WO2023174253A1 - Ai模型的处理方法及设备 - Google Patents

Ai模型的处理方法及设备 Download PDF

Info

Publication number
WO2023174253A1
WO2023174253A1 PCT/CN2023/081303 CN2023081303W WO2023174253A1 WO 2023174253 A1 WO2023174253 A1 WO 2023174253A1 CN 2023081303 W CN2023081303 W CN 2023081303W WO 2023174253 A1 WO2023174253 A1 WO 2023174253A1
Authority
WO
WIPO (PCT)
Prior art keywords
communication device
model
area
data
channel
Prior art date
Application number
PCT/CN2023/081303
Other languages
English (en)
French (fr)
Inventor
姜大洁
吴建明
Original Assignee
维沃移动通信有限公司
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 维沃移动通信有限公司 filed Critical 维沃移动通信有限公司
Publication of WO2023174253A1 publication Critical patent/WO2023174253A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413—MIMO systems
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/06—Testing, supervising or monitoring using simulated traffic

Definitions

  • This application belongs to the field of communication technology, and specifically relates to a processing method and equipment for an artificial intelligence (Artificial Intelligence, AI) model.
  • AI Artificial Intelligence
  • AI is currently widely used in various fields. By integrating AI into the field of wireless communications, throughput can be significantly improved, latency reduced, and user capacity increased.
  • the network usually provides a generalized sum for all terminals.
  • Cell-related AI models it is difficult for generalized AI models to effectively further improve communication system performance. For example, it is difficult to improve the feedback performance of multiple input multiple output-channel state information (MIMO-CSI).
  • MIMO-CSI multiple input multiple output-channel state information
  • Embodiments of the present application provide an AI model processing method and device, which can solve the problem that due to the generalization of the AI model, it is difficult to effectively further improve the performance of the communication system.
  • an AI model processing method including: a first communication device determines a first area; the first communication device performs at least one of the following: using data with characteristics of the first area to perform AI Model training: select an AI model matching the first area to perform the target communication service, wherein the AI model has characteristics of the first area.
  • a first communication device including: a determining module for determining a first area; an execution module for performing at least one of the following: performing an AI model using data with characteristics of the first area Training; selecting an AI model matching the first area to perform the target communication service, wherein the AI model has characteristics of the first area.
  • a communication device in a third aspect, includes a processor and a memory.
  • the memory stores a program or instructions that can be run on the processor.
  • the program or instructions are implemented when executed by the processor. The steps of the method as described in the first aspect.
  • a communication device including a processor and a communication interface, wherein the The processor is configured to determine the first area; and, perform at least one of the following: use data with characteristics of the first area to perform AI model training; select an AI model matching the first area to execute the target communication service, wherein, The AI model has characteristics of the first region.
  • an AI model processing system including: a terminal and a network side device.
  • the terminal can be used to perform the steps of the method described in the first aspect
  • the network side device can be used to perform the steps of the method described in the first aspect. The steps of the method described in this aspect.
  • a readable storage medium is provided. Programs or instructions are stored on the readable storage medium. When the programs or instructions are executed by a processor, the steps of the method described in the first aspect are implemented.
  • a chip in a seventh aspect, includes a processor and a communication interface.
  • the communication interface is coupled to the processor.
  • the processor is used to run programs or instructions to implement the method described in the first aspect. A step of.
  • a computer program/program product is provided, the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement the method described in the first aspect Method steps.
  • the first communication device determines the first area and performs at least one of the following: using data with characteristics of the first area for AI model training; selecting an AI model that matches the first area.
  • Execute the target communication service wherein the AI model has the characteristics of the first area. Since the AI model is adapted to the area characteristics, the characteristics can be trained and updated according to different areas, which is beneficial to improving the accuracy of the use of the AI model. and effectiveness, further improving communication system performance.
  • Figure 1 is a schematic diagram of a wireless communication system according to an embodiment of the present application.
  • Figure 2 is a schematic flow chart of an AI model processing method according to an embodiment of the present application.
  • Figure 3 is a schematic diagram of areas divided according to the embodiment of the present application.
  • Figure 4 is a schematic diagram of the behavior of a UE in different scenarios according to an embodiment of the present application.
  • Figure 5 is a schematic flow chart of an AI model processing method according to an embodiment of the present application.
  • Figure 6 is a schematic flow chart of an AI model processing method according to an embodiment of the present application.
  • Figure 7 is a schematic structural diagram of a first communication device according to an embodiment of the present application.
  • Figure 8 is a schematic structural diagram of a communication device according to an embodiment of the present application.
  • Figure 9 is a schematic structural diagram of a terminal according to an embodiment of the present application.
  • Figure 10 is a schematic structural diagram of a network side device according to an embodiment of the present application.
  • first, second, etc. in the description and claims of this application are used to distinguish similar objects and are not used to describe a specific order or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances so that the embodiments of the present application can be practiced in sequences other than those illustrated or described herein, and that "first" and “second” are distinguished objects It is usually one type, and the number of objects is not limited.
  • the first object can be one or multiple.
  • “and/or” in the description and claims indicates at least one of the connected objects, and the character “/" generally indicates that the related objects are in an "or” relationship.
  • LTE Long Term Evolution
  • LTE-Advanced, LTE-A Long Term Evolution
  • LTE-A Long Term Evolution
  • CDMA Code Division Multiple Access
  • TDMA Time Division Multiple Access
  • FDMA Frequency Division Multiple Access
  • OFDMA Orthogonal Frequency Division Multiple Access
  • SC-FDMA Single-carrier Frequency Division Multiple Access
  • NR New Radio
  • FIG. 1 shows a block diagram of a wireless communication system to which embodiments of the present application are applicable.
  • the wireless communication system includes a terminal 11 and a network side device 12.
  • the terminal 11 may be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer), or a notebook computer, a personal digital assistant (Personal Digital Assistant, PDA), a palmtop computer, a netbook, or a super mobile personal computer.
  • Tablet Personal Computer Tablet Personal Computer
  • laptop computer laptop computer
  • PDA Personal Digital Assistant
  • PDA Personal Digital Assistant
  • UMPC ultra-mobile personal computer
  • UMPC mobile Internet device
  • MID mobile Internet Device
  • AR augmented reality
  • VR virtual reality
  • robots wearable devices
  • WUE Vehicle User Equipment
  • PUE Pedestrian User Equipment
  • smart home home equipment with wireless communication functions, such as refrigerators, TVs, washing machines or furniture, etc.
  • game consoles personal computers (personal computer, PC), teller machine or self-service machine and other terminal-side devices.
  • Wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets) bracelets, smart anklets, etc.), smart wristbands, smart clothing, etc.
  • the network side device 12 may include an access network device or a core network device, where the access network device may also be called a radio access network device, a radio access network (Radio Access Network, RAN), a radio access network function or a wireless device.
  • Access network equipment may include a base station, a Wireless Local Area Network (WLAN) access point or a Wireless Fidelity (WiFi) node, etc.
  • the base station may be called a Node B, an Evolved Node B (eNB), or an access point.
  • BTS Base Transceiver Station
  • BSS Basic Service Set
  • ESS Extended Service Set
  • TRP Transmitting Receiving Point
  • the base station is not limited to specific technical terms. It should be noted that in this application, in the embodiment, only the base station in the NR system is taken as an example for introduction, and the specific type of the base station is not limited.
  • this embodiment of the present application provides an AI model processing method 200.
  • the method can be executed by the first communication device.
  • the method can be executed by software or hardware installed on the first communication device.
  • the method includes the following steps.
  • S202 The first communication device determines the first area.
  • the first communication device may be a mobile communication device, such as a terminal, etc.; it may also be a fixed communication device, such as a network side device, a fixed-position terminal, etc.
  • the first area corresponds to geographical coordinates; wherein the size of the first area is W ⁇ L, L is the length of the first area, W is the width of the first area, W, L is all positive numbers.
  • the first communication device includes a mobile communication device
  • the first communication device determining the first area includes: the mobile communication device determining the location information of the mobile communication device; the moving The communication device determines the first area based on the location information.
  • the terminal determines the first area according to its own location information, that is, the terminal is within the first area.
  • the first communication device includes a fixed communication device
  • the first communication device determining the first area includes: the fixed communication device determining the location information of the second communication device, and the second communication device determines the location information of the second communication device.
  • the communication device includes a mobile communication device; the fixed communication device determines the first area according to the location information, and the mobile communication device is within the first area.
  • the network side device determines the first area according to the location information of the terminal, and the terminal is within the first area.
  • the first communication device performs at least one of the following: using data with characteristics of the first area to perform AI model training; selecting an AI model matching the first area to execute the target communication service, wherein the AI model Having the characteristics of the first region.
  • the characteristics of the first area include: channel data between a second communication device and one or more first communication devices located in the first area; wherein, the first communication device A mobile communication device is included, and the second communication device includes a fixed communication device.
  • the first communication device includes a terminal (User Equipment, UE)
  • the second communication device includes an NR Node B (gNB)
  • the characteristics of the first area include: gNB and being in the first area Channel data between one or more UEs.
  • the range for AI model training includes: an area related to the maximum communication range of the fixed communication device.
  • the range for AI model training can be a range with gNB as the center and a radius of D max , where D max is the maximum communication distance between gNB and UE.
  • the first communication device includes a mobile communication device, such as a UE, and the mobile communication device in the first area can use data with characteristics of the first area for AI model training, and further The AI model matching the first area may be selected to perform a target communication service, such as channel prediction, channel compression, etc.
  • a target communication service such as channel prediction, channel compression, etc.
  • the first communication device includes a fixed communication device, such as a gNB.
  • the fixed communication device can use data with characteristics of the first area for AI model training, and can also choose to cooperate with the third area.
  • a region-matched AI model performs target communication services, such as channel prediction, channel compression, etc.
  • each area can correspond to a trained AI model; if the UE uses data with the characteristics of the first area AI model training is performed on characteristic data. Since there may be multiple UEs in the first area, each area can correspond to one or more trained AI models. For example, each UE independently trains and completes an AI model; for another example, Multiple UEs jointly train an AI model.
  • Embodiments of the present application can train AI models based on areas.
  • the trained AI models have area restrictions. Only when the mobile communication device (such as the receiving end or the sending end) enters the relevant area, the information related to this area will The AI model is used by the receiving end or the sending end, thereby improving the accuracy and effectiveness of the use of the AI model.
  • the first communication device determines the first area and performs at least one of the following: using data with characteristics of the first area to conduct AI model training;
  • the area matching AI model performs the target communication service, wherein the AI model has the characteristics of the first area. Since the AI model has the limitations of the area characteristics, it is beneficial to improve the accuracy and effectiveness of the use of the AI model and further improve communication. System performance.
  • Embodiment 200 mainly introduces the core ideas of the embodiments of the present application. The implementation details of the embodiments of the present application will be described in detail below in multiple scenarios.
  • an AI supervised learning (Supervised Learning) model (AI model for short) can be represented by a probability distribution function p(y
  • (y,x) can be regarded as the training data for AI model training, expressed as
  • (y n ,x n ) is the nth pair of input and output data
  • N is the total number of training data in the training data set.
  • the training of AI neural network parameters w can be solved by solving the cost To obtain the minimum value of function J(w), where the cost function J(w) is expressed as
  • the training data set can be divided into training data subsets, and the k-th data subset is associated with the parameter z k , then the training data set can be expressed as
  • x;w) can be expressed as
  • the AI model corresponding to the auxiliary parameter z k can be trained independently, and its cost function can be simplified as
  • Option 2 Method of distinguishing AI models based on regions
  • the wireless channel is far more affected by the UE than the network (such as gNB).
  • the behavior of the UE is unpredictable.
  • MIMO Multiple Input Multiple Output
  • the incident angle and emission angle of the UE and network-side transceiver antennas change, resulting in changes in the MIMO channel.
  • Scenario-2 in Figure 4 in the same geographical location, due to the rotation or swing of the antenna direction at the UE, the incident angle and emission angle of the UE will also change, resulting in changes in the MIMO channel. Therefore, the factors affecting the MIMO channel on the UE side will be far greater than those on the network side.
  • the generation process of the Space Channel Model is divided into multiple steps.
  • the first few steps mainly generate static or semi-static parameters associated with the MIMO channel, that is, within a certain time range, the parameters will not change due to time. Or the environment changes, and the MIMO channel changes.
  • the final steps generate relevant channel coefficients for cluster n and the pair of u-th receive antenna element and s-th transmit antenna element.
  • the MIMO channel coefficient is given by the following formula:
  • F rx, u, ⁇ and F rx, u, ⁇ are the receiving antenna unit u in the direction of the spherical basis vector respectively.
  • the field pattern of F tx,s, ⁇ and F tx,s, ⁇ are respectively the transmitting antenna unit s in the direction of the spherical basis vector.
  • field mode. is the position vector of the receiving antenna unit u, is the position vector of the transmitting antenna unit s, k n,m is the cross polarization power ratio (Cross Polarization Power Ratio) of the linear scale.
  • the second term is associated with the initial value, which can be obtained simply and effectively through training the AI model.
  • the first, fourth and sixth items are associated with the UE and are directly related to the behavior of the UE. Only by collecting a large number of AI training data sets can the AI model be effectively trained. Finally, the third and fifth items are related to the network side, and their behaviors are relatively stable.
  • the AI models can be distinguished by the geographical location of the UE.
  • the sixth item is associated with the Doppler effect.
  • the Doppler effect may occur.
  • the sixth item can predict and correct the influence of the Doppler effect.
  • the embodiments of this application simplify the complexity of each AI model and improve the accuracy and effectiveness of the use of AI models by dividing network areas, that is, restricting and distinguishing AI models to a certain extent.
  • the network can divide geographical coordinates into zones through a zone identifier (ie, zone identifier (ID), or Zone-ID) mechanism.
  • the network determines the network coverage area related to the maximum communication range D max based on the geographical location of the gNB, which is represented by Zone-ID. As shown in Figure 3, the size of each region is W ⁇ L, where L is the region length value and W is the region width value.
  • the training data set can be distinguished by the Zone-ID in the network coverage area. Therefore, the AI model obtained by training will have the characteristics of the Zone-ID.
  • AI model training process can be completed by the UE side or by the network side device.
  • uplink i.e., Up-Link
  • downlink i.e., Up-Link
  • the AI models mentioned in various embodiments of this application can be subdivided into uplink AI models and downlink AI models.
  • the AI model does not need to be subdivided, that is, a generalized AI model for uplink and downlink can be used.
  • the UE or the network side obtains the AI training data set (Training Data Set), that is, ⁇ D ⁇ n .
  • the training data set is divided into corresponding data subsets, namely
  • the superscripts UL and DL represent the uplink and downlink respectively, and n is the network cell.
  • the ID of k is the index of the k-th coverage area, and the index k is converted from the Zone-ID mapping, that is,
  • z is Zone-ID
  • It is the index mapping conversion function of the cell coverage area, which is determined by the high-level configuration.
  • the entire data set can be represented as
  • the AI model required by the nth network cell on the UE side can be represented by Table 1.
  • Table 1 AI model table for the nth network cell
  • the first communication device uses data with characteristics of the first area to perform AI model training, which can be divided into the following four examples.
  • the first communication device includes a mobile communication device, such as a UE
  • the AI model includes a downlink AI model
  • the first communication device uses an AI model with characteristics of the first area.
  • AI model training using data includes: the mobile communication device receives a downlink reference signal in the first area, and obtains data with characteristics of the first area based on the downlink reference signal; the mobile communication device uses the The data having the characteristics of the first area are used for AI model training.
  • the mobile communication device receives a downlink reference signal in the first area, and obtains data with characteristics of the first area based on the downlink reference signal; the mobile communication device uses the The data having the characteristics of the first area are used for AI model training.
  • the subsequent introduction of option 1 please refer to the subsequent introduction of option 1.
  • the first communication device includes a fixed communication device, such as a network side device
  • the AI model includes an uplink AI model
  • the first communication device uses an Performing AI model training on characteristic data includes: the fixed communication device receives an uplink reference signal, and obtains data with characteristics of the first area based on the uplink reference signal, and the uplink reference signal is located in the first area.
  • the subsequent introduction of option 2 please refer to the subsequent introduction of option 2.
  • the first communication device includes a fixed communication device, such as a network side device
  • the AI model includes a downlink AI model
  • the first communication device uses a network with the first area.
  • Performing AI model training on characteristic data includes: the fixed communication device receiving data with characteristics of the first area, and the data with characteristics of the first area being second communication within the first area.
  • the second communication device includes a mobile communication device; the fixed communication device uses the data with characteristics of the first area to perform AI model training.
  • the subsequent introduction of option three please refer to the subsequent introduction of option three.
  • the first communication device includes a fixed communication device, such as a network side device
  • the AI model includes a downlink AI model
  • the first communication device uses a network with the first area.
  • Performing AI model training on characteristic data includes: the fixed communication device receiving the first data, the first data being sent by a second communication device, the second communication device including a mobile communication device; the fixed communication device according to the The location information of the mobile communication device determines that the first data has the characteristics of the first area; the fixed communication device uses the data with the characteristics of the first area to perform AI model training.
  • the subsequent introduction of option four please refer to the subsequent introduction of option four.
  • the AI model can be obtained through UE-side training or network-side training, as shown in Table 2.
  • Option 1 UE executes the AI model training process
  • the AI model training process performed by the UE is generally only for the downlink MIMO channel.
  • the UE will perform real-time positioning and tracking of the UE based on methods such as Global Navigation Satellite System (GNSS) or 5G-NR positioning (i.e., 5G-NR Positioning).
  • GNSS Global Navigation Satellite System
  • 5G-NR positioning i.e., 5G-NR Positioning
  • the network performs real-time positioning and tracking of the UE, and notifies the UE of the positioning results, allowing the UE to obtain its own location coordinate information.
  • the UE When the UE enters the Connected Mode, the UE receives the downlink reference signal (i.e., Reference Signal), estimates and stores the MIMO channel data through means such as Minimum Mean Squared Error (MMSE).
  • MMSE Minimum Mean Squared Error
  • the downlink reference signal can be a periodic signal configured by the upper layer, or a periodic signal configured by the upper layer but triggered by the Media Access Control Element (MAC CE) or layer one (L1) signaling.
  • MAC CE Media Access Control Element
  • L1 layer one
  • periodic or aperiodic signals or directly transmit aperiodic L1 reference signals, such as Channel State Information-Reference Signal (CSI-RS).
  • CSI-RS Channel State Information-Reference Signal
  • Orthogonal frequency division multiplex (OFDM) symbols i.e., OFDM Symbol
  • time slot ie, Time Slot
  • the data subset of the downlink can be expressed as
  • L is the total number of AI training data for the k-th coverage area.
  • the UE side passes the collected AI training data subset To train the downlink AI model
  • the AI model training process performed by the network side can be for the uplink MIMO channel or the downlink MIMO channel. It is divided into the following options two, three and four.
  • Option 2 The network performs the AI model training process, targeting the uplink MIMO channel scenario.
  • the UE will timely determine the UE according to methods such as Global Navigation Satellite System (GNSS) or 5G-NR positioning (i.e., 5G-NR Positioning). position and track, and notify the network of the positioning results, allowing the network to obtain the location coordinate information of the UE.
  • GNSS Global Navigation Satellite System
  • 5G-NR Positioning 5G-NR positioning
  • the network will perform real-time positioning and tracking of the UE through methods such as 5G-NR positioning, so that the network can obtain the location coordinate information of the UE.
  • the network will configure the uplink reference signal for the UE through the higher layer, such as Sounding Reference Signal (SRS).
  • SRS Sounding Reference Signal
  • the uplink reference signal can be a periodic signal configured by the higher layer, a periodic or aperiodic signal configured by the higher layer but triggered by MAC-CE or L1 signaling, or a non-periodic L1 reference signal directly sent.
  • the network receives the reference signal, estimates and stores the MIMO channel data through MMSE and other means.
  • t is the time of the channel response, which can use orthogonal frequency division multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) symbols (i.e., OFDM Symbol) or by time Instead, k is the index of the coverage area, and l is the l-th data in the AI training data subset.
  • OFDM Orthogonal Frequency Division Multiplexing
  • L is the total number of AI training data for the k-th coverage area.
  • the network side passes the collected AI training data subset To train the uplink AI model
  • the AI model training process can be completed by the following two options.
  • Option three The UE will position and track the UE all the time based on methods such as GNSS or 5G-NR positioning (i.e., 5G-NR Positioning). Thus, the UE obtains its own location coordinate information.
  • methods such as GNSS or 5G-NR positioning (i.e., 5G-NR Positioning).
  • 5G-NR Positioning i.e., 5G-NR Positioning
  • the UE when the UE enters the Connected Mode, the UE receives the downlink reference signal (ie, Reference Signal), estimates and stores the MIMO channel data through MMSE and other means.
  • the downlink reference signal ie, Reference Signal
  • the downlink reference signal can be a periodic signal configured by the higher layer, a periodic or aperiodic signal configured by the higher layer but triggered by MAC-CE or L1 signaling, or the L1 reference signal can be sent directly, such as CSI Reference Signal (CSI-RS).
  • CSI-RS CSI Reference Signal
  • the UE will determine the AI training data while storing the MIMO channel data. Expressed by formula (2).
  • the data subset for the downlink It can be expressed by formula (3).
  • the UE uses the uplink channel resources (such as PUSCH) indicated by the network to convert the downlink data subset Report to the network.
  • uplink channel resources such as PUSCH
  • the network side passes the collected AI training data subset To train the downlink AI model
  • Option 4 The network will perform real-time positioning and tracking of the UE through 5G-NR positioning and other methods. Therefore, the network side obtains the location coordinate information of the UE side.
  • the UE When the UE enters the Connected Mode, the UE receives the downlink reference signal (ie, Reference Signal), estimates and stores the MIMO channel data through MMSE and other means.
  • the downlink reference signal ie, Reference Signal
  • the downlink reference signal can be a periodic signal configured by the higher layer, a periodic or aperiodic signal configured by the higher layer but triggered by MAC-CE or L1 signaling, or the L1 reference signal can be sent directly, such as ,CSI-RS.
  • the UE will determine the AI training data while storing the MIMO channel data.
  • l is the l-th data in the AI training data subset.
  • Downlink data subset can be expressed as
  • L is the total number of AI training data related to the nth gNB.
  • the UE uses the uplink channel resources indicated by the network, such as the Physical Uplink Shared Channel (PUSCH), to convert the downlink data subset Report to the network.
  • PUSCH Physical Uplink Shared Channel
  • the network side Since the network side knows the UE ID and its coordinate position, the network side distinguishes the AI training data subset associated with the kth area based on the coordinate position of the UE side. Likewise, data subsets It can be expressed by formula (3).
  • the network side passes the collected AI training data subset To train the downlink AI model
  • the AI model selection process is directly related to the AI model usage scenarios.
  • based on channel prediction For example, channel prediction based on CSI signals
  • the AI model trained on the data set only involves the UE side or the network side.
  • the AI model trained based on the data set of channel compression eg, channel compression estimated based on CSI signal
  • channel prediction is the channel obtained by the UE or the network through the existing time t (for example, the UE estimates the downlink channel based on the CSI-RS signal Or the network estimates the uplink channel based on the SRS signal. ) to predict the channel acquired at time t+ ⁇ , where ⁇ is the time length of channel prediction.
  • the UE passes the downlink channel at time t to predict the downlink channel at time t+ ⁇
  • the UE will use the predicted downlink channel Use uplink channel resources, such as Physical Uplink Control Channel (PUCCH), PUSCH, etc., to report to the network side, so that the network side can effectively determine the massive MIMO transmission method of the downlink.
  • PUCCH Physical Uplink Control Channel
  • PUSCH Physical Uplink Control Channel
  • the network side uses the downlink channel resources (such as PUCCH, PUSCH, etc.) reported by the UE side at time t. to predict the downlink channel at time t+ ⁇
  • the network side based on the predicted channel Can effectively determine the Massive MIMO transmission method of the downlink.
  • the network side passes the uplink channel at time t to predict the uplink channel at time t+ ⁇
  • the network side predicts the uplink channel To determine the Massive MIMO transmission method of the uplink, and use signaling, such as Physical Downlink Control Channel (Physical Downlink Control Channel, PDCCH), Physical Downlink Shared Channel (Physical Downlink Shared Channel, PDSCH), etc., to perform uplink through the UE Massive MIMO transmission of the link.
  • PDCCH Physical Downlink Control Channel
  • PDSCH Physical Downlink Shared Channel
  • the first communication device determines the first area and selects the AI model matching the first area to execute the target communication service.
  • the following two examples can be divided into the following two examples.
  • the first communication device includes a mobile communication device, the first communication device determines a first area; the first communication device selects an AI model execution target that matches the first area.
  • the communication service includes: the mobile communication device determines the location information of the mobile communication device; the mobile communication device determines the first area according to the location information; the mobile communication device determines the first area according to the index k , select the AI model corresponding to the index k among the K AI models that have been trained to perform channel prediction, and K is a positive integer.
  • the AI model includes an uplink AI model
  • the channel prediction includes predicting a channel at time t+ ⁇
  • the method further includes: the mobile communication device receives a channel at time t, and the channel at time t The channel is sent by a second communication device, including a fixed communication device.
  • the AI model includes an uplink AI model or a downlink AI model.
  • the AI model includes an uplink AI model or a downlink AI model.
  • the first communication device includes a fixed communication device, the first communication device determines a first area; the first communication device selects an AI model execution target that matches the first area.
  • the communication service includes: the fixed communication device determines the location information of the second communication device, and the second communication device includes a mobile communication device; the fixed communication device determines the first area according to the location information; the fixed communication The device selects the AI model corresponding to the index k among the K AI models that have been trained to perform channel prediction based on the index k of the first area, where K is a positive integer.
  • the AI model includes an uplink AI model
  • the channel prediction includes predicting a channel at time t+ ⁇
  • the method further includes: the fixed communication device receives a channel at time t, and the channel at time t The channel is sent by the mobile communication device.
  • the AI model includes an uplink AI model or a downlink AI model.
  • the AI model includes an uplink AI model or a downlink AI model.
  • AI model In the scenario of channel prediction, there are four options for the UE or network side to select the uplink and downlink AI models, as shown in Table 3.
  • Table 3 AI model selection method in channel prediction scenarios.
  • Option 1 The UE locates itself, or obtains the UE's own position coordinates through the network's positioning.
  • the UE determines the Zone-ID, that is, z, based on its own position coordinates, and obtains the zone index k corresponding to the AI model through formula (1).
  • the UE is in the trained K uplink AI models, that is, , select the kth AI model. According to AI model UE side prediction channel
  • the network can convert the uplink channel Tell the UE side, according to the AI model UE side prediction channel
  • the network positions the UE, or obtains the position coordinates of the UE through the positioning of the UE.
  • the network side determines the Zone-ID, that is, z, based on the location coordinates of the UE side, and obtains the zone index k corresponding to the AI model through formula (1); the network side determines the Zone-ID among the trained K uplink AI models, that is, , select the kth AI model.
  • Option three The UE locates itself, or obtains the UE's own position coordinates through the network's positioning; the UE determines the Zone-ID, that is, z, based on its own position coordinates, and obtains the corresponding zone index of the AI model through formula (1) k; the UE is in the trained K downlink AI models, that is, , select the kth AI model.
  • AI model UE side prediction channel According to AI model UE side prediction channel
  • the network locates the UE, or obtains the position of the UE through the positioning of the UE. location coordinates; the network side determines the Zone-ID, that is, z, based on the location coordinates of the UE side, and obtains the corresponding zone index k of the AI model through formula (1); the network side is among the trained K downlink AI models, that is, , select the kth AI model.
  • AI model Network side prediction channel
  • the UE can convert the downlink channel Tell the network side, according to the AI model Network side prediction channel
  • the definition of channel compression is the channel obtained by the UE through the existing time t (for example, the UE estimates the downlink channel based on the CSI-RS signal ).
  • the UE compresses the channel through the encoder of the AI neural network (ie, Encoder) and reports it to the network.
  • the network side decodes the received signal through the decoder of the AI neural network (i.e., Decoder) and restores the channel.
  • the AI model consists of an AI encoder model and an AI decoder model, respectively and or and means, among which
  • the first communication device determines the first area and selects the AI model matching the first area to execute the target communication service.
  • the following two examples can be divided into the following two examples.
  • the first communication device includes a mobile communication device, the first communication device determines a first area; the first communication device selects an AI model matching the first area to execute the target communication service
  • the method includes: the mobile communication device determines the location information of the mobile communication device; the mobile communication device determines the first area according to the location information; the mobile communication device determines the first area according to the index k of the first area. Select the AI compression model corresponding to the index k from the K AI models that have been trained to compress the channel to obtain compressed channel data, where K is a positive integer; the mobile communication device sends the compressed channel data to the second communication device, so
  • the second communication device includes a fixed communication device.
  • the method further includes: the mobile communication device sending an AI decoding model corresponding to the AI compression model to the fixed communication device; wherein the AI decoding model is used to decompress the compressed channel data.
  • the mobile communication device determines the location information of the mobile communication device including: the mobile communication device independently determines the location information of the mobile communication device; or, the mobile communication device determines the location information of the mobile communication device; The communication device receives the location information of the mobile communication device, and the location information is determined by the fixed communication device.
  • the mobile communication device determines the location information of the mobile communication device including: the mobile communication device independently determines the location information of the mobile communication device; or, the mobile communication device determines the location information of the mobile communication device; The communication device receives the location information of the mobile communication device, and the location information is determined by the fixed communication device.
  • the mobile communication device determines the location information of the mobile communication device including: the mobile communication device independently determines the location information of the mobile communication device; or, the mobile communication device determines the location information of the mobile communication device; The communication device receives the location information of the mobile communication device, and the location information is determined by the fixed communication device.
  • the first communication device includes a fixed communication device, the first communication device determines a first area; the first communication device selects an AI model matching the first area to execute the target communication service
  • the method includes: the fixed communication device determines the location information of the second communication device, and the second communication device includes a mobile communication device; the fixed communication device determines the first area according to the location information; the fixed communication device determines the first area according to the location information; For the index k of the first area, the AI compression model corresponding to the index k is selected from the K AI models that have been trained to compress the channel to obtain compressed channel data, where K is a positive integer; the fixed communication device transmits data to the The mobile communications device transmits the compressed channel data.
  • the method further includes: the fixed communication device sending an AI decoding model corresponding to the AI compression model to the mobile communication device; wherein the AI decoding model is used to decompress the compressed channel data.
  • the fixed communication device determines the location information of the second communication device including: the fixed communication device autonomously determines the location information of the second communication device; or the fixed communication device receives the second communication device the location information determined by the second communication device.
  • the fixed communication device determines the location information of the second communication device including: the fixed communication device autonomously determines the location information of the second communication device; or the fixed communication device receives the second communication device the location information determined by the second communication device.
  • the fixed communication device determines the location information of the second communication device including: the fixed communication device autonomously determines the location information of the second communication device; or the fixed communication device receives the second communication device the location information determined by the second communication device.
  • the selection of the AI model is determined based on whether the UE performs channel compression or the network performs channel compression.
  • Table 4 AI model selection method for channel compression scenarios.
  • Option 1 The UE itself performs positioning or updates the UE's own geographical location regularly or as needed.
  • the UE selects the AI encoder model of the corresponding area (i.e. Zone-ID) based on its own geographical location, i.e. Compress the channel.
  • the UE sends compressed channel data packets to the network through the physical channel (such as PUCCH or PUSCH), and can also notify the network of the corresponding AI decoder model.
  • the network side passes the AI decoder model Decoding the received packet restores the channel at time t.
  • the UE before or after the UE sends the compressed channel data packet to the network, the UE notifies the network of the corresponding AI decoder model.
  • the UE can notify the network of the corresponding AI decoder model through the index k of the AI model.
  • Option 2 The network performs positioning or updates the geographical location of the UE regularly or on demand.
  • the network notifies the UE to update the geographical location of the UE.
  • the UE selects the AI encoder model of the corresponding area (i.e. Zone-ID) based on its own geographical location, i.e. Compress the channel.
  • the UE sends compressed channel data packets to the network through the physical channel (such as PUCCH or PUSCH), and can also notify the network of the corresponding AI decoder model, that is,
  • the network side passes the AI decoder model Decoding the received packet restores the channel at time t.
  • the UE before or after sending the compressed channel data packet to the network, the UE notifies the network of the corresponding AI decoder model.
  • the UE can notify the network of the corresponding AI decoder model through the index k of the AI model.
  • Option three The UE itself performs positioning or updates the UE's own geographical location regularly or as needed.
  • the UE notifies the network to update the geographical location of the UE.
  • the network side selects the AI encoder model of the corresponding area (i.e., Zone-ID) based on the geographical location of the UE side, that is, Compress the channel.
  • the network sends compressed channel data packets to the UE through the physical channel (such as PDCCH or PDSCH), and can also notify the UE of the corresponding AI decoder model, that is,
  • the UE side passes the AI decoder model Decoding the received packet restores the channel at time t.
  • the network notifies the UE of the corresponding AI decoder model before or after sending the compressed channel data packet to the UE.
  • the network side can notify the UE side of the corresponding AI decoder model through the index k of the AI model.
  • the network performs positioning or updates the geographical location of the UE regularly or on demand.
  • the network side selects the AI encoder model of the corresponding area (i.e., Zone-ID) based on the geographical location of the UE side, that is, Compress the channel.
  • the network sends compressed channel data packets to the UE through the physical channel (such as PDCCH or PDSCH), and can also notify the UE of the corresponding AI decoder model, that is,
  • the UE side passes the AI decoder model Decoding the received packet recovers the channel at time t.
  • the network notifies the UE of the corresponding AI decoder model before or after sending the compressed channel data packet to the UE.
  • the network can notify the UE of the corresponding AI decoder model through the index k of the AI model.
  • the UE side Or the network side can select two or more different types of AI models at the same time based on the specific geographical location (region) of the UE side.
  • the AI model includes an AI prediction model and an AI compression model.
  • the first communication device selects an AI model matching the first area to execute the target communication service, including: the first The communication device selects the channel predicted by the AI prediction model at time t+ ⁇ ; the first communication device selects the AI compression model to compress the channel at time t+ ⁇ to obtain compressed channel data; the first communication device The device sends the compressed channel data to a second communications device.
  • the method further includes: the first communication device sending an AI decoding model corresponding to the AI compression model to the second communication device; wherein the AI decoding model is used to decompress the compressed channel data. .
  • the first communication device includes a fixed communication device, and the second communication device includes a mobile communication device; or, the first communication device includes a mobile communication device, and the second communication device includes a fixed communication device.
  • the first communication device includes a fixed communication device
  • the second communication device includes a mobile communication device
  • Option 1 The UE selects the AI model in the downlink channel prediction scenario based on the specific geographical location of the UE, that is, At the same time, the UE can also choose the AI encoder model in the downlink channel compression scenario, that is, The UE predicts related AI models through the downlink channel According to the downlink channel pair channel Make predictions.
  • the UE compresses the relevant AI encoder model through the downlink channel prediction channel Perform compression.
  • the UE sends compressed channel data packets to the network through the uplink physical channel (such as PUCCH or PUSCH), and can also notify the network of the corresponding AI decoder model, that is,
  • the network side passes the AI decoder model Decode the received data packet to recover the downlink channel at time t+ ⁇
  • the UE before or after the UE sends the compressed channel data packet to the network, the UE notifies the network of the corresponding AI decoder model.
  • the UE can notify the network of the corresponding AI decoder model through the index k of the AI model.
  • Option 2 The network selects the AI model in the uplink channel prediction scenario based on the specific geographical location of the UE, that is, At the same time, the network also selects an AI encoder model for the uplink channel compression scenario, that is The network side predicts related AI models through the uplink channel According to the uplink channel pair channel Make predictions.
  • the network side compresses the relevant AI encoder model through the uplink channel prediction channel Perform compression.
  • the network side uses the downlink physical channel (such as PDCCH or PDSCH) Send compressed channel data packets to the UE, and at the same time notify the UE of the corresponding AI decoder model, that is The UE side passes the AI decoder model Decode the received data packet to recover the uplink channel at time t+ ⁇
  • the network notifies the UE of the corresponding AI decoder model before or after sending the compressed channel data packet to the UE.
  • the network side can notify the UE side of the corresponding AI decoder model through the index k of the AI model.
  • the UE side or the network side can freely select the AI model related to channel prediction and the AI model related to channel compression to match, thereby effectively realizing the channel prediction and channel compression related processes.
  • This embodiment is about the AI model training, channel prediction and channel compression process performed by the UE. As shown in Figure 5, this embodiment includes the following steps:
  • Step 1 UE trains the downlink channel prediction AI model UE side training downlink channel compression AI encoder model and AI decoder model
  • Step 2 The UE side will use the AI decoder Model reports to gNB.
  • Step 3 The UE locates itself, or the gNB locates the UE and notifies the UE of the relevant UE's geographical location.
  • Step 4 The UE determines the Zone-ID, that is, z, based on its own position coordinates, and obtains the zone index k corresponding to the AI model through formula (1); the UE is among the K trained AI models, that is, , select the kth AI model
  • Step 5 Based on the AI model and p downlink channels before time t, that is, The UE predicts the channel at time t+ ⁇ , that is,
  • Step 6 The UE determines the Zone-ID, that is, z, based on its own position coordinates, and obtains the zone index k corresponding to the AI model through formula (1); the UE is among the K trained AI models, that is, , select the kth AI encoder model
  • Step 7 Based on the AI encoder model The predicted downlink channel at the UE end Compression is performed to generate predicted channel data packets.
  • AI decoder models Related index k include AI decoder models Related index k.
  • Step 8 The UE sends the predicted channel data packet to the gNB through PUCCH or PUSCH.
  • Step 9 Based on the relevant index k, gNB uses the AI decoder model Decode the predicted channel data packet to obtain the predicted channel
  • This embodiment is about the process of AI model training, channel prediction and channel compression performed by gNB. As shown in Figure 6, this embodiment includes the following steps:
  • Step 1 gNB trains the downlink channel prediction AI model gNB trains downlink channel compression AI encoder model and AI decoder model
  • Step 2 gNB applies the downlink channel prediction AI model to and AI encoder model Tell the UE side.
  • Step 3 The UE locates itself, or the gNB locates the UE and notifies the UE of the relevant UE's geographical location.
  • Step 4 The UE determines the Zone-ID, that is, z, based on its own position coordinates, and obtains the zone index k corresponding to the AI model through formula (1); the UE is among the K trained AI models, that is, , select the kth AI model
  • Step 5 Based on the AI model and p downlink channels before time t, that is, The UE predicts the channel at time t+ ⁇ , that is,
  • Step 6 The UE determines the Zone-ID, that is, z, based on its own position coordinates, and obtains the zone index k corresponding to the AI model through formula (1); the UE is among the K trained AI models, that is, , select the kth AI encoder model
  • Step 7 Based on the AI encoder model The predicted downlink channel at the UE end Compression is performed to generate predicted channel data packets.
  • AI decoder models Related index k include AI decoder models Related index k.
  • Step 8 The UE sends the predicted channel data packet to the gNB through PUCCH or PUSCH.
  • Step 9 Based on the relevant index k, gNB uses the AI decoder model Decode the predicted channel data packet to obtain the predicted channel
  • FIG. 7 is a schematic structural diagram of a first communication device according to an embodiment of the present application.
  • the first communication device may correspond to a terminal or a network side device in other embodiments.
  • the first communication device 700 includes the following modules.
  • the determining module 702 may be used to determine the first area.
  • the execution module 704 may be used to perform at least one of the following: use data with characteristics of the first area to perform AI model training; select an AI model matching the first area to execute the target communication service, wherein the AI The model has characteristics of said first region.
  • the first communication device determines the first area and performs at least one of the following: using data with characteristics of the first area for AI model training; selecting an AI model that matches the first area for execution Target communication service, wherein the AI model has characteristics of the first area. Since the AI model has limitations of regional characteristics, it is beneficial to improve the accuracy and effectiveness of the use of the AI model and further improve the performance of the communication system.
  • the first communication device 700 includes a mobile communication device, and the determining module 702 is configured to: determine the location information of the mobile communication device; determine the first communication device according to the location information. area.
  • the first communication device 700 includes a fixed communication device
  • the determining module 702 is configured to: determine the location information of a second communication device, where the second communication device includes a mobile communication device; The first area is determined according to the location information, and the mobile communication device is within the first area.
  • the first communication device 700 includes a mobile communication device
  • the AI model includes a downlink AI model
  • the execution module 704 is configured to: receive downlink data in the first area. reference signal, and obtain data with characteristics of the first area according to the downlink reference signal; and use the data with characteristics of the first area to perform AI model training.
  • the first communication device 700 includes a fixed communication device
  • the AI model includes an uplink AI model
  • the execution module 704 is configured to: receive an uplink reference signal, and perform The uplink reference signal obtains data with the characteristics of the first area.
  • the uplink reference signal is sent by a second communication device located in the first area.
  • the second communication device includes a mobile communication device; using the Data with characteristics of the first area are used for AI model training.
  • the first communication device 700 includes a fixed communication device
  • the AI model includes a downlink AI model
  • the execution module 704 is configured to: receive characteristics of the first area
  • the data having the characteristics of the first area is sent by a second communication device located in the first area, and the second communication device includes a mobile communication device; using the data having the characteristics of the first area
  • the regional characteristic data is used for AI model training.
  • the first communication device 700 includes a fixed communication device
  • the AI model includes a downlink AI model
  • the execution module 704 is configured to: receive the first data, the first The data is sent by a second communication device, and the second communication device includes a mobile communication device; it is determined according to the location information of the mobile communication device that the first data has the characteristics of the first area; using the The characteristic data of the first area are used for AI model training.
  • the first communication device 700 includes a mobile communication device; the determination module 702 is used to determine the location information of the mobile communication device; and determine the first area according to the location information. ;
  • the execution module 704 is used to perform according to the index k of the first area, Among the K AI models that have been trained, the AI model corresponding to the index k is selected for channel prediction, and K is a positive integer.
  • the first communication device 700 includes a fixed communication device; the determining module 702 is used to determine the location information of a second communication device, and the second communication device includes a mobile communication device; according to The location information determines the first area; the execution module 704 is used to select the AI model corresponding to the index k among the K AI models that have been trained according to the index k of the first area to perform channel prediction.
  • K is a positive integer.
  • the first communication device 700 includes a mobile communication device; the determination module 702 is used to determine the location information of the mobile communication device; and determine the first area according to the location information. ; The execution module 704 is used to select the AI compression model corresponding to the index k among the K AI models that have been trained according to the index k of the first region to compress the channel to obtain compressed channel data, where K is a positive Integer; sending the compressed channel data to a second communication device, the second communication device including a fixed communication device.
  • the first communication device 700 includes a fixed communication device; the determining module 702 is used to determine the location information of a second communication device, and the second communication device includes a mobile communication device; according to The location information determines the first area; the execution module 704 is used to select the AI compression model corresponding to the index k among the K AI models that have been trained according to the index k of the first area. Perform compression to obtain compressed channel data, K is a positive integer; send the compressed channel data to the mobile communication device.
  • the AI model includes an AI prediction model and an AI compression model
  • the execution module 704 is used to: select the channel predicted by the AI prediction model at time t+ ⁇ ; select the AI compression model The model compresses the channel at time t+ ⁇ to obtain compressed channel data; and sends the compressed channel data to the second communication device.
  • the first communication device 700 can refer to the process of the method 200 corresponding to the embodiment of the present application, and each unit/module in the first communication device 700 and the above-mentioned other operations and/or functions are respectively to implement the method. 200, and can achieve the same or equivalent technical effects. For the sake of simplicity, they will not be repeated here.
  • the first communication device in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or may be a component in the electronic device, such as an integrated circuit or chip.
  • the electronic device may be a terminal or other devices other than the terminal.
  • terminals may include but are not limited to the types of terminals 11 listed above, and other devices may be servers, network attached storage (Network Attached Storage, NAS), etc., which are not specifically limited in the embodiment of this application.
  • NAS Network Attached Storage
  • the first communication device provided by the embodiment of the present application can implement each process implemented by the method embodiments of Figures 2 to 6, and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • this embodiment of the present application also provides a communication device 800, including a processor 801 and memory 802.
  • the memory 802 stores programs or instructions that can be run on the processor 801.
  • the communication device 800 is a terminal, when the program or instructions are executed by the processor 801, the above AI model is implemented. Each step of the processing method embodiment, and can achieve the same technical effect.
  • the communication device 800 is a network-side device, when the program or instruction is executed by the processor 801, each step of the above AI model processing method embodiment is implemented, and the same technical effect can be achieved. To avoid duplication, the details will not be described here.
  • Embodiments of the present application also provide a terminal, including a processor and a communication interface.
  • the processor is configured to determine a first area; and perform at least one of the following: performing AI model training using data with characteristics of the first area. ; Select an AI model matching the first area to perform the target communication service, wherein the AI model has characteristics of the first area.
  • This terminal embodiment corresponds to the above-mentioned terminal-side method embodiment.
  • Each implementation process and implementation manner of the above-mentioned method embodiment can be applied to this terminal embodiment, and can achieve the same technical effect.
  • FIG. 9 is a schematic diagram of the hardware structure of a terminal that implements an embodiment of the present application.
  • the terminal 900 includes but is not limited to: a radio frequency unit 901, a network module 902, an audio output unit 903, an input unit 904, a sensor 905, a display unit 906, a user input unit 907, an interface unit 908, a memory 909, a processor 910, etc. At least some parts.
  • the terminal 900 may also include a power supply (such as a battery) that supplies power to various components.
  • the power supply may be logically connected to the processor 910 through a power management system, thereby managing charging, discharging, and power consumption through the power management system. Management and other functions.
  • the terminal structure shown in FIG. 9 does not constitute a limitation on the terminal.
  • the terminal may include more or fewer components than shown in the figure, or may combine certain components, or arrange different components, which will not be described again here.
  • the input unit 904 may include a graphics processing unit (Graphics Processing Unit, GPU) 9041 and a microphone 9042.
  • the graphics processor 9041 is responsible for the image capture device (GPU) in the video capture mode or the image capture mode. Process the image data of still pictures or videos obtained by cameras (such as cameras).
  • the display unit 906 may include a display panel 9061, which may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like.
  • the user input unit 907 includes a touch panel 9071 and at least one of other input devices 9072 .
  • Touch panel 9071 also known as touch screen.
  • the touch panel 9071 may include two parts: a touch detection device and a touch controller.
  • Other input devices 9072 may include but are not limited to physical keyboards, function keys (such as volume control keys, switch keys, etc.), trackballs, mice, and joysticks, which will not be described again here.
  • the radio frequency unit 901 after receiving downlink data from the network side device, can transmit it to the processor 910 for processing; in addition, the radio frequency unit 901 can send uplink data to the network side device.
  • the radio frequency unit 901 includes, but is not limited to, an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc.
  • Memory 909 may be used to store software programs or instructions as well as various data. Memory 909 can be mastered It should include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area can store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function). etc. Additionally, memory 909 may include volatile memory or nonvolatile memory, or memory 909 may include both volatile and nonvolatile memory. Among them, the non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electrically removable memory.
  • ROM Read-Only Memory
  • PROM programmable read-only memory
  • Erasable PROM Erasable PROM
  • EPROM electrically removable memory.
  • Volatile memory can be random access memory (Random Access Memory, RAM), static random access memory (Static RAM, SRAM), dynamic random access memory (Dynamic RAM, DRAM), synchronous dynamic random access memory (Synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (Double Data Rate SDRAM, DDRSDRAM), enhanced synchronous dynamic random access memory (Enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (Synch link DRAM) , SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DRRAM).
  • RAM Random Access Memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • synchronous dynamic random access memory Synchronous DRAM, SDRAM
  • Double data rate synchronous dynamic random access memory Double Data Rate SDRAM, DDRSDRAM
  • Enhanced SDRAM, ESDRAM synchronous link dynamic random access memory
  • Synch link DRAM synchronous link dynamic random access memory
  • SLDRAM direct memory bus random access memory
  • the processor 910 may include one or more processing units; optionally, the processor 910 integrates an application processor and a modem processor, where the application processor mainly handles operations related to the operating system, user interface, application programs, etc., Modem processors mainly process wireless communication signals, such as baseband processors. It can be understood that the above modem processor may not be integrated into the processor 910.
  • the processor 910 may be used to determine the first region; and perform at least one of the following: use data with characteristics of the first region to conduct AI model training; select an AI model matching the first region to execute Target communication service, wherein the AI model has characteristics of the first area
  • the terminal determines the first area and performs at least one of the following: using data with characteristics of the first area to conduct AI model training; selecting an AI model that matches the first area to perform target communication services , wherein the AI model has characteristics of the first region. Since the AI model has limitations of regional characteristics, it is beneficial to improve the accuracy and effectiveness of the use of the AI model and further improve the performance of the communication system.
  • the terminal 900 provided by the embodiment of the present application can also implement each process of the above-mentioned AI model processing method embodiment, and can achieve the same technical effect. To avoid duplication, the details will not be described here.
  • An embodiment of the present application also provides a network-side device, including a processor and a communication interface.
  • the processor is configured to determine a first area; and perform at least one of the following: performing AI using data with characteristics of the first area.
  • Model training select an AI model matching the first area to perform the target communication service, wherein the AI model has characteristics of the first area.
  • the network side device actually The embodiment corresponds to the above-mentioned network-side device method embodiment.
  • Each implementation process and implementation manner of the above-mentioned method embodiment can be applied to this network-side device embodiment, and can achieve the same technical effect.
  • the embodiment of the present application also provides a network side device.
  • the network side device 1000 includes: an antenna 101 , a radio frequency device 102 , a baseband device 103 , a processor 104 and a memory 105 .
  • the antenna 101 is connected to the radio frequency device 102 .
  • the radio frequency device 102 receives information through the antenna 101 and sends the received information to the baseband device 103 for processing.
  • the baseband device 103 processes the information to be sent and sends it to the radio frequency device 102.
  • the radio frequency device 102 processes the received information and then sends it out through the antenna 101.
  • the method performed by the network side device in the above embodiment can be implemented in the baseband device 103, which includes a baseband processor.
  • the baseband device 103 may include, for example, at least one baseband board on which multiple chips are disposed, as shown in FIG. Program to perform the network device operations shown in the above method embodiments.
  • the network side device may also include a network interface 106, which is, for example, a common public radio interface (CPRI).
  • a network interface 106 which is, for example, a common public radio interface (CPRI).
  • CPRI common public radio interface
  • the network side device 1000 in this embodiment of the present invention also includes: instructions or programs stored in the memory 105 and executable on the processor 104.
  • the processor 104 calls the instructions or programs in the memory 105 to execute each of the steps shown in Figure 7. The method of module execution and achieving the same technical effect will not be described in detail here to avoid duplication.
  • Embodiments of the present application also provide a readable storage medium.
  • Programs or instructions are stored on the readable storage medium.
  • the program or instructions are executed by a processor, each process of the above-mentioned AI model processing method embodiment is implemented, and can To achieve the same technical effect, to avoid repetition, we will not repeat them here.
  • the processor is the processor in the terminal described in the above embodiment.
  • the readable storage medium includes computer readable storage media, such as computer read-only memory ROM, random access memory RAM, magnetic disk or optical disk, etc.
  • An embodiment of the present application further provides a chip.
  • the chip includes a processor and a communication interface.
  • the communication interface is coupled to the processor.
  • the processor is used to run programs or instructions to implement the processing method of the above-mentioned AI model.
  • Each process in the example can achieve the same technical effect. To avoid repetition, we will not repeat it here.
  • chips mentioned in the embodiments of this application may also be called system-on-chip, system-on-a-chip, system-on-chip or system-on-chip, etc.
  • Embodiments of the present application further provide a computer program/program product.
  • the computer program/program product is stored in a storage medium.
  • the computer program/program product is executed by at least one processor to implement the above-mentioned AI model processing method.
  • Each process of the embodiment can achieve the same technical effect, To avoid repetition, they will not be repeated here.
  • Embodiments of the present application also provide an AI model processing system, including: a terminal and a network side device.
  • the terminal can be used to perform the steps of the AI model processing method as described above.
  • the network side device can be used to perform the above steps. The steps of the AI model processing method.
  • the methods of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. implementation.
  • the technical solution of the present application can be embodied in the form of a computer software product that is essentially or contributes to the existing technology.
  • the computer software product is stored in a storage medium (such as ROM/RAM, disk , CD), including several instructions to cause a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of this application.

Landscapes

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

Abstract

本申请实施例公开了一种AI模型的处理方法及设备，属于通信技术领域，本申请实施例的AI模型的处理方法包括：第一通信设备确定第一区域；所述第一通信设备执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。

Description

AI模型的处理方法及设备
交叉引用
本发明要求在2022年03月18日提交中国专利局、申请号为202210269877.5、发明名称为“AI模型的处理方法及设备”的中国专利申请的优先权，该申请的全部内容通过引用结合在本发明中。
技术领域
本申请属于通信技术领域，具体涉及一种人工智能(Artificial Intelligence，AI)模型的处理方法及设备。
背景技术
AI目前在各个领域获得了广泛的应用。通过将AI融入到无线通信领域，可以显著提升吞吐量、降低时延、提升用户容量。然而，在现实网络中，由于网络复杂度的限制、模型传输的限制以及终端不可预测性等原因，使得网络很难实现针对每个终端训练AI模型，网络通常是针对所有终端提供泛化的和蜂窝小区相关的AI模型。但是，泛化的AI模型很难有效地进一步提升通信系统性能，例如，难以提升多输入多输出-信道状态信息(MIMO-CSI)的反馈性能。
发明内容
本申请实施例提供一种AI模型的处理方法及设备，能够解决因AI模型泛化，很难有效地进一步提升通信系统性能的问题。
第一方面，提供了一种AI模型的处理方法，包括：第一通信设备确定第一区域；所述第一通信设备执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
第二方面，提供了一种第一通信设备，包括：确定模块，用于确定第一区域；执行模块，用于执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
第三方面，提供了一种通信设备，该通信设备包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第一方面所述的方法的步骤。
第四方面，提供了一种通信设备，包括处理器及通信接口，其中，所述 处理器用于确定第一区域；以及，执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
第五方面，提供了一种AI模型的处理系统，包括：终端及网络侧设备，所述终端可用于执行如第一方面所述的方法的步骤，所述网络侧设备可用于执行如第一方面所述的方法的步骤。
第六方面，提供了一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如第一方面所述的方法的步骤。
第七方面，提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如第一方面所述的方法的步骤。
第八方面，提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现如第一方面所述的方法的步骤。
在本申请实施例中，第一通信设备确定第一区域，并执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征，由于AI模型与区域特征相适应，因此可以根据不同的区域进行特征训练和更新，有利于提高AI模型使用的准确性和有效性，进一步提升通信系统性能。
附图说明
图1是根据本申请实施例的无线通信系统的示意图；
图2是根据本申请实施例的AI模型的处理方法的示意性流程图；
图3是根据本申请实施例中划分的区域示意图；
图4是根据本申请实施例中UE在不同场景下的行为示意图；
图5是根据本申请实施例的AI模型的处理方法的示意性流程图；
图6是根据本申请实施例的AI模型的处理方法的示意性流程图；
图7是根据本申请实施例的第一通信设备的结构示意图；
图8是根据本申请实施例的通信设备的结构示意图；
图9是根据本申请实施例的终端的结构示意图；
图10是根据本申请实施例的网络侧设备的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行 清楚描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员所获得的所有其他实施例，都属于本申请保护的范围。
本申请的说明书和权利要求书中的术语“第一”、“第二”等是用于区别类似的对象，而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换，以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施，且“第一”、“第二”所区别的对象通常为一类，并不限定对象的个数，例如第一对象可以是一个，也可以是多个。此外，说明书以及权利要求中“和/或”表示所连接对象的至少其中之一，字符“/”一般表示前后关联对象是一种“或”的关系。
值得指出的是，本申请实施例所描述的技术不限于长期演进型(Long Term Evolution，LTE)/LTE的演进(LTE-Advanced，LTE-A)系统，还可用于其他无线通信系统，诸如码分多址(Code Division Multiple Access，CDMA)、时分多址(Time Division Multiple Access，TDMA)、频分多址(Frequency Division Multiple Access，FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access，OFDMA)、单载波频分多址(Single-carrier Frequency Division Multiple Access，SC-FDMA)和其他系统。本申请实施例中的术语“系统”和“网络”常被可互换地使用，所描述的技术既可用于以上提及的系统和无线电技术，也可用于其他系统和无线电技术。以下描述出于示例目的描述了新空口(New Radio，NR)系统，并且在以下大部分描述中使用NR术语，但是这些技术也可应用于NR系统应用以外的应用，如第6代(6th Generation，6G)通信系统。
图1示出本申请实施例可应用的一种无线通信系统的框图。无线通信系统包括终端11和网络侧设备12。其中，终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)或称为笔记本电脑、个人数字助理(Personal Digital Assistant，PDA)、掌上电脑、上网本、超级移动个人计算机(ultra-mobile personal computer，UMPC)、移动上网装置(Mobile Internet Device，MID)、增强现实(augmented reality，AR)/虚拟现实(virtual reality，VR)设备、机器人、可穿戴式设备(Wearable Device)、车载设备(Vehicle User Equipment，VUE)、行人终端(Pedestrian User Equipment，PUE)、智能家居(具有无线通信功能的家居设备，如冰箱、电视、洗衣机或者家具等)、游戏机、个人计算机(personal computer，PC)、柜员机或者自助机等终端侧设备，可穿戴式设备包括：智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装等。需要说明的是，在本申请实施例并不限定终端 11的具体类型。网络侧设备12可以包括接入网设备或核心网设备，其中，接入网设备也可以称为无线接入网设备、无线接入网(Radio Access Network，RAN)、无线接入网功能或无线接入网单元。接入网设备可以包括基站、无线局域网(Wireless Local Area Network，WLAN)接入点或无线保真(Wireless Fidelity，WiFi)节点等，基站可被称为节点B、演进节点B(eNB)、接入点、基收发机站(Base Transceiver Station，BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set，BSS)、扩展服务集(Extended Service Set，ESS)、家用B节点、家用演进型B节点、发送接收点(Transmitting Receiving Point，TRP)或所述领域中其他某个合适的术语，只要达到相同的技术效果，所述基站不限于特定技术词汇，需要说明的是，在本申请实施例中仅以NR系统中的基站为例进行介绍，并不限定基站的具体类型。
下面结合附图，通过一些实施例及其应用场景对本申请实施例提供的AI模型的处理方法进行详细地说明。
如图2所示，本申请实施例提供一种AI模型的处理方法200，该方法可以由第一通信设备执行，换言之，该方法可以由安装在第一通信设备的软件或硬件来执行，该方法包括如下步骤。
S202：第一通信设备确定第一区域。
本申请各个实施例中，第一通信设备可以是移动通信设备，如终端等；还可以是固定通信设备，如网络侧设备，位置固定的终端等。
关于第一区域可以参见图3。可选地，所述第一区域与地理坐标对应；其中，所述第一区域的大小为W×L，L为所述第一区域的长度，W为所述第一区域的宽度，W、L均为正数。
可选地，在一个例子中，所述第一通信设备包括移动通信设备，所述第一通信设备确定第一区域包括：所述移动通信设备确定所述移动通信设备的位置信息；所述移动通信设备根据所述位置信息确定所述第一区域。该实施例例如，终端根据自身的位置信息确定第一区域，即终端处于第一区域内。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，所述第一通信设备确定第一区域包括：所述固定通信设备确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；所述固定通信设备根据所述位置信息确定所述第一区域，所述移动通信设备处于所述第一区域内。该实施例例如，网络侧设备根据终端的位置信息确定第一区域，终端处于第一区域内。
S204：第一通信设备执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
可选地，所述第一区域的特征包括：第二通信设备与处于所述第一区域内的一个或多个所述第一通信设备之间的信道数据；其中，所述第一通信设备包括移动通信设备，所述第二通信设备包括固定通信设备。在图3所示的例子中，第一通信设备包括终端(User Equipment，UE)，第二通信设备包括NR节点(NR Node B，gNB)，第一区域的特征包括：gNB与处于第一区域内的一个或多个UE之间的信道数据。
可选地，进行AI模型训练的范围包括：与固定通信设备的最大通信范围相关的区域。在图3所示的例子中，进行AI模型训练的范围可以是：以gNB为圆心，半径为Dmax的范围，其中，Dmax为gNB与UE的最大通信距离。
可选地，在一个例子中，所述第一通信设备包括移动通信设备，如UE，处于第一区域内的移动通信设备可以使用具有所述第一区域的特征的数据进行AI模型训练，还可以选择与所述第一区域匹配的AI模型执行目标通信业务，该目标通信业务例如，信道预测，信道压缩等。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，如gNB，固定通信设备可以使用具有所述第一区域的特征的数据进行AI模型训练，还可以选择与所述第一区域匹配的AI模型执行目标通信业务，该目标通信业务例如，信道预测，信道压缩等。
在图3所示的例子中，如果gNB使用具有所述第一区域的特征的数据进行AI模型训练，则每个区域可以对应一个训练完成的AI模型；如果UE使用具有所述第一区域的特征的数据进行AI模型训练，由于第一区域内可能存在多个UE，则每个区域可以对应一个或多个训练完成的AI模型，例如，每个UE独立训练完成一个AI模型；又例如，多个UE联合训练出一个AI模型。
本申请实施例可以基于区域进行AI模型的训练，这样，被训练出的AI模型具有区域的限制，只有当移动通信设备(如接收端或发送端)进入相关区域的时候，与此区域相关的AI模型被接收端或发送端所使用，从而提高对AI模型使用的准确性和有效性。
本申请实施例提供的AI模型的处理方法，第一通信设备确定第一区域，并执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征，由于AI模型具有区域特征的限制，有利于提高AI模型使用的准确性和有效性，进一步提升通信系统性能。
实施例200主要介绍了本申请实施例的核心思想，以下将分多个方案，对本申请实施例的实施细节进行详细说明。
方案一：AI模型的训练方法
在大多数情况下，AI监督学习(Supervised Learning)模型(简称AI模型)可以通过一个概率分布函数p(y|x；w)来表示，其中，x是已知数据向量(也可以被称为数据标签，即，Label)，y是AI模型的输入数据，w是AI模型权值向量或系数向量，需要通过AI模型训练获取。
值得注意的是，(y,x)可以被视为AI模型训练的训练数据，被表示为
其中，(yn,xn)是第n对输入输出数据，N是训练数据集中的训练数据总数。
通过简单地使用最大似然原理(Maximum Likelihood)，即，使用训练数据和模型预测之间的交叉熵(Cross-Entropy)作为成本函数(Cost Function)，AI神经网参数w的训练可以通过求解成本函数J(w)的最小值来获取，其中，成本函数J(w)被表示为
如果假设训练数据集可以被划分为训练数据子集，并且第k个数据子集与参数zk相关联，那么训练数据集可以被表示为
其中，z可以被视为辅助参数向量，z＝{z1,z2,…,zK}，训练数据子集(Xk，Yk)可以被表示为
其中，
当数据子集(Xk,Yk)与数据子集(Xi,Yi)完全独立的情况下，即，其中k≠i，概率分布函数p(y|x；w)可以被表示为
因此，训练数据和训练模型之间的交叉熵可以被表示为
如果辅助参数向量z是已知的或可以被估计的，则与辅助参数zk相对应的AI模型可以被独立训练，其成本函数可以被简化为
其中，k＝1,2,…,K，K是独立AI模型的总数。
值得注意的是，使用公式(0)来训练的AI模型方法要简单的多，且AI模型的训练会更加准确，推理性能也会大幅提高。
方案二：根据区域区分AI模型的方法
无线信道受UE端的影响远远大于网络端(如，gNB)的影响。一般情况下，UE端的行为是不可预测的。当UE端从一个位置移动到另一个位置，UE端与网络端间的多输入多输出(Multiple Input Multiple Output，MIMO)信道会产生变化，且移动距离越远MIMO信道变化越大。如图4中的场景-1，UE端与网络端收发天线的入射角和发射角发生变化，从而造成MIMO信道产生变化。另外，如图4中的场景-2，在同样的地理位置，UE端由于天线方向旋转或摇摆的原因，UE端的入射角和发射角也会发生变化，从而造成MIMO信道产生变化。因此，MIMO信道受UE端影响因素将远远大于受网络端影响因素。
空间信道模型(Space Channel Model，SCM)的生成过程分为多个步骤，前几个步骤主要产生静态或半静态的与MIMO信道相关联的参数，即，在一定时间范围内，不会由于时间或环境的变化，MIMO信道产生变化。最后几个步骤针对簇n以及第u个接收天线单元和第s个发射天线单元对生成相关信道系数。在相关信道系数生成过程中，对于第n簇和第m射线，MIMO信道系数由下公式给出:
其中，Frx，u，θ和Frx，u，φ分别为接收天线单元u在球面基矢量方向上和的场模式，Ftx,s,θ和Ftx,s,φ分别为发射天线单元s在球面基矢量方向上和的场模式。是接收天线单元u的位置向量，是发射天线单元s的位置向量，kn,m是线性尺度的交叉极化功率比(Cross Polarisation Power Ratio)。
值得注意的是，第二项与初始值相关联，可以通过对AI模型的训练简单且很有效地获取。第一项，第四项和第六项与UE端相关联，直接与UE端行为有关，必须通过收集大量的AI训练数据集，才能对AI模型进行有效的训练。最后，第三项和第五项与网络端相关联，其行为相对比较稳定，可以通过UE的地理位置，对AI模型进行区分。
可选的，第六项与多普勒效应相关联，在终端发生运动的情况下，可能会产生多普勒效应，通过第六项可以对多普勒效应的影响进行预测和修正。
本申请实施例通过对网络区域的划分，即对AI模型在一定程度上的限制和区分，从而达到简化每个AI模型的复杂度，并提高对AI模型使用的准确性和有效性。
网络可以通过区域标识(即，区域标识(Identifier，ID)，或Zone-ID)机制对地理坐标进行区域划分。网络根据gNB的地理位置决定与最大通信范围Dmax相关的网络覆盖区域，由Zone-ID来表示。如图3所示，每个区域的大小为W×L，其中，L为区域长度值，W为区域宽度值。在AI模型训练过程中，训练数据集可以通过网络覆盖区域中的Zone-ID来区分，因此，被训练而获取的AI模型将拥有区域Zone-ID的特征。
值得注意的是，AI模型训练过程可以由UE端完成，也可以由网络侧设备完成。
在蜂窝小区中，由于频谱的不同(如，频分复用(Frequency Division Duplex，FDD)机制的场景)，发送和接收电路的差异，上行链路(即，Up-Link)和下行链路(即，Down-Link)的MIMO信道是有区别的。可选地，本申请各个实施例中提到的AI模型可以被细分为上行链路AI模型和下行链路AI模型。
值得注意的是，如果蜂窝小区采用时分复用(Time Division Duplex，TDD)机制，则AI模型就没有必要被细分，即，使用上下行链路泛化的AI模型即可。
在此，我们假设UE端或网络端获取AI训练数据集(Training Data Set)，即，{D}n。根据上下行链路和蜂窝小区区域Zone-ID，训练数据集被区分为相应的数据子集，即
其中，上标UL和DL分别表示上行链路和下行链路，n是网络蜂窝小区 的ID，k是第k个覆盖区域的索引，索引k由Zone-ID映射转换而来，即，
其中，z是Zone-ID，是蜂窝小区覆盖区域的索引映射转换函数，由高层配置决定。
另外，整个数据集可以被表示为
因此，通过相应的数据子集或分别获取的上下行链路的训练AI模型为
或
其中，
或
是AI神经网函数。
根据最大通信范围Dmax，如果每个蜂窝小区覆盖K个区域作为AI模型训练的话，UE端第n个网络小区所要求的AI模型可以由表1表示。
表1:针对第n个网络小区的AI模型表
方案三：AI模型训练过程
在实施例200的基础上，S204中第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练可以分以下四个例子。
可选地，在一个例子中，所述第一通信设备包括移动通信设备，如UE，所述AI模型包括下行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：所述移动通信设备在所述第一区域内接收下行参考信号，并根据所述下行参考信号得到具有所述第一区域的特征的数据；所述移动通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。该实施例可以参见后续选项一的介绍。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，如网络侧设备，所述AI模型包括上行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：所述固定通信设备接收上行参考信号，并根据所述上行参考信号得到具有所述第一区域的特征的数据，所述上行参考信号是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；所述固定通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。该实施例可以参见后续选项二的介绍。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，如网络侧设备，所述AI模型包括下行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：所述固定通信设备接收具有所述第一区域的特征的数据，所述具有所述第一区域的特征的数据是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；所述固定通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。该实施例可以参见后续选项三的介绍。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，如网络侧设备，所述AI模型包括下行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：所述固定通信设备接收第一数据，所述第一数据是第二通信设备发送的，所述第二通信设备包括移动通信设备；所述固定通信设备根据所述移动通信设备的位置信息确定所述第一数据具有所述第一区域的特征；所述固定通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。该实施例可以参见后续选项四的介绍。
AI模型既可以通过UE端训练获取，也可以通过网络端训练获取，如表2所示。
表2:AI模型训练类型
选项一：UE端执行AI模型训练过程
UE端执行AI模型训练过程一般只是针对下行链路MIMO信道的。
UE端会根据全球导航卫星系统(Global Navigation Satellite System，GNSS)或5G-NR定位(即，5G-NR Positioning)等方法对UE端进行时时定位和跟踪。从而，UE端获取自己的位置坐标信息。
可选地，网络端对UE端进行时时定位和跟踪，并把定位结果通知UE端，让UE端获取自己的位置坐标信息。
当UE端进入连接态(Connected Mode)，UE端接收下行参考信号(即，Reference Signal)，通过最小均方误差(Minimum Mean Squared Error，MMSE)等手段对MIMO信道数据进行估计并储存。
可选地，下行参考信号可以是高层配置的周期性信号，也可以是高层配置但通过媒体接入控制控制单元(Media Access Control Control Element，MAC CE)或层一(L1)信令触发的周期性或非周期性信号，也可以是直接发送非周期的L1参考信号，如，信道状态信息参考信号(Channel State Information-Reference Signal，CSI-RS)。
值得注意的是，UE端储存MIMO信道数据主要目的是为了决定AI训练数据即，
其中，是与第n个gNB相关的下行链路的MIMO信道响应(即，Channel Response)矩阵，t是信道响应的时间，可以使用正交频分复用(Orthogonal frequency division multiplex，OFDM)符号(即，OFDM Symbol)也可以通过时隙(即，Time Slot)参数来代替，k是覆盖区域的索引，l是AI训练数据子集中的第l个数据。
因此，下行链路的数据子集可以被表示为
其中，L是针对第k个覆盖区域的AI训练数据总数。
UE端通过所收集的AI训练数据子集来训练下行AI模型
以下将介绍网络端执行AI模型训练过程，网络端执行AI模型训练过程可以针对上行链路MIMO信道，也可以针对下行链路MIMO信道，分以下选项二，选项三和选项四。
选项二：网络端执行AI模型训练过程，针对上行链路MIMO信道的场景。
UE端会根据全球导航卫星系统(Global Navigation Satellite System，GNSS)或5G-NR定位(即，5G-NR Positioning)等方法对UE端进行时时定 位和跟踪，并把定位结果通知网络端，让网络端获取UE端的位置坐标信息。
可选地，网络端会通过5G-NR定位等方法对UE端进行时时定位和跟踪，从而，网络端获取UE端的位置坐标信息。
可选地，网络端会通过高层针对UE端配置上行链路的参考信号，如探测参考信号(Sounding Reference Signal，SRS)。上行参考信号可以是高层配置的周期性信号，也可以是高层配置但通过MAC-CE或L1信令触发的周期性或非周期性信号，也可以是直接发送非周期的L1参考信号。
网络端接收参考信号，通过MMSE等手段对MIMO信道数据进行估计并储存。
值得注意的是，网络端储存MIMO信道数据主要目的是为了决定AI训练数据即，
其中，是与第n个gNB相关的上行链路的信道响应矩阵，t是信道响应的时间，可以使用正交频分复用(Orthogonal Frequency Division Multiplexing，OFDM)符号(即，OFDM Symbol)也可以通过时隙(即，Time Slot)参数来代替，k是覆盖区域的索引，l是AI训练数据子集中的第l个数据。
值得注意的是，上行链路的数据子集可以被表示为
其中，L是针对第k个覆盖区域的AI训练数据总数。
网络侧通过所收集的AI训练数据子集来训练上行AI模型
以下将介绍网络端执行AI模型训练过程针对下行链路MIMO信道的场景:这种情况下，AI模型训练过程可以由以下两个选项来完成。
选项三：UE端会根据GNSS或5G-NR定位(即，5G-NR Positioning)等方法对UE端进行时时定位和跟踪。从而，UE端获取自己的位置坐标信息。
同样地，当UE端进入连接态(Connected Mode)，UE端接收下行参考信号(即，Reference Signal)，通过MMSE等手段对MIMO信道数据进行估计并储存。
可选地，下行参考信号可以是高层配置的周期性信号，也可以是高层配置但通过MAC-CE或L1信令触发的周期性或非周期性信号，也可以是直接发送L1参考信号，如CSI参考信号(CSI Reference Signal，CSI-RS)。
值得注意的是，UE端在储存MIMO信道数据的同时，会决定AI训练数据由公式(2)表示。
同样地，下行链路的数据子集可以被公式(3)表示。
UE端通过网络端指示的上行信道资源(如，PUSCH)，将下行链路的数据子集汇报给网络端。
网络端通过所收集的AI训练数据子集来训练下行AI模型
选项四：网络端会通过5G-NR定位等方法对UE端进行时时定位和跟踪。从而，网络端获取UE端的位置坐标信息。
当UE端进入连接态(Connected Mode)，UE端接收下行参考信号(即，Reference Signal)，通过MMSE等手段对MIMO信道数据进行估计并储存。
可选地，下行参考信号可以是高层配置的周期性信号，也可以是高层配置但通过MAC-CE或L1信令触发的周期性或非周期性信号，也可以是直接发送L1参考信号，如，CSI-RS。
值得注意的是，UE端在储存MIMO信道数据同时，会决定AI训练数据即，
其中，是与第n个网络端相关的下行链路的信道响应(即，Channel Response)矩阵，t是信道响应的时间，可以使用OFDM符号(即，OFDM Symbol)也可以通过时隙(即，Time Slot)参数来代替，l是AI训练数据子集中的第l个数据。
值得注意的是，由于UE端没有相关的位置坐标信息，UE端不区分区域相关的数据子集。下行链路的数据子集可以被表示为
其中，L是针对第n个gNB相关的AI训练数据总数。
UE端通过网络端指示的上行信道资源，如，物理上行共享信道(Physical Uplink Shared Channel，PUSCH)，将下行链路的数据子集汇报给网络端。
由于网络端已知UE端ID和其坐标位置，网络端根据UE端的坐标位置区分与第k区域相关联的AI训练数据子集同样地，数据子集可以被公式(3)表示。
网络端通过所收集的AI训练数据子集来训练下行AI模型
方案四：AI模型选择过程
AI模型选择过程与AI模型使用场景直接关联。可选地，根据信道预测 (如，根据CSI信号对信道预测)的数据集训练的AI模型仅仅涉及到UE端或网络端。可选地，根据信道压缩(如，根据CSI信号估计的信道压缩)的数据集训练的AI模型会同时涉及到UE端和网络端。
以下将介绍AI模型应用在信道预测中的场景。
信道预测的定义是，UE端或网络端通过现有时间t获取的信道(如，UE端根据CSI-RS信号估计下行信道或网络端根据SRS信号估计上行信道)来预测时间t+Δ获取的信道，其中，Δ是信道预测的时间长度。
可选地，UE端通过时间t的下行信道来预测时间t+Δ的下行信道UE端会把预测的下行信道使用上行链路的信道资源，如，物理上行控制信道(Physical Uplink Control Channel，PUCCH)，PUSCH等汇报给网络端，从而网络端能够有效地决定下行链路的大规模(Massive)MIMO传输方式。
可选地，网络端通过UE端在时间t使用上行链路的信道资源(如，PUCCH，PUSCH等)汇报的下行信道来预测时间t+Δ的下行信道网络端根据预测的信道能够有效地决定下行链路的Massive MIMO传输方式。
可选地，网络端通过时间t的上行信道来预测时间t+Δ的上行信道网络端根据预测的上行信道来决定上行链路的Massive MIMO传输方式，并使用信令，如，物理下行控制信道(Physical Downlink Control Channel，PDCCH)，物理下行共享信道(Physical Downlink Shared Channel，PDSCH)等，通过UE端进行上行链路的Massive MIMO传输。
在实施例200的基础上，S204中第一通信设备确定第一区域，并选择与所述第一区域匹配的AI模型执行目标通信业务可以分以下两个例子。
可选地，在一个例子中，所述第一通信设备包括移动通信设备，所述第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：所述移动通信设备确定所述移动通信设备的位置信息；所述移动通信设备根据所述位置信息确定所述第一区域；所述移动通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择与所述索引k对应的AI模型进行信道预测，K是正整数。
可选地，所述AI模型包括上行链路AI模型，所述信道预测包括预测时刻t+Δ的信道，所述方法还包括：所述移动通信设备接收时刻t的信道，所述时刻t的信道是第二通信设备发送的，所述第二通信设备包括固定通信设备。
可选地，所述AI模型包括上行链路AI模型或下行链路AI模型。该实施例可以参见后续选项一和选项三的介绍。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，所述第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：所述固定通信设备确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；所述固定通信设备根据所述位置信息确定所述第一区域；所述固定通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI模型进行信道预测，K是正整数。
可选地，所述AI模型包括上行链路AI模型，所述信道预测包括预测时刻t+Δ的信道，所述方法还包括：所述固定通信设备接收时刻t的信道，所述时刻t的信道是所述移动通信设备发送的。
可选地，所述AI模型包括上行链路AI模型或下行链路AI模型。该实施例可以参见后续选项二和选项四的介绍。
AI模型在信道预测的场景中，UE端或网络端选择上行和下行链路的AI模型的方法有四种选项，如表3所示。
表3:针对信道预测场景中的AI模型选择方法。
选项一：UE端对自己进行定位，或通过网络端的定位获取UE端自身的位置坐标。UE端根据自身的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k。UE端在训练好的K个上行AI模型中，即，中，选择第k个AI模型。根据AI模型UE端预测信道
值得注意的是，选项一的应用场景是，网络端可以将上行信道告诉UE端，根据AI模型UE端预测信道
选项二：网络端对UE端进行定位，或通过UE端的定位获取UE端的位置坐标。网络端根据UE端的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；网络端在训练好的K个上行AI模型中，即，中，选择第k个AI模型。根据AI模型网络端预测信道
选项三：UE端对自己进行定位，或通过网络端的定位获取UE端自身的位置坐标；UE端根据自身的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；UE端在训练好的K个下行AI模型中，即，中，选择第k个AI模型。根据AI模型UE端预测信道
选项四：网络端对UE端进行定位，或通过UE端的定位获取UE端的位 置坐标；网络端根据UE端的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；网络端在训练好的K个下行AI模型中，即，中，选择第k个AI模型。根据AI模型网络端预测信道
值得注意的是，选项四的应用场景是，UE端可以将下行信道告诉网络端，根据AI模型网络端预测信道
以下将介绍AI模型应用在信道压缩中的场景。
信道压缩的定义是，UE端通过现有时间t获取的信道(如，UE端根据CSI-RS信号估计下行信道)。UE端通过AI神经网的编码器(即，Encoder)对信道进行压缩，并报告给网络端。网络端通过AI神经网的解码器(即，Decoder)对接收信号进行解码，恢复信道。
有效地，AI模型由AI编码器模型和AI解码器模型组成，分别由和或和表示，其中

或

在实施例200的基础上，S204中第一通信设备确定第一区域，并选择与所述第一区域匹配的AI模型执行目标通信业务可以分以下两个例子。
可选地，在一个例子中，所述第一通信设备包括移动通信设备，第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：所述移动通信设备确定所述移动通信设备的位置信息；所述移动通信设备根据所述位置信息确定所述第一区域；所述移动通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；所述移动通信设备向第二通信设备发送所述压缩信道数据，所述第二通信设备包括固定通信设备。
可选地，所述方法还包括：所述移动通信设备向所述固定通信设备发送所述AI压缩模型对应的AI解码模型；其中，所述AI解码模型用于解压所述压缩信道数据。
可选地，所述移动通信设备确定所述移动通信设备的位置信息包括：所述移动通信设备自主确定所述移动通信设备的位置信息；或者，所述移动通 信设备接收所述移动通信设备的位置信息，所述位置信息所述固定通信设备确定的。该实施例可以参见后续选项一和选项二的介绍。
可选地，在一个例子中，所述第一通信设备包括固定通信设备，第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：所述固定通信设备确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；所述固定通信设备根据所述位置信息确定所述第一区域；所述固定通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；所述固定通信设备向所述移动通信设备发送所述压缩信道数据。
可选地，所述方法还包括：所述固定通信设备向所述移动通信设备发送所述AI压缩模型对应的AI解码模型；其中，所述AI解码模型用于解压所述压缩信道数据。
可选地，所述固定通信设备确定第二通信设备的位置信息包括：所述固定通信设备自主确定所述第二通信设备的位置信息；或者，所述固定通信设备接收所述第二通信设备的位置信息，所述位置信息所述第二通信设备确定的。该实施例可以参见后续选项三和选项四的介绍。
在信道压缩中的场景中，AI模型的选择是根据UE端执行信道压缩还是网络端执行信道压缩决定的。AI模型的选择方法有四种选项，如表4所示。
表4:针对信道压缩场景中的AI模型选择方法。
选项一：UE端自身定期或根据需求执行定位或更新UE端自身地理位置。UE端根据自身地理位置，选择相应区域(即，Zone-ID)的AI编码器模型，即对信道进行压缩。UE端通过物理信道(如，PUCCH或PUSCH)发送压缩信道数据包给网络端，同时还可以通知网络端对应的AI解码器模型网络端通过AI解码器模型对接收的数据包进行解码恢复在时间t的信道。
可选地，在选项一中，UE端发送压缩信道数据包给网络端之前或之后，UE端通知网络端对应的AI解码器模型
可选地，在选项一中，UE端可以通过AI模型的索引k通知网络端对应的AI解码器模型。
选项二：网络端对UE端定期或根据需求执行定位或更新UE端的地理位置。网络端通知UE端，更新UE端的地理位置。UE端根据自身地理位置，选择相应区域(即，Zone-ID)的AI编码器模型，即对信道进行压缩。UE端通过物理信道(如，PUCCH或PUSCH)发送压缩信道数据包给网络端，同时可以通知网络端对应的AI解码器模型，即网络端通过AI解码器模型对接收的数据包进行解码恢复在时间t的信道。
可选地，在选项二中，UE端发送压缩信道数据包给网络端之前或之后，通知网络端对应的AI解码器模型
可选地，在选项二中，UE端可以通过AI模型的索引k通知网络端对应的AI解码器模型。
选项三：UE端自身定期或根据需求的执行定位或更新UE端自身地理位置。UE端通知网络端，更新UE端的地理位置。网络端根据UE端的地理位置，选择相应区域(即，Zone-ID)的AI编码器模型，即对信道进行压缩。网络端通过物理信道(如，PDCCH或PDSCH)发送压缩信道数据包给UE端，同时可以通知UE端对应的AI解码器模型，即UE端通过AI解码器模型对接收的数据包进行解码恢复在时间t的信道。
可选地，在选项三中，网络端发送压缩信道数据包给UE端之前或之后，通知UE端对应的AI解码器模型
可选地，在选项三中，网络端可以通过AI模型的索引k通知UE端对应的AI解码器模型。
选项四：网络端对UE端定期或根据需求执行定位或更新UE端的地理位置。网络端根据UE端的地理位置，选择相应区域(即，Zone-ID)的AI编码器模型，即对信道进行压缩。网络端通过物理信道(如，PDCCH或PDSCH)发送压缩信道数据包给UE端，同时可以通知UE端对应的AI解码器模型，即UE端通过AI解码器模型对接收的数据包进行解码恢复在时间t的信道。
可选地，在选项四中，网络端发送压缩信道数据包给UE端之前或之后，通知UE端对应的AI解码器模型
可选地，在选项四中，网络端可以通过AI模型的索引k通知UE端对应的AI解码器模型。
以下将介绍AI模型在信道预测和信道压缩中的场景。更有效地，UE端 或网络端可以根据UE端具体地理位置(区域)，同时对两个以上不同类型的AI模型进行选择。
在实施例200的基础上，所述AI模型包括AI预测模型以及AI压缩模型，S204中所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：所述第一通信设备选择所述AI预测模型预测时刻t+Δ的信道；所述第一通信设备选择所述AI压缩模型对所述时刻t+Δ的信道进行压缩，得到压缩信道数据；所述第一通信设备向第二通信设备发送所述压缩信道数据。
可选地，所述方法还包括：所述第一通信设备向所述第二通信设备发送所述AI压缩模型对应的AI解码模型；其中，所述AI解码模型用于解压所述压缩信道数据。
可选地，所述第一通信设备包括固定通信设备，所述第二通信设备包括移动通信设备；或者，所述第一通信设备包括移动通信设备，所述第二通信设备包括固定通信设备。该实施例可以参见后续选项一和选项二的介绍。
选项一：UE端根据UE端具体地理位置选择针对下行信道预测场景中的AI模型，即同时UE端还可以选择针对下行信道压缩场景中的AI编码器模型，即UE端通过下行信道预测相关的AI模型根据下行信道对信道进行预测。
可选地，UE端通过下行信道压缩相关的AI编码器模型对预测信道进行压缩。UE端通过上行物理信道(如，PUCCH或PUSCH)发送压缩信道数据包给网络端，同时可以通知网络端对应的AI解码器模型，即网络端通过AI解码器模型对接收的数据包进行解码恢复在时间t+Δ的下行信道
可选地，在选项一中，UE端发送压缩信道数据包给网络端之前或之后，UE端通知网络端对应的AI解码器模型
可选地，在选项一中，UE端可以通过AI模型的索引k通知网络端对应的AI解码器模型。
选项二：网络端根据UE端具体地理位置选择针对上行信道预测场景中的AI模型，即同时网络端还选择针对上行信道压缩场景中的AI编码器模型，即网络端通过上行信道预测相关的AI模型根据上行信道对信道进行预测。
可选地，网络端通过上行信道压缩相关的AI编码器模型对预测信道进行压缩。网络端通过下行物理信道(如，PDCCH或PDSCH) 发送压缩信道数据包给UE端，同时可以通知UE端对应的AI解码器模型，即UE端通过AI解码器模型对接收的数据包进行解码恢复在时间t+Δ的上行信道
可选地，在选项二中，网络端发送压缩信道数据包给UE端之前或之后，通知UE端对应的AI解码器模型
可选地，在选项二中，网络端可以通过AI模型的索引k通知UE端对应的AI解码器模型。
可选地，UE端或网络端可以自由地选择信道预测相关的AI模型和信道压缩相关的AI模型进行搭配，从而有效地实现信道预测和信道压缩相关过程。
为详细说明本申请实施例提供的AI模型的处理方法，以下将结合几个具体的实施例进行说明。
实施例一
本实施例是关于由UE执行AI模型训练，信道预测和信道压缩过程，如图5所示，该实施例包括如下步骤：
步骤1：UE端训练下行信道预测AI模型UE端训练下行信道压缩AI编码器模型和AI解码器模型
步骤2：UE端将AI解码器模型报告给gNB。
步骤3：UE端对自己定位，或gNB对UE端定位并通知UE端相关UE端的地理位置。
步骤4：UE端根据自身的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；UE端在训练好的K个AI模型中，即，中，选择第k个AI模型
步骤5：根据AI模型和时间t以前的p个下行信道，即UE端预测在时间t+Δ的信道，即
步骤6：UE端根据自身的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；UE端在训练好的K个AI模型中，即，中，选择第k个AI编码器模型
步骤7：根据AI编码器模型UE端对预测的下行信道进行压缩，生成预测信道数据包。其中包括AI解码器模型相关的索引k。
步骤8：UE端通过PUCCH或PUSCH发送预测信道数据包给gNB。
步骤9：根据相关的索引k，gNB使用AI解码器模型对预测信道数据包进行解码，获取预测信道
实施例二：
本实施例是关于由gNB执行AI模型训练，信道预测和信道压缩过程，如图6所示，该实施例包括如下步骤：
步骤1：gNB训练下行信道预测AI模型gNB训练下行信道压缩AI编码器模型和AI解码器模型
步骤2：gNB将下行信道预测AI模型和AI编码器模型告诉给UE端。
步骤3：UE端对自己定位，或gNB对UE端定位并通知UE端相关UE端的地理位置。
步骤4：UE端根据自身的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；UE端在训练好的K个AI模型中，即，中，选择第k个AI模型
步骤5：根据AI模型和时间t以前的p个下行信道，即UE端预测在时间t+Δ的信道，即
步骤6：UE端根据自身的位置坐标确定Zone-ID，即z，并通过公式(1)获取AI模型对应区域索引k；UE端在训练好的K个AI模型中，即，中，选择第k个AI编码器模型
步骤7：根据AI编码器模型UE端对预测的下行信道进行压缩，生成预测信道数据包。其中包括AI解码器模型相关的索引k。
步骤8：UE端通过PUCCH或PUSCH发送预测信道数据包给gNB。
步骤9：根据相关的索引k，gNB使用AI解码器模型对预测信道数据包进行解码，获取预测信道
图7是根据本申请实施例的第一通信设备的结构示意图，该第一通信设备可以对应于其他实施例中的终端或网络侧设备。如图7所示，第一通信设备700包括如下模块。
确定模块702，可以用于确定第一区域。
执行模块704，可以用于执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
本申请实施例中，第一通信设备确定第一区域，并执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征，由于AI模型具有区域特征的限制，有利于提高AI模型使用的准确性和有效性，进一步提升通信系统性能。
可选地，作为一个实施例，所述第一通信设备700包括移动通信设备，所述确定模块702，用于：确定所述移动通信设备的位置信息；根据所述位置信息确定所述第一区域。
可选地，作为一个实施例，所述第一通信设备700包括固定通信设备，所述确定模块702，用于：确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；根据所述位置信息确定所述第一区域，所述移动通信设备处于所述第一区域内。
可选地，作为一个实施例，所述第一通信设备700包括移动通信设备，所述AI模型包括下行链路AI模型，所述执行模块704，用于：在所述第一区域内接收下行参考信号，并根据所述下行参考信号得到具有所述第一区域的特征的数据；使用所述具有所述第一区域的特征的数据进行AI模型训练。
可选地，作为一个实施例，所述第一通信设备700包括固定通信设备，所述AI模型包括上行链路AI模型，所述执行模块704，用于：接收上行参考信号，并根据所述上行参考信号得到具有所述第一区域的特征的数据，所述上行参考信号是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；使用所述具有所述第一区域的特征的数据进行AI模型训练。
可选地，作为一个实施例，所述第一通信设备700包括固定通信设备，所述AI模型包括下行链路AI模型，所述执行模块704，用于：接收具有所述第一区域的特征的数据，所述具有所述第一区域的特征的数据是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；使用所述具有所述第一区域的特征的数据进行AI模型训练。
可选地，作为一个实施例，所述第一通信设备700包括固定通信设备，所述AI模型包括下行链路AI模型，所述执行模块704，用于：接收第一数据，所述第一数据是第二通信设备发送的，所述第二通信设备包括移动通信设备；根据所述移动通信设备的位置信息确定所述第一数据具有所述第一区域的特征；使用所述具有所述第一区域的特征的数据进行AI模型训练。
可选地，作为一个实施例，所述第一通信设备700包括移动通信设备；所述确定模块702，用于确定所述移动通信设备的位置信息；根据所述位置信息确定所述第一区域；所述执行模块704，用于根据所述第一区域的索引k， 在训练完成的K个AI模型中选择与所述索引k对应的AI模型进行信道预测，K是正整数。
可选地，作为一个实施例，所述第一通信设备700包括固定通信设备；所述确定模块702，用于确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；根据所述位置信息确定所述第一区域；所述执行模块704，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI模型进行信道预测，K是正整数。
可选地，作为一个实施例，所述第一通信设备700包括移动通信设备；所述确定模块702，用于确定所述移动通信设备的位置信息；根据所述位置信息确定所述第一区域；所述执行模块704，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；向第二通信设备发送所述压缩信道数据，所述第二通信设备包括固定通信设备。
可选地，作为一个实施例，所述第一通信设备700包括固定通信设备；所述确定模块702，用于确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；根据所述位置信息确定所述第一区域；所述执行模块704，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；向所述移动通信设备发送所述压缩信道数据。
可选地，作为一个实施例，所述AI模型包括AI预测模型以及AI压缩模型，所述执行模块704，用于：选择所述AI预测模型预测时刻t+Δ的信道；选择所述AI压缩模型对所述时刻t+Δ的信道进行压缩，得到压缩信道数据；向第二通信设备发送所述压缩信道数据。
根据本申请实施例的第一通信设备700可以参照对应本申请实施例的方法200的流程，并且，该第一通信设备700中的各个单元/模块和上述其他操作和/或功能分别为了实现方法200中的相应流程，并且能够达到相同或等同的技术效果，为了简洁，在此不再赘述。
本申请实施例中的第一通信设备可以是电子设备，例如具有操作系统的电子设备，也可以是电子设备中的部件，例如集成电路或芯片。该电子设备可以是终端，也可以为除终端之外的其他设备。示例性的，终端可以包括但不限于上述所列举的终端11的类型，其他设备可以为服务器、网络附属存储器(Network Attached Storage，NAS)等，本申请实施例不作具体限定。
本申请实施例提供的第一通信设备能够实现图2至图6的方法实施例实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
可选的，如图8所示，本申请实施例还提供一种通信设备800，包括处 理器801和存储器802，存储器802上存储有可在所述处理器801上运行的程序或指令，例如，该通信设备800为终端时，该程序或指令被处理器801执行时实现上述AI模型的处理方法实施例的各个步骤，且能达到相同的技术效果。该通信设备800为网络侧设备时，该程序或指令被处理器801执行时实现上述AI模型的处理方法实施例的各个步骤，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供一种终端，包括处理器和通信接口，所述处理器用于确定第一区域；以及，执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。该终端实施例与上述终端侧方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该终端实施例中，且能达到相同的技术效果。具体地，图9为实现本申请实施例的一种终端的硬件结构示意图。
该终端900包括但不限于：射频单元901、网络模块902、音频输出单元903、输入单元904、传感器905、显示单元906、用户输入单元907、接口单元908、存储器909以及处理器910等中的至少部分部件。
本领域技术人员可以理解，终端900还可以包括给各个部件供电的电源(比如电池)，电源可以通过电源管理系统与处理器910逻辑相连，从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。图9中示出的终端结构并不构成对终端的限定，终端可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置，在此不再赘述。
应理解的是，本申请实施例中，输入单元904可以包括图形处理单元(Graphics Processing Unit，GPU)9041和麦克风9042，图形处理器9041对在视频捕获模式或图像捕获模式中由图像捕获装置(如摄像头)获得的静态图片或视频的图像数据进行处理。显示单元906可包括显示面板9061，可以采用液晶显示器、有机发光二极管等形式来配置显示面板9061。用户输入单元907包括触控面板9071以及其他输入设备9072中的至少一种。触控面板9071，也称为触摸屏。触控面板9071可包括触摸检测装置和触摸控制器两个部分。其他输入设备9072可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆，在此不再赘述。
本申请实施例中，射频单元901接收来自网络侧设备的下行数据后，可以传输给处理器910进行处理；另外，射频单元901可以向网络侧设备发送上行数据。通常，射频单元901包括但不限于天线、放大器、收发信机、耦合器、低噪声放大器、双工器等。
存储器909可用于存储软件程序或指令以及各种数据。存储器909可主 要包括存储程序或指令的第一存储区和存储数据的第二存储区，其中，第一存储区可存储操作系统、至少一个功能所需的应用程序或指令(比如声音播放功能、图像播放功能等)等。此外，存储器909可以包括易失性存储器或非易失性存储器，或者，存储器909可以包括易失性和非易失性存储器两者。其中，非易失性存储器可以是只读存储器(Read-Only Memory，ROM)、可编程只读存储器(Programmable ROM，PROM)、可擦除可编程只读存储器(Erasable PROM，EPROM)、电可擦除可编程只读存储器(Electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory，RAM)，静态随机存取存储器(Static RAM，SRAM)、动态随机存取存储器(Dynamic RAM，DRAM)、同步动态随机存取存储器(Synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM，DDRSDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(Synch link DRAM，SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM，DRRAM)。本申请实施例中的存储器909包括但不限于这些和任意其它适合类型的存储器。
处理器910可包括一个或多个处理单元；可选的，处理器910集成应用处理器和调制解调处理器，其中，应用处理器主要处理涉及操作系统、用户界面和应用程序等的操作，调制解调处理器主要处理无线通信信号，如基带处理器。可以理解的是，上述调制解调处理器也可以不集成到处理器910中。
其中，处理器910，可以用于确定第一区域；以及，执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征
本申请实施例中，终端确定第一区域，并执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征，由于AI模型具有区域特征的限制，有利于提高AI模型使用的准确性和有效性，进一步提升通信系统性能。
本申请实施例提供的终端900还可以实现上述AI模型的处理方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供一种网络侧设备，包括处理器和通信接口，所述处理器用于确定第一区域；以及，执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。该网络侧设备实 施例与上述网络侧设备方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该网络侧设备实施例中，且能达到相同的技术效果。
具体地，本申请实施例还提供了一种网络侧设备。如图10所示，该网络侧设备1000包括：天线101、射频装置102、基带装置103、处理器104和存储器105。天线101与射频装置102连接。在上行方向上，射频装置102通过天线101接收信息，将接收的信息发送给基带装置103进行处理。在下行方向上，基带装置103对要发送的信息进行处理，并发送给射频装置102，射频装置102对收到的信息进行处理后经过天线101发送出去。
以上实施例中网络侧设备执行的方法可以在基带装置103中实现，该基带装置103包括基带处理器。
基带装置103例如可以包括至少一个基带板，该基带板上设置有多个芯片，如图10所示，其中一个芯片例如为基带处理器，通过总线接口与存储器105连接，以调用存储器105中的程序，执行以上方法实施例中所示的网络设备操作。
该网络侧设备还可以包括网络接口106，该接口例如为通用公共无线接口(common public radio interface，CPRI)。
具体地，本发明实施例的网络侧设备1000还包括：存储在存储器105上并可在处理器104上运行的指令或程序，处理器104调用存储器105中的指令或程序执行图7所示各模块执行的方法，并达到相同的技术效果，为避免重复，故不在此赘述。
本申请实施例还提供一种可读存储介质，所述可读存储介质上存储有程序或指令，该程序或指令被处理器执行时实现上述AI模型的处理方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
其中，所述处理器为上述实施例中所述的终端中的处理器。所述可读存储介质，包括计算机可读存储介质，如计算机只读存储器ROM、随机存取存储器RAM、磁碟或者光盘等。
本申请实施例另提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现上述AI模型的处理方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
应理解，本申请实施例提到的芯片还可以称为系统级芯片，系统芯片，芯片系统或片上系统芯片等。
本申请实施例另提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现上述AI模型的处理方法实施例的各个过程，且能达到相同的技术效果， 为避免重复，这里不再赘述。
本申请实施例还提供了一种AI模型的处理系统，包括：终端及网络侧设备，所述终端可用于执行如上所述的AI模型的处理方法的步骤，所述网络侧设备可用于执行如上所述的AI模型的处理方法的步骤。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外，需要指出的是，本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能，还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能，例如，可以按不同于所描述的次序来执行所描述的方法，并且还可以添加、省去、或组合各种步骤。另外，参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以计算机软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述，但是本申请并不局限于上述的具体实施方式，上述的具体实施方式仅仅是示意性的，而不是限制性的，本领域的普通技术人员在本申请的启示下，在不脱离本申请宗旨和权利要求所保护的范围情况下，还可做出很多形式，均属于本申请的保护之内。

Claims (39)

  1. 一种人工智能AI模型的处理方法，包括：
    第一通信设备确定第一区域；
    所述第一通信设备执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
  2. 根据权利要求1所述的方法，其中，所述第一通信设备包括移动通信设备，所述第一通信设备确定第一区域包括：
    所述移动通信设备确定所述移动通信设备的位置信息；
    所述移动通信设备根据所述位置信息确定所述第一区域。
  3. 根据权利要求1所述的方法，其中，所述第一通信设备包括固定通信设备，所述第一通信设备确定第一区域包括：
    所述固定通信设备确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；
    所述固定通信设备根据所述位置信息确定所述第一区域，所述移动通信设备处于所述第一区域内。
  4. 根据权利要求1所述的方法，其中，所述第一区域与地理坐标对应；
    其中，所述第一区域的大小为W×L，L为所述第一区域的长度，W为所述第一区域的宽度，W、L均为正数。
  5. 根据权利要求1所述的方法，其中，所述第一区域的特征包括：
    第二通信设备与处于所述第一区域内的一个或多个所述第一通信设备之间的信道数据；
    其中，所述第一通信设备包括移动通信设备，所述第二通信设备包括固定通信设备。
  6. 根据权利要求1所述的方法，其中，进行AI模型训练的范围包括：与固定通信设备的最大通信范围相关的区域。
  7. 根据权利要求1所述的方法，其中，所述第一通信设备包括移动通信设备，所述AI模型包括下行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：
    所述移动通信设备在所述第一区域内接收下行参考信号，并根据所述下行参考信号得到具有所述第一区域的特征的数据；
    所述移动通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。
  8. 根据权利要求1所述的方法，其中，所述第一通信设备包括固定通信设备，所述AI模型包括上行链路AI模型，所述第一通信设备使用具有所述 第一区域的特征的数据进行AI模型训练包括：
    所述固定通信设备接收上行参考信号，并根据所述上行参考信号得到具有所述第一区域的特征的数据，所述上行参考信号是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；
    所述固定通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。
  9. 根据权利要求1所述的方法，其中，所述第一通信设备包括固定通信设备，所述AI模型包括下行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：
    所述固定通信设备接收具有所述第一区域的特征的数据，所述具有所述第一区域的特征的数据是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；
    所述固定通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。
  10. 根据权利要求1所述的方法，其中，所述第一通信设备包括固定通信设备，所述AI模型包括下行链路AI模型，所述第一通信设备使用具有所述第一区域的特征的数据进行AI模型训练包括：
    所述固定通信设备接收第一数据，所述第一数据是第二通信设备发送的，所述第二通信设备包括移动通信设备；
    所述固定通信设备根据所述移动通信设备的位置信息确定所述第一数据具有所述第一区域的特征；
    所述固定通信设备使用所述具有所述第一区域的特征的数据进行AI模型训练。
  11. 根据权利要求1所述的方法，其中，所述第一通信设备包括移动通信设备，所述第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：
    所述移动通信设备确定所述移动通信设备的位置信息；
    所述移动通信设备根据所述位置信息确定所述第一区域；
    所述移动通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择与所述索引k对应的AI模型进行信道预测，K是正整数。
  12. 根据权利要求11所述的方法，其中，所述AI模型包括上行链路AI模型或下行链路AI模型。
  13. 根据权利要求11所述的方法，其中，所述AI模型包括上行链路AI模型，所述信道预测包括预测时刻t+Δ的信道，所述方法还包括：
    所述移动通信设备接收时刻t的信道，所述时刻t的信道是第二通信设备 发送的，所述第二通信设备包括固定通信设备。
  14. 根据权利要求1所述的方法，其中，所述第一通信设备包括固定通信设备，所述第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：
    所述固定通信设备确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；
    所述固定通信设备根据所述位置信息确定所述第一区域；
    所述固定通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI模型进行信道预测，K是正整数。
  15. 根据权利要求14所述的方法，其中，所述AI模型包括上行链路AI模型或下行链路AI模型。
  16. 根据权利要求15所述的方法，其中，所述AI模型包括上行链路AI模型，所述信道预测包括预测时刻t+Δ的信道，所述方法还包括：
    所述固定通信设备接收时刻t的信道，所述时刻t的信道是所述移动通信设备发送的。
  17. 根据权利要求1所述的方法，其中，所述第一通信设备包括移动通信设备，第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：
    所述移动通信设备确定所述移动通信设备的位置信息；
    所述移动通信设备根据所述位置信息确定所述第一区域；
    所述移动通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；
    所述移动通信设备向第二通信设备发送所述压缩信道数据，所述第二通信设备包括固定通信设备。
  18. 根据权利要求17所述的方法，其中，所述方法还包括：
    所述移动通信设备向所述固定通信设备发送所述AI压缩模型对应的AI解码模型；
    其中，所述AI解码模型用于解压所述压缩信道数据。
  19. 根据权利要求17所述的方法，其中，所述移动通信设备确定所述移动通信设备的位置信息包括：
    所述移动通信设备自主确定所述移动通信设备的位置信息；或者，
    所述移动通信设备接收所述移动通信设备的位置信息，所述位置信息所述固定通信设备确定的。
  20. 根据权利要求1所述的方法，其中，所述第一通信设备包括固定通 信设备，第一通信设备确定第一区域；所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：
    所述固定通信设备确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；
    所述固定通信设备根据所述位置信息确定所述第一区域；
    所述固定通信设备根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；
    所述固定通信设备向所述移动通信设备发送所述压缩信道数据。
  21. 根据权利要求20所述的方法，其中，所述方法还包括：
    所述固定通信设备向所述移动通信设备发送所述AI压缩模型对应的AI解码模型；
    其中，所述AI解码模型用于解压所述压缩信道数据。
  22. 根据权利要求20所述的方法，其中，所述固定通信设备确定第二通信设备的位置信息包括：
    所述固定通信设备自主确定所述第二通信设备的位置信息；或者，
    所述固定通信设备接收所述第二通信设备的位置信息，所述位置信息所述第二通信设备确定的。
  23. 根据权利要求1所述的方法，其中，所述AI模型包括AI预测模型以及AI压缩模型，所述第一通信设备选择与所述第一区域匹配的AI模型执行目标通信业务包括：
    所述第一通信设备选择所述AI预测模型预测时刻t+Δ的信道；
    所述第一通信设备选择所述AI压缩模型对所述时刻t+Δ的信道进行压缩，得到压缩信道数据；
    所述第一通信设备向第二通信设备发送所述压缩信道数据。
  24. 根据权利要求23所述的方法，其中，所述方法还包括：
    所述第一通信设备向所述第二通信设备发送所述AI压缩模型对应的AI解码模型；
    其中，所述AI解码模型用于解压所述压缩信道数据。
  25. 根据权利要求23所述的方法，其中，
    所述第一通信设备包括固定通信设备，所述第二通信设备包括移动通信设备；或者，
    所述第一通信设备包括移动通信设备，所述第二通信设备包括固定通信设备。
  26. 一种第一通信设备，包括：
    确定模块，用于确定第一区域；
    执行模块，用于执行如下至少之一：使用具有所述第一区域的特征的数据进行AI模型训练；选择与所述第一区域匹配的AI模型执行目标通信业务，其中，所述AI模型具有所述第一区域的特征。
  27. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括移动通信设备，所述确定模块，用于：
    确定所述移动通信设备的位置信息；
    根据所述位置信息确定所述第一区域。
  28. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括固定通信设备，所述确定模块，用于：
    确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；
    根据所述位置信息确定所述第一区域，所述移动通信设备处于所述第一区域内。
  29. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括移动通信设备，所述AI模型包括下行链路AI模型，所述执行模块，用于：
    在所述第一区域内接收下行参考信号，并根据所述下行参考信号得到具有所述第一区域的特征的数据；
    使用所述具有所述第一区域的特征的数据进行AI模型训练。
  30. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括固定通信设备，所述AI模型包括上行链路AI模型，所述执行模块，用于：
    接收上行参考信号，并根据所述上行参考信号得到具有所述第一区域的特征的数据，所述上行参考信号是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；
    使用所述具有所述第一区域的特征的数据进行AI模型训练。
  31. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括固定通信设备，所述AI模型包括下行链路AI模型，所述执行模块，用于：
    接收具有所述第一区域的特征的数据，所述具有所述第一区域的特征的数据是处于所述第一区域内的第二通信设备发送的，所述第二通信设备包括移动通信设备；
    使用所述具有所述第一区域的特征的数据进行AI模型训练。
  32. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括固定通信设备，所述AI模型包括下行链路AI模型，所述执行模块，用于：
    接收第一数据，所述第一数据是第二通信设备发送的，所述第二通信设备包括移动通信设备；
    根据所述移动通信设备的位置信息确定所述第一数据具有所述第一区域 的特征；
    使用所述具有所述第一区域的特征的数据进行AI模型训练。
  33. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括移动通信设备；
    所述确定模块，用于确定所述移动通信设备的位置信息；根据所述位置信息确定所述第一区域；
    所述执行模块，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择与所述索引k对应的AI模型进行信道预测，K是正整数。
  34. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括固定通信设备；
    所述确定模块，用于确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；根据所述位置信息确定所述第一区域；
    所述执行模块，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI模型进行信道预测，K是正整数。
  35. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括移动通信设备；
    所述确定模块，用于确定所述移动通信设备的位置信息；根据所述位置信息确定所述第一区域；
    所述执行模块，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；向第二通信设备发送所述压缩信道数据，所述第二通信设备包括固定通信设备。
  36. 根据权利要求26所述的第一通信设备，其中，所述第一通信设备包括固定通信设备；
    所述确定模块，用于确定第二通信设备的位置信息，所述第二通信设备包括移动通信设备；根据所述位置信息确定所述第一区域；
    所述执行模块，用于根据所述第一区域的索引k，在训练完成的K个AI模型中选择所述索引k对应的AI压缩模型对信道进行压缩，得到压缩信道数据，K是正整数；向所述移动通信设备发送所述压缩信道数据。
  37. 根据权利要求26所述的第一通信设备，其中，所述AI模型包括AI预测模型以及AI压缩模型，所述执行模块，用于：
    选择所述AI预测模型预测时刻t+Δ的信道；
    选择所述AI压缩模型对所述时刻t+Δ的信道进行压缩，得到压缩信道数据；
    向第二通信设备发送所述压缩信道数据。
  38. 一种通信设备，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求1至25任一项所述的方法的步骤。
  39. 一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如权利要求1至25任一项所述的方法的步骤。
PCT/CN2023/081303 2022-03-18 2023-03-14 Ai模型的处理方法及设备 WO2023174253A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210269877.5 2022-03-18
CN202210269877.5A CN116806029A (zh) 2022-03-18 2022-03-18 Ai模型的处理方法及设备

Publications (1)

Publication Number Publication Date
WO2023174253A1 true WO2023174253A1 (zh) 2023-09-21

Family

ID=88022265

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/081303 WO2023174253A1 (zh) 2022-03-18 2023-03-14 Ai模型的处理方法及设备

Country Status (2)

Country Link
CN (1) CN116806029A (zh)
WO (1) WO2023174253A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110121205A (zh) * 2019-04-30 2019-08-13 维沃移动通信有限公司 一种终端设备的控制方法及终端设备
US20210185700A1 (en) * 2019-12-13 2021-06-17 Qualcomm Incorporated Scheduling request associated with artificial intelligence information
WO2021193989A1 (ko) * 2020-03-25 2021-09-30 엘지전자 주식회사 무선 통신 시스템에서 무선 신호를 송수신하는 방법 및 장치
CN114071484A (zh) * 2020-07-30 2022-02-18 华为技术有限公司 基于人工智能的通信方法和通信装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110121205A (zh) * 2019-04-30 2019-08-13 维沃移动通信有限公司 一种终端设备的控制方法及终端设备
US20210185700A1 (en) * 2019-12-13 2021-06-17 Qualcomm Incorporated Scheduling request associated with artificial intelligence information
WO2021193989A1 (ko) * 2020-03-25 2021-09-30 엘지전자 주식회사 무선 통신 시스템에서 무선 신호를 송수신하는 방법 및 장치
CN114071484A (zh) * 2020-07-30 2022-02-18 华为技术有限公司 基于人工智能的通信方法和通信装置

Also Published As

Publication number Publication date
CN116806029A (zh) 2023-09-26

Similar Documents

Publication Publication Date Title
CN114363921B (zh) Ai网络参数的配置方法和设备
US20240154773A1 (en) Tci state determining method and apparatus, terminal, and network-side device
US20240088970A1 (en) Method and apparatus for feeding back channel information of delay-doppler domain, and electronic device
WO2023066288A1 (zh) 模型请求方法、模型请求处理方法及相关设备
WO2023174253A1 (zh) Ai模型的处理方法及设备
WO2023174325A1 (zh) Ai模型的处理方法及设备
CN116415476A (zh) 模型的构建方法、装置及通信设备
WO2024061287A1 (zh) 人工智能ai模型传输方法、装置、终端及介质
WO2023179651A1 (zh) 波束处理方法、装置及设备
WO2023160547A1 (zh) 信息响应方法、信息发送方法、终端及网络侧设备
WO2023179540A1 (zh) 信道预测方法、装置及无线通信设备
WO2023103912A1 (zh) 分集传输方法、终端及网络侧设备
WO2023109828A1 (zh) 数据收集方法及装置、第一设备、第二设备
WO2024001981A1 (zh) 预编码矩阵的指示方法、终端及网络侧设备
WO2024027576A1 (zh) 一种ai网络模型的性能监督方法、装置和通信设备
WO2024067281A1 (zh) Ai模型的处理方法、装置及通信设备
WO2023198094A1 (zh) 模型输入的确定方法及通信设备
WO2023088387A1 (zh) 信道预测方法、装置、ue及系统
WO2024032694A1 (zh) Csi预测处理方法、装置、通信设备及可读存储介质
WO2024012303A1 (zh) 一种ai网络模型交互方法、装置和通信设备
WO2023179649A1 (zh) 人工智能模型的输入处理方法、装置及设备
WO2024093997A1 (zh) 确定模型适用性的方法、装置及通信设备
WO2023151650A1 (zh) 信息激活方法、终端及网络侧设备
CN117997457A (zh) 参考信号确定方法、终端及网络侧设备
CN116963093A (zh) 模型调整方法、信息传输方法、装置及相关设备

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: 23769755

Country of ref document: EP

Kind code of ref document: A1