下行信道数据(即，第i份第二子下行信道数据)。又例如，N＝1，M个数据空间均对应第一字典，则对于第i个第一信息的标识，接入网设备可以在第一字典中确定第i个第一信息的标识，从而确定第i个第一信息的标识在第一字典中对应的第一信息，即，恢复出第i个第一信息对应的第二子下行信道数据(即，第i份第二子下行信道数据)。再例如，M>N>1，则对于第i个第一信息的标识，接入网设备根据第一顺序能够明确第i个第一信息的标识对应的数据空间，例如为第i个数据空间。接入网设备进一步根据数据空间的索引与字典索引之间的对应关系，可以确定第i个数据空间对应的字典，例如第一字典。则接入网设备可以确定第i个第一信息的标识在第一字典中对应的第一信息，即，恢复出第i个第一信息对应的第二子下行信道数据(即，第i份第二子下行信道数据)。
在理想状态下，接入网设备所得到的M份第二子下行信道数据，与UE获得的M份 第一子下行信道数据，可以是相同的数据。例如，第i份第一子下行信道数据与第i份第二子下行信道数据是相同的数据。在实际应用中，接入网设备所得到的M份第二子下行信道数据，与UE获得的M份第一子下行信道数据，这之间可能会出现偏差。UE根据字典得到第一信息的过程相当于是对M份第一子下行信道数据进行量化，即，UE向接入网设备发送的是量化信息，接入网设备是根据量化信息和字典恢复出了M份第二子下行信道数据，在量化和恢复的过程中可能会有一些损失，因此可能导致M份第二子下行信道数据与M份第一子下行信道数据之间有一定的偏差。例如，第i份第一子下行信道数据与第i份第二子下行信道数据可能是不同的数据。但随着字典精度的提高以及传输质量的提高等，可以使得M份第二子下行信道数据与M份第一子下行信道数据之间的偏差趋于减小。
S505、接入网设备根据M份第二子下行信道数据，恢复得到下行信道信息。或者说，接入网设备根据M份第二子下行信道数据，重构下行信道矩阵，例如重构第一下行信道矩阵。
如果在S501中，UE是将复数矩阵C complex作为第一下行信道数据，或者是从复数矩阵C complex中选取了连续的F列作为第一下行信道数据，则接入网设备得到M份第二子下行信道数据后，可将M份第二子下行信道数据拼接，得到的信息例如称为角度时延域系数，角度时延域系数为一个矩阵，可表示为
Figure PCTCN2022142946-appb-000017
或者，如果在S501中，UE是将复数矩阵C complex进行压缩处理得到了压缩信息，并将该压缩信息作为了第一下行信道数据，则接入网设备所得到的M份第二子下行信道数据实际上是M个子压缩信息。可选的，接入网设备可对M个子压缩信息进行恢复，得到K个恢复信息，K为正整数，K可以等于M，也可以不等于M。例如，UE是根据编码器网络得到了压缩信息，则接入网设备侧可设置与该编码器网络对应的解码器网络，接入网设备可将M份第二子下行信道数据输入该解码器网络，该解码器网络就可输出K个恢复信息。接入网设备可将K个恢复信息进行拼接，得到角度时延域系数为一个矩阵，可表示为
Figure PCTCN2022142946-appb-000018
可参考图7，为UE将压缩信息作为第一下行信道数据、接入网设备需恢复压缩信息的示意图。在图7中，例如UE将第二下行信道数据输入编码器网络，编码器网络对第二下行信道数据进行压缩，编码器网络输出压缩信息，该压缩信息可作为第一下行信道数据。或者在图7中，UE也可以将复数矩阵C complex输入编码器网络，编码器网复数矩阵C complex进行压缩，编码器网络输出压缩信息，该压缩信息可作为第一下行信道数据。UE将该压缩信息划分到M个数据空间中，图7以M＝4为例，则UE得到4份第一子下行信道数据。UE通过4个字典处理4份第一子下行信道数据，得到4个第一信息的标识，图7中的圆圈表示字典，C M表示第M个字典包括的元素的个数，log 2C M表示第M个字典对应的传输比特数。可选的，在图7中，字典对应的传输比特数可以通过向上取整得到，例如图7中的log 2C M也可替换为
Figure PCTCN2022142946-appb-000019
log 2C 1也可替换为
Figure PCTCN2022142946-appb-000020
等。或者，图7中，字典对应 的传输比特数还可以通过向下取整得到，具体不做限制。其中，一个第一信息为一份第一子下行信道数据在相应的字典中对应的元素。图7以N＝M为例。UE向接入网设备发送4个第一信息的标识，接入网设备接收4个第一信息的标识后，可根据4个字典恢复出4个子压缩信息，接入网设备将4个子压缩信息进行拼接等处理，再将得到的信息输入解码器网络，以得到恢复信息，接入网设备根据恢复信息可恢复出下行信道信息。其中，编码器网络在对复数矩阵C complex进行压缩时，可能需要用到码本，相应的，解码器网络在恢复压缩信息时，也可能需要用到码本。或者，码本也可以称为字典，但该字典与本公开所述的N个字典是不同的。
无论通过如上哪种方式得到角度时延域系数，接入网设备都可根据角度时延域系数
Figure PCTCN2022142946-appb-000021
恢复得到下行信道信息。在理想状态下，恢复得到的下行信道信息与第一下行信道矩阵可以是相同的信息。
例如，如果第一下行信道数据为第一下行信道矩阵，则
Figure PCTCN2022142946-appb-000022
接入网设备将
Figure PCTCN2022142946-appb-000023
进行逆变换，可获得恢复的下行信道(或者说，重构的下行信道)。例如，接入网设备将
Figure PCTCN2022142946-appb-000024
进行逆变换的一种方式如下：
Figure PCTCN2022142946-appb-000025
Figure PCTCN2022142946-appb-000026
就表示恢复得到的下行信道，公式7中的
Figure PCTCN2022142946-appb-000027
的维度为N txN rx*N RB。例如，可将
Figure PCTCN2022142946-appb-000028
直接作为恢复得到的下行信道信息，或者也可以通过矩阵变换等方式，将
Figure PCTCN2022142946-appb-000029
的维度转换为与第一下行信道矩阵的维度相同的维度，维度转换后得到的即为恢复的下行信道信息。
又例如，如果第一下行信道数据为根据第一下行信道矩阵得到的特征向量，则
Figure PCTCN2022142946-appb-000030
接入网设备将
Figure PCTCN2022142946-appb-000031
进行逆变换，可获得恢复的下行信道的特征子空间。例如，接入网设备将
Figure PCTCN2022142946-appb-000032
进行逆变换的一种方式如下：
Figure PCTCN2022142946-appb-000033
Figure PCTCN2022142946-appb-000034
就表示恢复得到的下行信道的特征子空间，公式8中的
Figure PCTCN2022142946-appb-000035
的维度为N tx*N rb。例如，可将
Figure PCTCN2022142946-appb-000036
直接作为恢复得到的下行信道信息。
在本公开中，UE可将第一下行信道数据划分到M个数据空间中，根据不同的数据空间对应的字典可以确定各份第一子下行信道数据对应的第一信息，不同的数据空间能够表征不同的位置信息，或者说能够表征不同的信道环境信息，UE反馈不同的数据空间对应的第一信息，可以使得接入网设备能够明确第一信息与环境信息之间的对应关系，从而使得UE所反馈的第一信息能够反映实际的通信环境，提高了UE所反馈的第一信息的准确性。接入网设备根据UE所反馈的第一信息，能够恢复得到较为准确的下行信道。
图5所示的实施例介绍的是网络推理的过程。在网络推理的过程中涉及到了字典，字典可以通过网络训练得到。其中，用于训练得到字典的方式可能有多种。例如，如果UE侧不设置编码器网络，接入网设备侧不设置解码器网络，或者，即使UE侧设置了编码器 网络，接入网设备侧设置了解码器网络，编码器网络和解码器网络可以与字典共同训练，也可以不与字典共同训练。如果只需通过训练得到字典，而无需得到编解码网络，可参考接下来介绍的本公开的另一种通信方法，通过该方法介绍一种网络训练的过程，通过该训练过程可得到字典。请参考图8，为该方法的流程图。
S801、第一节点获得M份第三子下行信道数据。其中，每份第三子下行信道数据对应于M个数据空间中的一个数据空间。本公开中的M个数据空间，与图5所示的实施例中所述的M个数据空间，可以是相同的特征。
M份第三子下行信道数据例如是根据第三下行信道数据获得的。例如，UE或第一节点可将第三下行信道数据划分到M个数据空间，或者理解为，UE或第一节点可将第三下行信道数据划分为M份，从而得到M份第三子下行信道数据。第三下行信道数据例如为原始的下行信道数据，例如将本实施例中的原始的下行信道数据称为第三下行信道矩阵；或者，第三下行信道数据也可以是对第四下行信道数据进行预处理所得到的数据，第四下行信道数据是根据第三下行信道矩阵得到的；或者，第三下行信道数据还可能是神经网络输出的数据。其中，第三下行信道矩阵可视为训练数据，或者称为训练样本。例如，可认为第三下行信道矩阵包括一个或多个训练数据，例如第三下行信道矩阵实际上包括一个或多个子下行信道矩阵，其中的一个子下行信道矩阵可视为一个训练数据。这里的第三子下行信道矩阵之间可能是彼此独立的，并不是包括在一个大的矩阵中，即，并不将第三下行信道矩阵视为大的矩阵，第三下行信道矩阵可理解为对一个或多个第三子下行信道矩阵的统称。
如果第三下行信道数据是对第四下行信道数据进行预处理得到的，那么就涉及到预处理过程。关于第四下行信道数据的预处理过程，可参考图5所示的实施例的S501中对于第二下行信道数据的预处理过程的介绍。
本公开中，第一节点例如为UE，或者为接入网设备，或也可以是第三方设备(例如AI节点等)等。训练过程可以是在线训练也可以是离线训练。
在得到第三下行信道数据后，第一节点可将第三下行信道数据划分到M个数据空间中，得到M份第三子下行信道数据。其中，第三子下行信道数据与数据空间是一一对应的关系，例如，M份第三子下行信道数据中的第i份第三子下行信道数据对应M个数据空间中的第i个数据空间，i可以取从1至M的整数。
第一节点要将第三下行信道数据划分到M个数据空间中，首先需要确定M个数据空间，或者说确定M个数据空间的划分方式。以第一节点是UE或接入网设备为例。例如，M个数据空间的划分方式通过协议预定义，则UE和接入网设备都可以根据协议确定M个数据空间的划分方式。或者，M个数据空间的划分方式由接入网设备确定，接入网设备可向UE发送第四指示信息，第四指示信息可指示M个数据空间的划分方式，UE根据第四指示信息就能确定M个数据空间的划分方式。或者，M个数据空间的划分方式可由UE确定，UE可向接入网设备发送第五指示信息，第五指示信息可指示M个数据空间的划分方式，接入网设备根据第五指示信息可确定M个数据空间的划分方式。
本公开中，例如协议规定M个数据空间的一种划分方式为，M＝4，这4个数据空间分别包括某个数据的4个部分，这4个部分分别为该数据包括的极化1+实部的部分、该数据包括的极化1+虚部的部分、该数据包括的极化2+实部的部分、该数据包括的极化2+虚部的部分。例如，将第三下行信道数据划分到这4个数据空间中，则划分得到的4份第三子 下行信道数据分别包括第三下行信道数据包括的极化1+实部的部分、第三下行信道数据包括的极化1+虚部的部分、第三下行信道数据包括的极化2+实部的部分、第三下行信道数据包括的极化2+虚部的部分。从天线形态来看，天线阵子为双极化，极化1和极化2就表示两种极化方向，两种极化方向之间可以认为相互独立；从复数的性质来看，数据包括实部和虚部，实部和虚部的处理过程也是相对独立的。因此，可按照天线极化方向与复数的实虚部来划分数据空间，使得各个数据空间可以独立处理。通过划分数据空间，每个数据空间的大小是原始数据的1/M，不同的数据空间能够表征不同的环境信息。可选地，数据空间的划分方式也可以是非等分的方式，不予限制。
可选的，在网络推理阶段，UE和接入网设备也可以确定M个数据空间的划分方式，确定方式与本公开是类似的，或者，第一节点也可以将M个数据空间的划分方式指示给UE和/或接入网设备。即，图5所示的实施例中，UE要将第一下行信道数据划分到M个数据空间中，也需要先确定M个数据空间的划分方式，则可以利用本公开所提供的M个数据空间的划分方式。其中，在网络推理阶段和网络训练阶段，所应用的M个数据空间的划分方式是一致的。
S802、第一节点进行聚类(clustering)训练，以得到N个字典。
在图5所示的实施例中介绍了，N可能等于M，也可能等于1，还可能M>N>1，对于这几种方案，第一节点的训练过程可能有所不同，下面分别介绍。
1、N＝M，即，数据空间与字典一一对应。
第一节点可根据M份第三子下行信道数据中的第i份第三子下行信道数据进行聚类训练，以得到对应于第i个数据空间的字典(例如第一字典)，第i份第三子下行信道数据对应于第i个数据空间。也就是说，第一节点可以在每个数据空间内分别进行训练，从而得到每个数据空间对应的字典，共可以得到M个字典。聚类是按照某种特定标准(例如距离)，将一个数据集分割成不同的类或簇，使得同一个簇内的数据对象的相似性尽可能大，且不在同一个簇中的数据对象的差异性也尽可能地大。即，聚类后可以使得同一类的数据尽可能聚集到一起，不同类的数据尽量分离。每一类数据都有一个类中心值。如果本公开采用聚类的方式进行网络模型训练，则字典所包括的元素也可以称为类中心值。
在一个数据空间内训练，是为了得到该数据空间对应的元素，这些元素就可以作为该数据空间对应的字典所包括的元素。
一个数据空间对应的字典所包括的元素数量，可以与该数据空间对应的比特开销有关。例如一种比特开销为48比特(bit)，这是M个数据空间对应的总的传输开销。例如各个数据空间对应的比特开销均相等，假设共4个数据空间，其中的每个数据空间对应的传输开销，就是总的比特开销的1/M，那么一个数据空间对应的比特开销为12bits。12bits最多可以承载2 12个标识，因此第一字典所包括的元素的个数需要小于或等于2 12。可见，不同的比特开销所对应的元素的个数不同。可选的，第一节点可以根据不同的比特开销，分别训练不同的字典，比特开销与字典可以一一对应。也就是说，第一节点为一个数据空间可以训练一个或多个字典，如果训练多个字典，则多个字典可以对应不同的比特开销，从而UE在进行网络推理时可以根据当前的比特开销选择合适的字典。
例如，在网络推理过程中，数据空间对应的比特开销一般是由网络侧决定，例如由接入网设备确定，接入网设备可以根据与UE之间的实时的信道情况来决定。则UE在进行图5所示的实施例所述的网络推理过程时，接入网设备可先向UE下发信息，以确定本次 传输的比特开销，该信息可指示M个数据空间对应的总的比特开销，或者也可确定一个数据空间对应的比特开销，UE根据接入网设备指示的比特开销就可以选择合适的字典。例如接入网设备指示总的比特开销为48bits，另外M＝4，则UE可确定每个数据空间的传输开销为48/4＝12bits，例如UE在确定第i个数据空间对应的字典时，如果第i个数据空间对应多个字典(不同的字典对应不同的比特开销)，那么UE可以从中选择与12bits对应的字典进行网络推理。
另外，第一节点还可以根据第i份第三子下行信道数据的维度，确定第一字典所包括的元素的维度。第一字典所包括的元素的维度，也可以认为是第一字典的深度，其与用于训练第一字典的子下行信道数据的维度相关。因此第一节点根据第i份第三子下行信道数据的维度可以确定第一字典所包括的元素的维度。例如，第一节点可将第i份第三子下行信道数据转换为矢量，该矢量的长度就是第一字典所包括的元素的维度。例如，第i份第三子下行信道数据为第三下行信道数据所包括的极化1+实部的部分，这部分例如为维度是[16,13]的矩阵，第一节点可将该矩阵转换为长度为16*13的矢量，那么第一字典所包括的元素的维度就是16*13。
其中，第一节点要将一个矩阵转换为一个矢量，可以按行进行转换，也可以按列进行转换，在网络推理过程中，UE也需要执行该转换过程。为了使得接入网设备能够恢复得到较为准确的下行信道信息，UE的转换顺序需要UE和接入网设备都知晓。例如，UE的转换顺序可通过协议预定义，或者UE的转换顺序可由接入网设备确定并告知UE，或者UE的转换顺序可由UE确定并告知接入网设备。而网络推理过程中的转换顺序与网络训练过程中的转换顺序可以是一致的。
在确定第一字典所包括的元素的数量以及第一字典所包括的元素的维度后，第一节点就可以根据第i份第三子下行信道数据进行聚类训练，从而得到第一字典。对于每个数据空间，第一节点都可以按照类似方式训练，从而可得到M个字典。
2、N＝1，即，M个数据空间对应一个字典，例如该字典称为第一字典。
第一节点可根据M份第三子下行信道数据进行聚类训练，以得到对应于M个数据空间的字典(例如第一字典)。也就是说，第一节点得到M份第三子下行信道数据后，可以统一训练，从而得到一个字典，该字典就对应M个数据空间中的每个数据空间。利用M份第三子下行信道数据训练一个字典，对于该字典来说相当于采样数据(或者说训练数据)就增加了M倍，训练数据更为丰富，使得该字典所包括的元素更为丰富细致，有利于接入网设备恢复得到更为准确的下行信道信息。
利用M份第三子下行信道数据进行训练，是为了得到M个数据空间对应的元素，这些元素就可以作为训练得到的一个字典所包括的元素。
可选的，第一节点可以根据不同的比特开销，分别训练不同的字典，比特开销与字典可以一一对应。也就是说，第一节点可以训练一个或多个字典，如果训练多个字典，则多个字典可以对应不同的比特开销，从而UE在进行网络推理时可以根据当前的比特开销选择合适的字典。
另外，第一节点还可以根据第三下行信道数据的维度，确定第一字典所包括的元素的维度。确定方式可参考前文。第一字典例如为第一节点训练得到的其中一个字典。
在确定第一字典所包括的元素的数量以及第一字典所包括的元素的维度后，第一节点就可以根据M份第三子下行信道数据进行聚类训练，从而得到第一字典。
或者，在N＝1的情况下，第一节点也可以对于每个数据空间分别进行训练，例如第一节点采用N＝M时的训练方式，但在训练时需要附加一个条件，即，在每个数据空间内训练得到的字典均相同。这样，虽然第一节点可以训练得到N个字典，但N个字典都是相同的，相当于第一节点还是得到了一个字典。
3、M>N>1，即，M个数据空间对应N个字典。
第一节点可根据M份第三子下行信道数据中的至少一份第三子下行信道数据进行聚类训练，以得到N个字典中的一个字典(例如第一字典)。例如第三子下行信道数据与数据空间一一对应，则至少一份第三子下行信道数据对应至少一个数据空间。一个字典可对应一个或多个数据空间，例如字典索引与数据空间的索引之间具有对应关系，那么第一节点在训练一个字典时，就根据该字典对应的数据空间内的第三子下行信道数据进行训练。
可选的，第一节点可以根据不同的比特开销，分别训练不同的字典，比特开销与字典可以一一对应。也就是说，一个字典对应一个或多个数据空间，那么对于这一个或多个数据空间来说，第一节点可根据不同的比特开销训练一个或多个字典，如果训练多个字典，则多个字典可以对应不同的比特开销，从而UE在进行网络推理时可以根据当前的比特开销选择合适的字典。
另外，第一节点还可以根据第三下行信道数据的维度，确定第一字典所包括的元素的维度。确定方式可参考前文。第一字典例如为第一节点训练得到的其中一个字典。
在确定第一字典所包括的元素的数量以及第一字典所包括的元素的维度后，第一节点就可以根据至少一份第三子下行信道数据进行聚类训练，从而得到第一字典。
本公开中可应用的聚类训练方式可以有多种，例如K-Means聚类方法等。另外，在神经网络的训练过程中，可以定义损失函数，损失函数描述了神经网络的输出值与理想目标值之间的差距或差异。在聚类训练过程中可以不使用损失函数，或者也可以使用损失函数。例如一种损失函数为，将多个训练样本与聚类中心之间的距离中的最小值作为目标，或者将多个训练样本中与聚类中心之间最相关的训练样本作为目标等。除此之外，损失函数还可能是其他函数，本公开并不限制损失函数的实现方式。字典的训练过程就是通过调整字典的参数，使得损失函数的取值小于门限，或者使得损失函数的取值满足目标需求的过程。调整字典的参数，例如包括调整字典的元素等。
关于字典所包括的内容，在图5所示的实施例已有介绍，不多赘述。
通过如上过程，第一节点就训练得到了N个字典，从而UE在进行图5所示的实施例所述的网络推理过程时可以使用N个字典，接入网设备在进行下行信道信息的恢复时也可以使用N个字典。通过划分数据空间以及N个字典，可以体现下行信道所对应的环境信息，有助于接入网设备恢复出更为准确的下行信道信息。
为了便于理解，下面用一些附图作为示例，以介绍本公开涉及的网络训练过程和网络推理过程。
可参考图9A，为本公开提供的训练过程和网络推理过程的一种示意图，图9A以及后续的各个附图所涉及的训练过程中，均以UE执行训练过程为例。其中，从第三下行信道数据至q 1～q 4，即，向接入网设备发送信息之前，可认为是训练过程；而图9A的全过程又可视为网络推理过程，网络推理过程也可认为是对一个数据的处理过程，当然该数据实际上并不是用于训练的训练数据，只是该数据的处理过程与训练数据是类似的。训练过程包括通过对多个训练数据进行聚类训练的方法得到字典，处理过程可以看做通过得到的字典 对下行信道数据进行表示。
在训练过程中，假设训练数据是原始的下行信道数据，该原始的下行信道数据实际上包括多个训练数据(或称为训练样本)。UE对原始的下行信道数据中的每个训练数据进行处理，得到特征向量，其维度为[N tx＝32,N sb＝13]。UE对该特征向量进行预处理，得到该特征向量的稀疏系数，多个训练数据对应的特征向量的稀疏系数可作为第三下行信道数据。由于网络训练采用实数训练，因此数据输入分为实部和虚部两部分，第三下行信道数据的维度例如为[E,2,32,13]。其中，第三下行信道数据的维度中的“E”，视为训练数据的数量，即，此时第三下行信道数据可以视为包括了E个训练数据，E为正整数。第三下行信道数据的维度中的“2”表示实部和虚部，“32”表示N tx，“13”表示N sb。
将第三下行信道数据划分到M个数据空间中，或者说，将第三下行信道数据划分为M份，图9A以M＝4为例，则可划分得到4份第三子下行信道数据，这4份第三子下行信道数据分别为y 1、y 2、y 3、y 4。其中y 1、y 2、y 3、y 4的维度均为[S,16*13]，S表示一份第三子下行信道数据所对应的训练数据的数量，16*13例如为待训练的字典的维度。图9A中的q 1～q 4表示4个待训练的字典，即，图9A以N＝M为例。UE按照聚类方式训练这4个字典。可选的，如果是离线训练的方式，这4个字典的信息可以约定在协议中，或者由UE发送给接入网设备。如果是在线训练的方式，这4个字典的信息可以由UE发送给接入网设备。
在推理过程中，例如，UE根据这4个训练得到的字典和4份第一子下行信道数据，可得到4个第一信息，其中一个第一信息为一份第一子下行信道数据在相应的字典中对应的元素。
例如UE向接入网设备发送4个第一信息的标识，其中每个第一信息的标识可占用X个比特。接入网设备接收4个第一信息的标识后，可根据4个字典恢复出4个第一信息，接入网设备再将4个第一信息进行拼接等处理，可恢复出下行信道信息，或者说，可以重构下行信道矩阵。
可再参考图9B，为本公开提供的训练过程的另一种示意图。假设UE对原始的下行信道数据中的每个训练数据进行处理，得到特征向量，其维度为[N tx＝32，N sb＝13]。UE对该特征向量进行预处理，得到该特征向量的稀疏系数，原始的下行信道数据包括的多个训练数据对应的特征向量的稀疏系数可作为第三下行信道数据。由于网络训练采用实数训练，因此数据输入分为实部和虚部两部分，第三下行信道数据的维度例如为[E,2,32,13]。其中，E为训练数据的数量，E为正整数。
将第三下行信道数据划分到M个数据空间中，或者说，将第三下行信道数据划分为M份，图9B以M＝4为例，划分得到的4份第三子下行信道数据可统一表示为y 1，即，y 1可视为包括4份第三子下行信道数据。y 1的维度为[4*S,16*13]，其中，S表示一份第三子下行信道数据对应的训练数据的数量，这4份第三子下行信道数据对应的训练数据的数量就是4*S，另外，16*13例如为待训练的字典的维度。图9B中的q 1表示M个数据空间统一对应的一个待训练的字典，即，图9B以N＝1为例。UE按照聚类方式训练该字典。
在网络推理过程中，对于每个数据空间，UE可以独立在相应的字典中找到对应的第一信息。如果按照图9B，M个数据空间统一对应一个字典，那么在网络推理过程中，UE对于每个数据空间，都可以在图9B训练出的该字典中找到对应的第一信息。继续以图9B为例，则UE可确定4个第一信息。例如UE向接入网设备发送这4个第一信息的标识，接入网设备接收4个第一信息的标识后，可根据4个字典恢复出4个第一信息，接入网设 备再将4个第一信息进行拼接等处理，可恢复出下行信道信息，或者说，可以重构下行信道矩阵。
图9A和图9B中，都以第三下行信道数据是原始的下行信道数据(或，特征向量)的预处理结果为例。根据图5所示的实施例的介绍可知，第三下行信道数据还可能是从预处理结果中取出的连续的F列数据。如果是这种情况，则字典所包括的元素的维度可能会有所变化。例如，第i份第三子下行信道数据是对从预处理结果中取出的F列数据划分了M份后的其中一份，则第i份第三子下行信道数据例如为维度是[16,F]的矩阵，UE可将该矩阵转换为长度为16*F的矢量，那么第一字典所包括的元素的维度就是16*F。F一般小于子带数量，这样可以减小字典占用的存储空间。
可参考图9C，为本公开提供的训练过程的又一种示意图。假设UE对原始的下行信道数据中的每个训练数据进行处理，得到特征向量，其维度为[N tx＝32,N sb＝13]。UE对该特征向量进行预处理，得到该特征向量的稀疏系数，原始的下行信道数据包括的多个训练数据对应的特征向量的稀疏系数可作为第三下行信道数据。由于网络训练采用实数训练，因此数据输入分为实部和虚部两部分，第三下行信道数据的维度例如为[E,2,32,13]。其中，E表示训练数据的数量，E为正整数。从第三下行信道数据中取出连续的F列数据，将这连续的F列数据划分到M个数据空间中，得到M份第三子下行信道数据。图9C以M＝4为例，划分得到的4份第三子下行信道数据分别为y 1、y 2、y 3、y 4。其中y 1、y 2、y 3、y 4的维度均为[S,16*F]。图9C中的q 1～q 4表示4个字典，即，图9C以N＝M为例。UE按照聚类方式训练这4个字典。
在网络推理过程中，对于每个数据空间，UE可以独立在相应的字典中找到对应的第一信息。继续以图9C为例，则UE可根据图9C训练出的4个字典确定4个第一信息。例如UE向接入网设备发送这4个第一信息的标识，接入网设备接收4个第一信息的标识后，可根据4个字典恢复出4个第一信息，接入网设备再将4个第一信息进行拼接等处理，可恢复出下行信道信息，或者说，可以重构下行信道矩阵。
可再参考图9D，为本公开提供的训练过程的再一种示意图。假设UE对原始的下行信道数据中的每个训练数据进行处理，得到特征向量，其维度为[N tx＝32,N sb＝13]。UE对该特征向量进行预处理，得到该特征向量的稀疏系数，原始的下行信道数据包括的多个训练数据对应的特征向量的稀疏系数可作为第三下行信道数据。由于网络训练采用实数训练，因此数据输入分为实部和虚部两部分，第三下行信道数据的维度例如为[E,2,32,13]。
从第三下行信道数据中取出连续的F列数据，将这连续的F列数据划分到M个数据空间中，得到M份第三子下行信道数据。图9D以M＝4为例，这4份第三子下行信道数据可统一表示为y 1，其中y 1的维度为[4*S,16*13]。图9B中的q 1表示M个数据空间统一对应的一个字典，即，图9C以N＝1为例。UE按照聚类方式训练该字典。
在网络推理过程中，对于每个数据空间，UE可以独立在相应的字典中找到对应的第一信息。如果按照图9D，M个数据空间统一对应一个字典，那么在网络推理过程中，UE对于每个数据空间，都可以在图9D训练出的该字典中找到对应的第一信息。继续以图9D为例，则UE可确定4个第一信息。例如UE向接入网设备发送这4个第一信息的标识，接入网设备接收4个第一信息的标识后，可根据4个字典恢复出4个第一信息，接入网设备再将4个第一信息进行拼接等处理，可恢复出下行信道信息，或者说，可以重构下行信道矩阵。
图8所示的实施例介绍的网络训练过程是训练得到字典的过程，而在前文也介绍了，UE侧可能设置编码器网络，接入网设备侧可能设置与该编码器网络对应的解码器网络，则还有一种网络训练过程为，将编码器网络、解码器网络以及字典进行联合训练的过程。接下来介绍本公开的又一种通信方法，通过该方法介绍联合训练过程。请参考图10，为该方法的流程图。
S1001、第二节点获得第五下行信道数据。
第五下行信道数据例如为原始的下行信道数据；或者，第五下行信道数据也可以是对原始的下行信道数据进行预处理所得到的数据；或者，第五下行信道数据还可能是神经网络输出的数据。其中，原始的下行信道数据可视为训练数据，或者称为训练样本。在训练字典的过程中，第二节点要对训练样本进行训练。所述原始的下行信道数据可以包括一个或多个训练数据。
如果第五下行信道数据是对原始的下行信道数据进行预处理得到的，那么就涉及到预处理过程。关于原始的下行信道数据的预处理过程，可参考图5所示的实施例的S501中对于第二下行信道数据的预处理过程的介绍。
本公开中，第二节点例如为UE，或者为接入网设备，或也可以是第三方设备(例如AI节点等)等。训练过程可以是在线训练也可以是离线训练。第二节点与图8所示的实施例所述的第一节点可以是同一个节点，或者也可以是不同的节点。
第二节点可利用第五下行信道数据，对编码器网络、字典、以及解码器网络进行联合训练。下面通过S1002～S1006来介绍训练过程。
S1002、第二节点将第五下行信道数据输入编码器网络，得到编码器网络输出的第六下行信道数据。
该编码器网络是需要训练的编码器网络，第二节点将第五下行信道数据输入编码器网络，该编码器网络可对第五下行信道数据进行压缩等处理。该编码器网络在处理后，会输出第六下行信道数据。
S1003、第二节点获得M份第六子下行信道数据。其中，每份第六子下行信道数据对应于M个数据空间中的一个数据空间。本公开中的M个数据空间，与图5所示的实施例中所述的M个数据空间，可以是相同的特征。
M份第六子下行信道数据是根据第六下行信道数据得到的，例如将第六下行信道数据划分到M个数据空间中，就可得到M份第六子下行信道数据。关于S1001的更多内容，可参考图8所示的实施例中的S801。
S1004、第二节点根据M份第六子下行信道数据和N个待训练的字典，得到M个第三信息。
例如在训练字典的过程中，第二节点根据M个数据空间中的第i个数据空间对待训练的字典分别进行训练，i取从1至M的整数，则第二节点可训练M个待训练的字典。第二节点根据M个数据空间中的第i个数据空间对该数据空间对应的待训练的字典进行训练，例如一种训练方式为，对于M份第六子下行信道数据中的第i份第六子下行信道数据，第二节点根据第i个数据空间对应的待训练的字典，得到第i份第六子下行信道数据对应的第三信息，则第二节点共可以得到M个第三信息。例如，第i份第六子下行信道数据对应的第三信息，是第i份第六子下行信道数据在第i个数据空间对应的待训练的字典中所对应的元素。
其中，在训练开始之前，可以设置一个初始模型作为待训练的字典，通过原始的下行信道数据对该初始模型进行多轮训练(其中，采用一个训练数据进行训练的过程可视为一轮训练过程)，在训练完成后可得到在网络推理阶段使用的字典。因此，上述的第i个数据空间对应的待训练的字典，可能是初始模型，也可能是对初始模型进行至少一轮训练后所得到的中间模型。
可选的，在训练字典的过程中，第二节点还可以根据M个数据空间对待训练的字典进行训练，则第二节点可训练出M个相同的字典或者1个字典。第二节点根据M个数据空间对待训练的字典进行训练，例如一种训练方式为，对于M份第六子下行信道数据中的第i份第六子下行信道数据，第二节点根据待训练的字典，得到第i份第六子下行信道数据对应的第三信息，则第二节点共可以得到M个第三信息。例如，第i份第六子下行信道数据对应的第三信息，是第i份第六子下行信道数据在待训练的字典中所对应的元素。
S1005、对于M个第三信息中的第i个第三信息，第二节点根据M个数据空间中的第i个数据空间对应的待训练的字典，恢复得到第i份第五子下行信道数据。i取从1至M的整数，则第二节点共可以得到M份第五子下行信道数据。
在理想状态下，第二节点所得到的M份第五子下行信道数据，与第二节点所得到的M份第六子下行信道数据，可以是相同的数据。例如，第i份第六子下行信道数据与第i份第五子下行信道数据是相同的数据。在实际应用中，M份第六子下行信道数据与M份第五子下行信道数据，这之间可能会出现偏差，关于该内容可参考图5所示的实施例中的S504。
关于S1005的更多内容，可参考图5所示的实施例中的S504。
S1006、第二节点将M份第五子下行信道数据输入解码器网络，得到解码器网络输出的L个恢复信息，L为正整数。或者，第二节点将M份第五子下行信道数据进行拼接，并将拼接后的子下行信道数据输入解码器网络，得到解码器网络输出的第一恢复信息。
该解码器网络是需要训练的解码器网络，也是与S1002中的编码器网络对应的解码器网络。
例如所述原始的下行信道数据包括多个训练数据，其中一个训练数据可包括子训练数据和标签。第二节点可将子训练数据输入编码器网络，再经过解码器网络后，解码器网络会输出推理结果(例如本公开所述的L个恢复信息或第一恢复信息)。第二节点可根据损失函数计算得到推理结果与标签之间的误差，基于该误差，第二节点可采用反向传播优化算法(或者称为模型优化算法等)，优化编码器网络和/或解码器网络的参数。通过大量训练数据对编码器网络和解码器网络进行训练，使得解码器网络的输出与标签之间的差异小于预设值后完成神经网络的训练。
需要说明的是，如上介绍的对于编码器网络和解码器网络的训练过程，是采用了监督学习的训练方式，即，基于训练数据与标签，利用损失函数实现编码器网络和解码器网络的训练。还有可能，智能模型的训练过程也可以采用无监督学习，利用算法学习训练数据的内在模式，实现基于训练数据完成智能模型的训练。智能模型的训练过程还可以采用强化学习，通过与环境进行交互获取环境反馈的激励信号，从而学习解决问题的策略，实现模型的优化。本公开对于模型的训练方法和模型的类型等不予限制。
根据上述介绍可知，第二节点在训练编码器网络和解码器网络时，可按照某个损失函数来进行训练。可选的，可以为M个数据空间设置同一种损失函数，即，对于M个数据空间中的任一个数据空间，第二节点均可按照该损失函数进行联合训练。例如解码器网络 输出的是L个恢复信息，则可将经过解码器网络恢复得到的L个恢复信息所拼接得到的数据与第五下行信道数据之间的均方差(mean square error，MSE)作为该损失函数，或者将经过解码器网络恢复得到的L个恢复信息所拼接得到的数据与第三下行信道数据之间的相关性作为该损失函数等；或者，解码器网络输出的是第一恢复信息，则可将第一恢复信息与第五下行信道数据之间的MSE作为该损失函数，或者将第一恢复信息与第三下行信道数据之间的相关性作为该损失函数等。
或者，也可以为不同的数据空间分别设置不同的损失函数。例如解码器网络输出的是L个恢复信息，则可将经过解码器网络恢复得到的恢复信息与输入编码器网络的数据之间的MSE作为一个数据空间对应的损失函数。其中，该损失函数所对应的经过解码器网络恢复得到的恢复信息，是经过解码器网络恢复得到的与该数据空间对应的恢复信息；该损失函数所对应的输入编码器网络的数据，是指输入编码器网络的第五下行信道数据中与该数据空间对应的数据。
通过如上过程，第二节点就对编码器网络、解码器网络以及字典进行了联合训练，可得到N个字典，以及得到编码器网络和对应的解码器网络，从而UE在进行图5所示的实施例所述的网络推理过程时可以使用N个字典和编码器网络，接入网设备在进行下行信道信息的恢复时也可以使用N个字典和解码器网络。通过划分数据空间以及N个字典，可以体现下行信道所对应的环境信息，有助于接入网设备恢复出更为准确的下行信道信息。如果图5所示的实施例中不使用编码器网络和解码器网络，则可以采用图8所示的实施例提供的网络训练方式单独训练字典；如果图5所示的实施例中需要使用编码器网络和解码器网络，则可以采用图10所示的实施例提供的网络训练方式来联合训练得到编码器网络、解码器网络以及字典。
例如可参考图11，为本公开提供的训练过程和网络推理过程的一种示意图，该训练过程例如联合训练得到了编码器网络、解码器网络和字典，该网络推理过程会用到编码器网络、解码器网络和字典。其中，从第三下行信道数据至q 1～q 4，即，向接入网设备发送信息之前，可认为是训练过程；而图11的全过程又可视为网络推理过程，网络推理过程也可认为是对一个训练数据的处理过程，当然该数据实际上并不是用于训练的训练数据，只是该数据的处理过程与训练数据是一致的。
在训练过程中，假设原始的下行信道数据可包括多个训练数据。UE对原始的下行信道数据中的每个训练数据进行处理，得到特征向量，其维度为[N tx＝32,N sb＝13]。UE对该特征向量进行预处理，得到该特征向量的稀疏系数。UE利用编码器网络对该特征向量的稀疏系数进行压缩，得到压缩信息。多个训练数据对应的压缩信息可作为第三下行信道数据。
UE将该第三下行信道数据划分到4个数据空间中，得到4份第三子下行信道数据，这4份第三子下行信道数据分别为y 1、y 2、y 3、y 4。其中y 1、y 2、y 3、y 4的维度均为[S,16*13]，S表示一份第三子下行信道数据对应的训练数据的数量，16*13例如为待训练的字典的维度。图11中的q 1～q 4表示4个待训练的字典，即，图11以N＝M为例。UE按照聚类方式训练这4个字典。
在推理过程中，例如，UE根据这4个待训练的字典和4份第一子下行信道数据，可得到4个第一信息。其中一个第一信息为一份第一子下行信道数据在相应的字典中对应的元素。
例如UE向接入网设备发送4个第一信息的标识，其中每个第一信息的标识可占用X个比特。接入网设备接收4个第一信息的标识后，可根据4个字典恢复出4个子压缩信息，接入网设备将4个子压缩信息进行拼接等处理，再将处理后的结果输入解码器网络，以得到解码器网络输出的恢复信息。接入网设备在得到解码器网络输出的恢复信息后，再根据恢复信息可恢复出下行信道信息。其中，在训练过程中可以应用图10所示的实施例所述的损失函数，使得训练得到的编解码网络性能更好。
基于前述的方法实施例，介绍本公开提供的通信装置。
可以理解的是，为了实现上述方法中的功能，接入网设备和UE等包括了执行各个功能相应的硬件结构和/或软件模块。本领域技术人员应该很容易意识到，结合本公开描述的各示例的单元及方法步骤，本公开能够以硬件或硬件和计算机软件相结合的形式来实现。某个功能究竟以硬件还是计算机软件驱动硬件的方式来执行，取决于技术方案的特定应用场景和设计约束条件。
本公开提供一种通信装置，该通信装置例如包括处理单元和收发单元(或者，称为通信单元)，处理单元可用于实现图5所示的实施例、图8所示的实施例或图10所示的实施例所述的UE的处理功能，收发单元可用于实现图5所示的实施例、图8所示的实施例或图10所示的实施例所述的UE的全部收发功能或部分收发功能。或者，处理单元可用于实现图5所示的实施例、图8所示的实施例或图10所示的实施例所述的接入网设备所实现的处理功能，收发单元可用于实现图5所示的实施例、图8所示的实施例或图10所示的实施例所述的接入网设备的全部收发功能或部分收发功能。
可选的，处理单元和/或收发单元可通过虚拟模块实现，例如处理单元可通过软件功能单元或虚拟装置实现，收发单元可通过软件功能单元或虚拟装置实现。或者，处理单元和/或收发单元也可通过实体装置(例如电路系统和/或处理器等)实现。对于处理单元和收发单元通过实体装置实现的情况，下面进行介绍。
图12给出了本公开提供的一种通信装置的结构示意图。所述通信装置1200可以是图5所示的实施例、图8所示的实施例或图10所示的实施例所述的UE、该UE的电路系统或能够应用于该UE的电路系统等，用于实现上述方法实施例中对应于UE的方法。或者，所述通信装置1200可以是图5所示的实施例、图8所示的实施例或图10所示的实施例所述的接入网设备、该接入网设备的电路系统或能够应用于该接入网设备的电路系统等，用于实现上述方法实施例中对应于接入网设备的方法。具体的功能可以参见上述方法实施例中的说明。其中，例如一种电路系统为芯片系统。
通信装置1200包括一个或多个处理器1201。处理器1201可以实现一定的控制功能。所述处理器1201可以是通用处理器或者专用处理器等。例如，包括：基带处理器，中央处理器等。所述基带处理器可以用于对通信协议以及通信数据进行处理。所述中央处理器可以用于对通信装置1200进行控制，执行软件程序和/或处理数据。不同的处理器可以是独立的器件，也可以是设置在一个或多个处理电路中，例如，集成在一个或多个专用集成电路上。
可选的，通信装置1200中包括一个或多个存储器1202，用以存储指令1204，所述指令1204可在所述处理器上被运行，使得通信装置1200执行上述方法实施例中描述的方法。可选的，所述存储器1202中还可以存储有数据。所述处理器和存储器可以单独设置，也可以集成在一起。该存储器可以是非易失性存储器，比如硬盘(hard disk drive，HDD)或 固态硬盘(solid-state drive，SSD)等，还可以是易失性存储器(volatile memory)，例如随机存取存储器(random-access memory，RAM)。存储器是能够用于携带或存储具有指令或数据结构形式的期望的程序代码并能够由计算机存取的任何其他介质，但不限于此。本公开中的存储器还可以是电路或者其它任意能够实现存储功能的装置，用于存储程序指令和/或数据。
可选的，通信装置1200可以存储指令1203(有时也可以称为代码或程序)，所述指令1203可以在所述处理器上被运行，使得所述通信装置1200执行上述实施例中描述的方法。处理器1201中可以存储数据。
例如，所述处理单元通过一个或多个处理器1201实现，或者，所述处理单元通过一个或多个处理器1201以及一个或多个存储器1202实现，或者，所述处理单元通过一个或多个处理器1201、一个或多个存储器1202、以及指令1203实现。
可选的，通信装置1200还可以包括收发器1205以及天线1206。收发器1205可以称为收发单元、收发机、收发电路、收发器，输入输出接口等，用于通过天线1206实现通信装置1200的收发功能。例如，所述收发单元通过收发器1205实现，或者，所述收发单元通过收发器1205以及天线1206实现。
可选的，通信装置1200还可以包括以下一个或多个部件：无线通信模块，音频模块，外部存储器接口，内部存储器，通用串行总线(universal serial bus，USB)接口，电源管理模块，天线，扬声器，麦克风，输入输出模块，传感器模块，马达，摄像头，或显示屏等等。可以理解，在一些实施例中，通信装置1200可以包括更多或更少部件，或者某些部件集成，或者某些部件拆分。这些部件可以是硬件，软件，或者软件和硬件的组合实现。
本公开中描述的处理器1201和收发器1205可实现在集成电路(integrated circuit，IC)、模拟IC、射频集成电路(radio frequency identification，RFID)、混合信号IC、专用集成电路(application specific integrated circuit，ASIC)、印刷电路板(printed circuit board，PCB)、或电子设备等上。实现本文描述的通信装置，可以是独立设备(例如，独立的集成电路，手机等)，或者可以是较大设备中的一部分(例如，可嵌入在其他设备内的模块)，具体可以参照前述各个实施例关于UE，以及接入网设备的说明，在此不再赘述。
本公开提供了一种终端设备，该终端设备可用于前述各个实施例中。所述终端设备包括用以实现图5所示的实施例、图8所示的实施例或图10所示的实施例所述的UE功能的相应的手段(means)、单元和/或电路。例如，终端设备，包括收发模块(或者，称为收发单元)，用以支持终端设备实现收发功能，和，处理模块(或者，称为处理单元)，用以支持终端设备对信号进行处理。
本公开还提供一种接入网设备，该接入网设备可用于前述各个实施例中。所述接入网设备包括用以实现图5所示的实施例、图8所示的实施例或图10所示的实施例所述的接入网设备功能的相应的手段(means)、单元和/或电路。例如，接入网设备，包括收发模块(或者，称为收发单元)，用以支持接入网设备实现收发功能，和，处理模块(或者，称为处理单元)，用以支持接入网设备对信号进行处理。
本公开提供的技术方案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。在计算机上加载和执行所述计算机程序指令时，全部或部分地产生按照本公开所述的流程或功能。所述计算机可以是通用计算机、专用计算机、 计算机网络、接入网设备、终端设备、AI节点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可以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，数字视频光盘(digital video disc，DVD))、或者半导体介质等。
以上所述，仅为本公开的具体实施方式，但本公开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公开揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本公开的保护范围之内。因此，本公开的保护范围应所述以权利要求的保护范围为准。

Claims (15)

  1. 一种通信方法，其特征在于，包括：
    获得M份第一子下行信道数据，其中每份第一子下行信道数据对应于M个数据空间中的一个数据空间，M为大于1的整数；
    对于所述M份第一子下行信道数据中的第i份第一子下行信道数据，根据与所述M个数据空间中的第i个数据空间对应的第一字典确定所述第i份第一子下行信道数据对应的第一信息，共确定M个第一信息，i取从1至M的整数，所述第i份第一子下行信道数据对应于所述第i个数据空间，所述第一字典包括多个元素，所述第i份第一子下行信道数据对应的第一信息对应于所述多个元素中的P个元素，P为正整数；
    发送第一指示信息，所述第一指示信息用于指示所述M个第一信息。
  2. 根据权利要求1所述的方法，其特征在于，所述第一指示信息用于指示所述M个第一信息的标识，其中，发送所述第一信息，包括：
    按照第一顺序发送所述M个第一信息的标识，所述第一顺序为所述M个数据空间的排列顺序。
  3. 根据权利要求2所述的方法，其特征在于，
    所述第一顺序为预定义的顺序；或，
    接收第二指示信息，所述第二指示信息用于指示所述第一顺序；或，
    确定所述第一顺序，并发送第三指示信息，所述第三指示信息用于指示所述第一顺序。
  4. 根据权利要求1～3任一项所述的方法，其特征在于，所述M份第一子下行信道数据是根据第一下行信道数据得到的，其中，
    所述第一下行信道数据为预处理结果；或，
    所述第一下行信道数据包括预处理结果中连续的F列数据；或，
    所述第一下行信道数据为对预处理结果进行压缩所得到的压缩信息；
    其中，所述预处理结果是对第二下行信道数据进行预处理得到的。
  5. 根据权利要求1～4任一项所述的方法，其特征在于，
    所述M个数据空间的划分方式为预定义；或，
    接收第四指示信息，所述第四指示信息用于指示所述M个数据空间的划分方式；或，
    确定所述M个数据空间的划分方式，并发送第五指示信息，所述第五指示信息用于指示所述M个数据空间的划分方式。
  6. 一种通信方法，其特征在于，包括：
    接收第一指示信息，所述第一指示信息用于指示M个第一信息，M为大于1的整数；
    对于所述M个第一信息中的第i个第一信息，根据M个数据空间中的第i个数据空间对应的第一字典，恢复第i份第二子下行信道数据，共得到M份第二子下行信道数据，所述第i个第一信息对应于所述第i个数据空间，i取从1至M的整数，所述第一字典包括多个元素，所述第i份第二子下行信道数据对应的第一信息对应于所述多个元素中的P个元素；
    根据所述M份第二子下行信道数据，恢复得到下行信道信息。
  7. 根据权利要求6所述的方法，其特征在于，接收第一指示信息，包括：
    按照第一顺序接收所述M个第一信息的标识，所述第一顺序为所述M个数据空间的 排列顺序。
  8. 根据权利要求7所述的方法，其特征在于，
    所述第一顺序为预定义的顺序；或，
    发送第二指示信息，所述第二指示信息用于指示所述第一顺序；或，
    接收第三指示信息，所述第三指示信息用于指示所述第一顺序。
  9. 根据权利要求6～8任一项所述的方法，其特征在于，所述M个数据空间对应M个字典，其中每个数据空间对应一个字典；或，所述M个数据空间均对应于同一个字典。
  10. 根据权利要求6～9任一项所述的方法，其特征在于，根据所述M份第二子下行信道数据，恢复得到下行信道信息，包括：
    根据所述M份第二子下行信道数据得到压缩信息；
    根据所述压缩信息得到所述下行信道信息。
  11. 根据权利要求6～10任一项所述的方法，其特征在于，
    所述M个数据空间的划分方式为预定义；或，
    发送第四指示信息，所述第四指示信息用于指示所述M个数据空间的划分方式；或，
    接收第五指示信息，所述第五指示信息用于指示所述M个数据空间的划分方式。
  12. 一种通信装置，其特征在于，用于实现如权利要求1～5任一项所述的方法，或用于实现如权利要求6～11任一项所述的方法。
  13. 一种通信装置，其特征在于，包括处理器和存储器，所述存储器和所述处理器耦合，所述处理器用于执行如权利要求1～5任一项所述的方法，或用于执行如权利要求6～11任一项所述的方法。
  14. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质用于存储计算机程序，当所述计算机程序在计算机上运行时，使得所述计算机执行如权利要求1～5中任一项所述的方法，或使得所述计算机执行如权利要求6～11中任一项所述的方法。
  15. 一种通信系统，其特征在于，包括用于实现如权利要求1～5任一项所述的方法的装置，和/或，用于实现如权利要求6～11任一项所述的方法的装置。
PCT/CN2022/142946 2021-12-31 2022-12-28 一种通信方法及装置 WO2023125699A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111663303.8 2021-12-31
CN202111663303.8A CN116419269A (zh) 2021-12-31 2021-12-31 一种通信方法及装置

Publications (1)

Publication Number Publication Date
WO2023125699A1 true WO2023125699A1 (zh) 2023-07-06

Family

ID=86997998

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/142946 WO2023125699A1 (zh) 2021-12-31 2022-12-28 一种通信方法及装置

Country Status (2)

Country Link
CN (1) CN116419269A (zh)
WO (1) WO2023125699A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107113797A (zh) * 2015-03-20 2017-08-29 华为技术有限公司 一种信道状态信息的传输方法、用户设备及接入网设备
WO2019137445A1 (zh) * 2018-01-12 2019-07-18 华为技术有限公司 信道状态信息的测量方法和装置
WO2020143580A1 (zh) * 2019-01-11 2020-07-16 华为技术有限公司 用于构建预编码向量的向量指示方法和通信装置
WO2020156103A1 (zh) * 2019-01-30 2020-08-06 华为技术有限公司 信息反馈方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107113797A (zh) * 2015-03-20 2017-08-29 华为技术有限公司 一种信道状态信息的传输方法、用户设备及接入网设备
WO2019137445A1 (zh) * 2018-01-12 2019-07-18 华为技术有限公司 信道状态信息的测量方法和装置
WO2020143580A1 (zh) * 2019-01-11 2020-07-16 华为技术有限公司 用于构建预编码向量的向量指示方法和通信装置
WO2020156103A1 (zh) * 2019-01-30 2020-08-06 华为技术有限公司 信息反馈方法及装置

Also Published As

Publication number Publication date
CN116419269A (zh) 2023-07-11

Similar Documents

Publication Publication Date Title
US11956031B2 (en) Communication of measurement results in coordinated multipoint
US20240137082A1 (en) Communication method and apparatus
US20230412430A1 (en) Inforamtion reporting method and apparatus, first device, and second device
CN115053465B (zh) 一种信息传输方法及装置
US20240187283A1 (en) Channel information feedback method, channel information recovery method, and apparatus
WO2023116808A1 (zh) 一种通信方法及装置
WO2023126007A1 (zh) 信道信息传输方法及装置
WO2023125699A1 (zh) 一种通信方法及装置
WO2023202514A1 (zh) 一种通信方法及装置
WO2023174108A1 (zh) 一种通信方法及装置
WO2023279947A1 (zh) 一种通信方法及装置
WO2023231881A1 (zh) 一种模型应用方法及装置
WO2024008004A1 (zh) 一种通信方法及装置
WO2023198018A1 (zh) 一种上行信号发送、接收方法及装置
WO2023006096A1 (zh) 一种通信方法及装置
WO2024001744A1 (zh) 一种信道状态信息的上报方法及通信装置
WO2023115254A1 (zh) 处理数据的方法及装置
WO2024094177A1 (zh) 信道状态信息确定方法、装置及可读存储介质
WO2023231934A1 (zh) 一种通信方法及装置
WO2023185758A1 (zh) 数据的传输方法和通信装置
US20240187052A1 (en) Communication method and apparatus
WO2024046288A1 (zh) 通信方法和装置
WO2023125996A1 (zh) 一种上行预编码方法及装置
CN117200838A (zh) 一种通信方法及装置
CN118355610A (zh) 处理数据的方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22914950

Country of ref document: EP

Kind code of ref document: A1