Communication method, communication device, communication equipment and storage medium

Abstract

Provided are a communication method, a communication device, a communication device and a storage medium, the method comprising: a network device sending first CSI generation model information to a terminal device, deploying a first CSI reconstruction model in a decoder, the first CSI reconstruction model corresponding to a first CSI generation model corresponding to the first CSI generation model information; and the terminal equipment deploys a first CSI generation model at the encoder according to the first CSI generation model information. By adopting the method, the deployment of the double-side AI model can be realized by sending the CSI generation model information to the terminal equipment through the network equipment.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a communication method, a communication apparatus, a communication device, and a storage medium. Background Art

[0002] The 3rd Generation Partnership Project (3GPP) has defined three initial use cases in its Artificial Intelligence (AI) / Machine Learning (ML) Study Item (SI). These include enhanced Channel State Information (CSI) feedback. While enhanced CSI feedback includes the deployment of two-sided AI models, 3GPP has not yet defined how to implement this. Summary of the Invention

[0003] Based on this, the present application provides a communication method, a communication device, a communication equipment and a storage medium, which can realize the deployment of bilateral AI models by sending CSI generation model information to terminal devices through network devices.

[0004] In a first aspect, the present application provides a communication method, applied to a terminal device, wherein the terminal device includes an encoder, and the method includes:

[0005] Receiving first channel state information CSI generation model information from a network device;

[0006] A first CSI generation model is deployed at the encoder according to the first CSI generation model information.

[0007] In one of the embodiments, the first CSI generation model information is carried in a radio resource control RRC message, a downlink control information DCI or a media access control MAC control element CE.

[0008] In one embodiment, deploying the first CSI generation model at the encoder according to the first CSI generation model information includes:

[0009] In a case where the terminal device supports a version of the CSI generation model corresponding to the first CSI generation model information, the first CSI generation model is deployed on the encoder according to the first CSI generation model information.

[0010] In one embodiment, the method further comprises:

[0011] generating first CSI using the first CSI generation model;

[0012] sending the first CSI to the network device;

[0013] receiving second CSI from the network device, where the second CSI is reconstructed by using a first CSI reconstruction model corresponding to the first CSI generation model;

[0014] determining a similarity between the second CSI and third CSI, where the third CSI is original CSI corresponding to the first CSI;

[0015] When the similarity is less than a threshold, sending a CSI feedback method switching request to the network device;

[0016] receiving information of a target CSI feedback mode from the terminal device, where the target CSI feedback mode corresponds to a traditional CSI feedback mode, or a second CSI generation model and a second CSI reconstruction model, where the first CSI generation model is different from the second CSI generation model, and the first CSI reconstruction model is different from the second CSI reconstruction model;

[0017] According to the information of the target CSI feedback mode, the CSI feedback mode is switched to the target CSI feedback mode.

[0018] In one embodiment, the method further comprises:

[0019] generating first CSI using the first CSI generation model;

[0020] sending the first CSI to the network device;

[0021] receiving first indication information from the network device, where the first indication information is used to instruct a CSI feedback mode switch, the first indication information including information of a target CSI feedback mode, where the target CSI feedback mode corresponds to a traditional CSI feedback mode, or a second CSI generation model and a second CSI reconstruction model, where the first CSI generation model is different from the second CSI generation model, and the first CSI reconstruction model is different from the second CSI reconstruction model;

[0022] According to the first indication information, the CSI feedback mode is switched to the target CSI feedback mode.

[0023] In one embodiment, the method further comprises:

[0024] receiving first generation parameter information from a server;

[0025] updating initial parameters of an initial CSI generation model deployed in the encoder according to the first generation parameter information to obtain a third CSI generation model;

[0026] training the third CSI generation model to obtain a fourth CSI generation model;

[0027] determining first gradient information according to parameters of the fourth CSI generation model;

[0028] sending the first gradient information to the server;

[0029] receiving second generation parameter information from the server, where the second generation parameter information is determined according to the first gradient information;

[0030] updating parameters of the fourth CSI generation model according to the second generation parameter information to obtain a fifth CSI generation model;

[0031] When the fifth CSI generation model converges, the first CSI generation model is determined to be the fifth CSI generation model.

[0032] In one embodiment, the receiving of first generation parameter information from the server includes:

[0033] The first generation parameter information is received from the server in a broadcast manner.

[0034] In one embodiment, the method further comprises:

[0035] receiving second indication information from the network device, where the second indication information is used to indicate a target quantization dictionary;

[0036] determining the target quantization dictionary according to the second indication information;

[0037] The training of the third CSI generation model to obtain a fourth CSI generation model includes:

[0038] The third CSI generation model is trained according to the target quantization dictionary to obtain a fourth CSI generation model.

[0039] In one embodiment, the method further comprises:

[0040] determining a target codebook according to the target quantization dictionary;

[0041] The training of the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model includes:

[0042] The third CSI generation model is trained according to the target quantization dictionary and the target codebook to obtain a fourth CSI generation model.

[0043] In one embodiment, the training of the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model includes:

[0044] When there is no conflict in the target quantization dictionary, the third CSI generation model is trained according to the target quantization dictionary to obtain a fourth CSI generation model.

[0045] In one embodiment, the second indication information includes an identifier of a target quantization dictionary, and the method further includes:

[0046] Receive information from the dictionary library;

[0047] The determining the target quantization dictionary according to the second indication information includes:

[0048] The target quantization dictionary is acquired from the dictionary library according to the information of the dictionary library and the identifier of the target quantization dictionary.

[0049] In one embodiment, the CSI generation model information includes one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

[0050] In a second aspect, the present application further provides a communication method, applied to a network device, wherein the network device includes a decoder, and the method includes:

[0051] Sending first channel state information CSI generation model information to the terminal device;

[0052] A first CSI reconstruction model is deployed in the decoder, where the first CSI reconstruction model corresponds to the first CSI generation model corresponding to the first CSI generation model information.

[0053] In one of the embodiments, the first CSI generation model information is carried in a radio resource control RRC message, a downlink control information DCI or a media access control MAC control element CE.

[0054] In one embodiment, the method further comprises:

[0055] receiving first CSI from the terminal device;

[0056] reconstructing the first CSI using the first CSI reconstruction model to obtain second CSI;

[0057] Sending the second CSI to the terminal device;

[0058] Receiving a CSI feedback method switching request from the terminal device;

[0059] Sending information of a target CSI feedback mode to the terminal device, where the target CSI feedback mode corresponds to a traditional CSI feedback mode or a second CSI generation model, and the first CSI generation model is different from the second CSI generation model;

[0060] The reconstruction mode is switched to a target CSI reconstruction mode corresponding to the target CSI feedback mode, where the target CSI reconstruction mode corresponds to a traditional CSI reconstruction mode or a second CSI reconstruction model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

[0061] In one embodiment, the method further comprises:

[0062] receiving first CSI from the terminal device;

[0063] reconstructing the first CSI using the first CSI reconstruction model to obtain second CSI;

[0064] Determine a similarity between the second CSI and third CSI, where the third CSI is original CSI corresponding to the first CSI generated by the terminal device;

[0065] If the similarity is less than a threshold, sending first indication information to the terminal device, where the first indication information is used to instruct a CSI feedback mode to be switched, and the first indication information includes information of a target CSI feedback mode, where the target CSI feedback mode corresponds to a traditional CSI feedback mode or a second CSI generation model, and the first CSI generation model is different from the second CSI generation model;

[0066] The reconstruction mode is switched to a target CSI reconstruction mode corresponding to the target CSI feedback mode, where the target CSI reconstruction mode corresponds to a traditional CSI reconstruction mode or a second CSI reconstruction model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

[0067] In one embodiment, the method further comprises:

[0068] receiving first reconstruction parameter information from a server;

[0069] updating initial parameters of an initial CSI reconstruction model deployed in the decoder according to the first reconstruction parameter information to obtain a third CSI reconstruction model;

[0070] Training the third CSI reconstruction model to obtain a fourth CSI reconstruction model;

[0071] determining second gradient information according to parameters of the fourth CSI reconstruction model;

[0072] sending the second gradient information to the server;

[0073] receiving second reconstruction parameter information from the server, where the second reconstruction parameter information is determined according to the second gradient information;

[0074] updating parameters of the fourth CSI reconstruction model according to the second reconstruction parameter information to obtain a fifth CSI reconstruction model;

[0075] When the fifth CSI reconstruction model converges, the first CSI reconstruction model is determined to be the fifth CSI reconstruction model.

[0076] In one embodiment, the receiving first reconstruction parameter information from the server includes:

[0077] The first reconstruction parameter information is received from the server in a broadcast manner.

[0078] In one embodiment, the method further comprises:

[0079] Sending second indication information to the terminal device, where the second indication information is used to indicate a target quantization dictionary;

[0080] The training of the third CSI reconstruction model to obtain a fourth CSI reconstruction model includes:

[0081] The third CSI reconstruction model is trained according to the target quantization dictionary to obtain a fourth CSI reconstruction model.

[0082] In one embodiment, the method further comprises:

[0083] determining a target codebook according to the target quantization dictionary;

[0084] The training of the third CSI reconstruction model according to the target quantization dictionary to obtain a fourth CSI reconstruction model includes:

[0085] The third CSI reconstruction model is trained according to the target quantization dictionary and the target codebook to obtain a fourth CSI reconstruction model.

[0086] In one embodiment, the second indication information includes an identifier of a target quantization dictionary, and the method further includes:

[0087] The dictionary library information is sent to the terminal device.

[0088] In one embodiment, the CSI generation model information includes one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

[0089] In a third aspect, the present application further provides a communication device, applied to a terminal device, wherein the terminal device includes an encoder, and the device includes:

[0090] A receiving module, configured to receive first channel state information CSI generation model information from a network device;

[0091] A deployment module is used to deploy a first CSI generation model on the encoder according to the first CSI generation model information.

[0092] In a fourth aspect, the present application further provides a communication device, applied to a network device, wherein the network device includes a decoder, and the device includes:

[0093] A sending module, configured to send first channel state information CSI generation model information to a terminal device;

[0094] A deployment module is used to deploy a first CSI reconstruction model in the decoder, where the first CSI reconstruction model corresponds to the first CSI generation model corresponding to the first CSI generation model information.

[0095] In the fifth aspect, the present application also provides a communication device, which can be a terminal device. The communication device includes a memory and a processor, and the memory stores a computer program. When the processor executes the computer program, it implements the steps of the method provided in the first aspect or any embodiment of the first aspect.

[0096] In the sixth aspect, the present application also provides a communication device, which may be a network device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method provided in the second aspect or any embodiment of the second aspect when executing the computer program.

[0097] In the seventh aspect, the present application also provides a communication system, which includes the communication equipment provided in the fifth and sixth aspects.

[0098] In an eighth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above methods are implemented.

[0099] In a ninth aspect, the present application also provides a computer program product, comprising a computer program, which implements the steps of the above methods when executed by a processor.

[0100] In this application, the network device sends first CSI generation model information to the terminal device and deploys the first CSI reconstruction model in the decoder; the terminal device deploys the first CSI generation model in the encoder based on the first CSI generation model information, and the first CSI generation model corresponds to the first CSI reconstruction model. It can be seen that when CSI needs to be transmitted between the terminal device and the network device, the network device can send CSI generation model information to the terminal device so that the terminal device deploys the CSI generation model in the encoder based on the CSI generation model information. In addition, the network device deploys the CSI reconstruction model in the decoder to achieve the deployment of a bilateral AI model with CSI feedback enhancement. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0102] Figure 1 This is a schematic diagram of a network architecture provided by an embodiment of the present application;

[0103] Figure 2 This is a flow chart of a communication method provided in an embodiment of the present application;

[0104] Figure 3 This is a flow chart of another communication method provided in an embodiment of the present application;

[0105] Figure 4 This is a flow chart of another communication method provided in an embodiment of the present application;

[0106] Figure 5 This is a flow chart of another communication method provided in an embodiment of the present application;

[0107] Figure 6 This is a flow chart of another communication method provided in an embodiment of the present application;

[0108] Figure 7 A schematic structural diagram of a communication device provided in an embodiment of the present application;

[0109] Figure 8 A schematic structural diagram of another communication device provided in an embodiment of the present application;

[0110] Figure 9 A schematic diagram of the structure of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0111] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0112] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, such that the embodiments of the present disclosure described herein can be implemented in an order other than that illustrated or described herein.

[0113] The present application provides a communication method, a communication apparatus, a communication device, and a storage medium, which can realize the deployment of a bilateral AI model by sending CSI generation model information to a terminal device through a network device.

[0114] In order to better understand the embodiments of the present application, the network architecture of the present application is first described below.

[0115] Figure 1 This is a network architecture diagram provided by an embodiment of the present application. Figure 1 As shown, the network architecture may include a server 101, a network device 102, and a terminal device 103. The server 101 and the network device 102, the server 101 and the terminal device 103, and the network device 102 and the terminal device 103 may all be connected via a network.

[0116] Terminal device 103 includes an encoder, which may be deployed with or loaded with an initial CSI generation model. Network device 102 includes a decoder, which may be deployed with or loaded with an initial CSI reconstruction model. The initial CSI generation model corresponds to the initial CSI reconstruction model, i.e., the initial CSI generation model is the CSI generation model corresponding to the initial CSI reconstruction model.

[0117] Server 101 can broadcast initial generation parameters of the initial CSI generation model and initial reconstruction parameters of the initial CSI reconstruction model. After receiving the initial generation parameters, terminal device 103 can update the parameters of the initial CSI generation model based on the initial generation parameters and then train the initial CSI generation model based on the generated training data. After receiving the initial reconstruction parameters, network device 102 can update the parameters of the initial CSI reconstruction model based on the initial reconstruction parameters and then train the initial CSI reconstruction model based on the reconstructed training data.

[0118] During a training process of the initial CSI generation model and the initial CSI reconstruction model, the terminal device 103 may use the initial CSI generation model to generate compressed CSI based on one of the training data, and then may send the compressed CSI to the network device 102. The terminal device may also update the parameters of the initial CSI reconstruction model based on the generated compressed CSI and the label data in the training data.

[0119] After receiving the compressed CSI, the network device 102 may use the initial CSI reconstruction model to decompress the compressed CSI to obtain reconstructed CSI, and may update the parameters of the initial reconstruction model according to the reconstructed CSI and the reconstructed training data.

[0120] After the terminal device 103 and the network device 102 complete the above-mentioned training once or multiple trainings, the terminal device 103 may generate first generation gradient information based on the parameters of the initial CSI generation model and may send the first generation gradient information to the server 101. The network device 102 may generate first reconstruction gradient information based on the parameters of the initial CSI reconstruction model and may send the first reconstruction gradient information to the server 101. After receiving the first generation gradient information and the first reconstruction gradient information, the server 101 may determine target generation parameter information based on the first generation gradient information and the first reconstruction gradient information, and may then send the target reconstruction parameter information to the network device 102 and the target generation parameter information to the terminal device 103. After receiving the target reconstruction parameter information, the network device 102 may update the parameters of the initial CSI reconstruction model based on the target reconstruction parameter information. After receiving the target generation parameter information, the terminal device 103 may update the parameters of the initial CSI generation model based on the target generation parameter information.

[0121] It should be understood that the above is one round of training of the CSI model. In one case, after a round of training is completed, the network device 102 can determine whether the trained initial CSI reconstruction model has converged, and the terminal device 103 can determine whether the trained CSI generation model has converged. In another case, after a round of training is completed, the server 101 can determine whether the trained initial CSI reconstruction model and the trained initial CSI generation model have converged. If the trained initial CSI generation model has not converged, or the trained initial CSI reconstruction model has not converged, the next round of training can be continued until the trained initial CSI reconstruction model and the trained initial CSI generation model have converged, and then the target CSI generation model and the target CSI reconstruction model are obtained. The target CSI generation model and the target CSI reconstruction model can be stored in the network device 102, or in the server 101, or in other servers or devices.

[0122] If the target CSI generation model and target CSI reconstruction model are to be used later, the network device 102 may send information about the target CSI generation model to the terminal device 103, and may deploy the target CSI reconstruction model in the decoder. After receiving the target CSI generation model information, the terminal device may deploy the target CSI generation model in the encoder based on the target CSI generation model information. The target CSI generation model may then be used to generate CSI and sent to the network device 102. After receiving the CSI reported by the terminal device 103, the network device 102 may reconstruct the CSI using the target CSI reconstruction model.

[0123] Terminal equipment, also known as User Equipment (UE), Subscriber Station (STA), Mobile Station (MS), Mobile Terminal (MT), etc., is a device that provides voice and / or data connectivity to users. The terminal device can be a handheld terminal, a very small aperture terminal (VSAT), a laptop computer, a customer premises equipment (CPE) laptop computer, a subscriber unit, a cellular phone, a smart phone, a wireless data card, a personal digital assistant (PDA) computer, a tablet computer, a handheld device (handheld), a laptop computer, a cordless phone or a wireless local loop (WLL) station, a machine type communication (MTC) terminal, a wearable device (such as a smart watch, a smart bracelet, a pedometer, etc.), an in-vehicle device (such as a car, a bicycle, an electric car, an airplane, a ship, a train, a high-speed rail, etc.), a virtual reality (VR) device, an augmented reality (AR) device, or a wearable device (such as a smart watch, a smart bracelet, a pedometer, etc.). The present invention relates to wireless terminals for autonomous driving, wireless terminals for telemedicine, wireless terminals for smart grid, wireless terminals for transportation safety, wireless terminals for smart city, wireless terminals for smart home, flight equipment (such as intelligent robots, hot air balloons, drones, airplanes, etc.), or other devices that can access the Internet. This application does not limit the specific technology and specific device form adopted by the terminal device.

[0124] Network equipment is a device that provides wireless access to terminal devices and is primarily responsible for functions such as wireless resource management, Quality of Service (QoS) flow management, data compression, and encryption on the air interface side. Network equipment can be a Radio Access Network (RAN) device or a centralized unit (CU) within a RAN device. RAN equipment can include various forms of base stations, such as macro base stations, micro base stations (also known as small stations), relay stations, and access points. RAN equipment can also include Wireless Fidelity (WiFi) access points (APs). RAN equipment can also include Worldwide interoperability for Microwave access (WiMax) base stations (BSs).

[0125] The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CSI model management, CSI model training, as well as big data and artificial intelligence platforms.

[0126] It should be noted that Figure 1 The network architecture shown is not limited to including only the devices shown in the figure, but may also include other devices not shown in the figure, which will not be listed one by one in the present invention.

[0127] It should be noted that Figure 1 The network architecture shown can be the network architecture of the fifth generation mobile communication technology (5th Generation Mobile Networks, 5G), the network architecture of the sixth generation mobile communication technology (6th Generation Mobile Networks, 6G), or the network architecture of other communication systems.

[0128] Based on the above network architecture, Figure 2 This is a flow chart of a communication method provided by an embodiment of the present application. Figure 2 As shown, the communication method may include the following steps.

[0129] 201. The network device sends first CSI generation model information to the terminal device.

[0130] Correspondingly, the terminal device receives the first CSI generation model information from the network device.

[0131] The first CSI generation model information is information about the first CSI generation model. The CSI generation model is a model used to generate or predict CSI and compress the generated or predicted CSI.

[0132] The first CSI generation model is a pre-trained CSI generation model. The CSI generation model can be stored in a network device, a dedicated server, or other devices. Figure 1 The servers in the can be the same server or different servers.

[0133] When the CSI generation model is stored in a network device, the first CSI generation model information may include data of the first CSI generation model. When the CSI generation model is stored in a server or other device, the first CSI generation model information may include an identifier of the first CSI generation model.

[0134] When CSI needs to be transmitted between the terminal device and the network device, the network device may send first CSI generation model information to the terminal device.

[0135] The CSI needs to be transmitted between the terminal device and the network device. This can be understood as the terminal device needing to report the CSI to the network device, or the network device needing the terminal device to report the CSI.

[0136] The first CSI generation model information may be carried in a Radio Resource Control (RRC) message. The network device may send an RRC message to the terminal device. The terminal device may receive an RRC message from the network device. The RRC message may include the first CSI generation model information. The RRC message may be an RRC Setup message, an RRC Reconfiguration message, or other RRC message.

[0137] The first CSI generation model information may be carried in downlink control information (DCI). The network device may send the DCI to the terminal device. The terminal device may receive the DCI from the network device. The DCI may include the first CSI generation model information.

[0138] The first CSI generation model information may be carried in a Media Access Control (MAC) Control Element (CE). The network device may send a MAC CE to the terminal device. The terminal device may receive the MAC CE from the network device. The MAC CE may include the first CSI generation model information.

[0139] AI / ML model parameters can be defined according to the hierarchical structure of Open Neural Network Exchange (ONNX). AI / ML model parameters may include a metadata layer, a core data layer, and an extended data layer. The metadata layer may include basic attribute parameters of the AI / ML model, such as the AI / ML model's identifier, version information, and cyclic redundancy check (CRC) checksum. The core data layer may include key parameters of the AI / ML model, such as the weights of the convolutional layer and the fully connected layer of the AI / ML model. The extended data layer includes customized parameters.

[0140] For example, AI / ML model parameters can be defined as follows:

[0141] message ModelPackage{

[0142] string model_id = 1; / / Model unique identifier (such as 3GPP_CSI_Encoder_v2)

[0143] repeated float base_weights = 2; / / core layer parameters (convolution / fully connected layer)

[0144] repeated float extension_weights=3; / / Manufacturer-defined extension layer parameters

[0145] QuantizationConfig quant_config = 6; / / Quantization configuration

[0146] }

[0147] The CSI generation model information may include one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

[0148] The metadata layer information of the CSI generation model may include one or more of an identifier of the CSI generation model, version information, and a CRC check code.

[0149] 202. The terminal device deploys the first CSI generation model in the encoder according to the first CSI generation model information.

[0150] The terminal device includes an encoder. After receiving the first CSI generation model information from the network device, the terminal device can deploy or load the first CSI generation model in the encoder according to the first CSI generation model information.

[0151] In the case where the first CSI generation model information includes data of the first CSI generation model, the terminal device can directly deploy the first CSI generation model in the encoder according to the first CSI generation model information.

[0152] When the first CSI generation model information includes an identifier of the first CSI generation model, the terminal device may first obtain data of the first CSI generation model according to the identifier of the first CSI generation model, and then deploy the first CSI generation model in the encoder according to the data of the first CSI generation model.

[0153] The terminal device may first determine whether the first CSI generation model information is successfully received. If it is determined that the first CSI generation model information is successfully received, the terminal device may send an Acknowledgment (ACK) message to the network device, or may not send an ACK message to the network device. If the network device receives an ACK message, or does not receive a Negative Acknowledgment (NACK) message within a preset time, it may be determined that the terminal device has successfully received the first CSI generation model information.

[0154] If the terminal device determines that the first CSI generation model information has not been successfully received, the terminal device may send a NACK message to the network device. After receiving the NACK message, the network device may retransmit the first CSI generation model information to the terminal device until the terminal device successfully receives the first CSI generation model information.

[0155] The terminal device can verify the first CSI generation model information based on the CRC check code included in the metadata layer information of the first CSI generation model (i.e., the information of the metadata layer). If the verification passes, it can be determined that the first CSI generation model information is received successfully. If the verification fails, it can be determined that the first CSI generation model information is received unsuccessfully.

[0156] 203. The network device deploys a first CSI reconstruction model in a decoder.

[0157] The network device includes a decoder. After the network device sends the first CSI generation model information to the terminal device, the first CSI reconstruction model may be deployed or loaded in the decoder. The first CSI reconstruction model corresponds to the first CSI generation model corresponding to the first CSI generation model information, i.e., the first CSI reconstruction model is the CSI reconstruction model corresponding to the first CSI generation model corresponding to the first CSI generation model information. The CSI reconstruction model is a model used to reconstruct CSI generated using the CSI generation model.

[0158] In the case where the network device stores a CSI reconstruction model, the network device may first determine the first CSI reconstruction model corresponding to the first CSI generation model, and then deploy the first CSI reconstruction model in the decoder according to data of the first CSI reconstruction model.

[0159] When the network device does not store a CSI reconstruction model, the network device can first determine the first CSI reconstruction model corresponding to the first CSI generation model, obtain data of the first CSI reconstruction model, and then deploy the first CSI reconstruction model in the decoder based on the data of the first CSI reconstruction model.

[0160] In some embodiments, after the encoder deploys the first CSI generation model, the terminal device may send information indicating the successful deployment of the first CSI generation model to the network device. This information may be carried in an RRC message, such as an RRCReconfigurationComplete message, or in other messages. After determining that the terminal device has successfully deployed the first CSI generation model, i.e., after receiving this information, the network device may deploy the first CSI reconstruction model in the decoder, thereby avoiding unnecessary deployment and reducing power consumption of the network device.

[0161] exist Figure 2 In the communication method shown, when CSI needs to be transmitted between the terminal device and the network device, the network device can send CSI generation model information to the terminal device so that the terminal device deploys the CSI generation model in the encoder according to the CSI generation model information. In addition, the network device deploys the CSI reconstruction model in the decoder to realize the deployment of a bilateral AI model with CSI feedback enhancement.

[0162] In some embodiments, step 202 may include:

[0163] In a case where the terminal device supports the CSI generation model of a version corresponding to the first CSI generation model information, the terminal device deploys the first CSI generation model in the encoder according to the first CSI generation model information.

[0164] The terminal device supports the CSI generation model version corresponding to the first CSI generation model information, which can be understood as the terminal device being able to deploy the CSI generation model corresponding to the first CSI generation model information, or the terminal device being compatible with the CSI generation model corresponding to the first CSI generation model information.

[0165] After receiving the first CSI generation model information, the terminal device may first determine the version of the first CSI generation model. Then, based on the terminal device's capability information, it may determine whether the terminal device supports that version of the CSI generation model. If the terminal device supports that version of the CSI generation model, the first CSI generation model may be deployed in the encoder based on the first CSI generation model information. The terminal device may determine the version of the first CSI generation model based on the version information of the first CSI generation model included in the metadata layer information of the first CSI generation model. Version information can be understood as version information.

[0166] If the terminal device does not support the CSI generation model version, the terminal device may send version incompatibility information to the network device. After receiving the version incompatibility information from the terminal device, the network device may send third indication information to the terminal device to instruct it to switch the CSI feedback mode. The third indication information may indicate switching the CSI generation model version, switching the CSI generation model, or switching the CSI feedback mode.

[0167] When the third indication information indicates switching the version of the CIS generation model, the second indication information may include information about the version after the switching. When the third indication information indicates switching the version of the CIS generation model, it indicates that the CSI generation model after the switching is the same as the CSI generation model before the switching, but the version is different.

[0168] When the third indication information indicates switching the CSI generation model, the third indication information may include second CSI generation model information. The second CSI generation model information is information about the second CSI generation model. The first CSI generation model is different from the second CSI generation model.

[0169] In the case where the third indication information indicates switching the CSI feedback mode, the third indication information may include a traditional CSI feedback mode. The traditional CSI feedback mode is a traditional CSI feedback mode.

[0170] As can be seen, if the terminal device supports the version of the CSI generation model, the terminal device can deploy the CSI generation model, avoiding deployment of the CSI generation model version that the terminal device does not support, thus avoiding unnecessary deployment and reducing the power consumption of the terminal device. In addition, it can also avoid CSI reporting failures caused by deployment of the CSI generation model version that the terminal device does not support, thus reducing the transmission of unnecessary information and saving communication resources.

[0171] Based on the above network architecture, Figure 3 This is a flow chart of another communication method provided by an embodiment of the present application. Figure 3 As shown, the communication method may include the following steps.

[0172] 301. The network device sends first CSI generation model information to the terminal device.

[0173] Correspondingly, the terminal device receives the first CSI generation model information from the network device.

[0174] The detailed description of step 301 can refer to step 201 and will not be repeated here.

[0175] 302. The terminal device deploys the first CSI generation model in the encoder according to the first CSI generation model information.

[0176] The detailed description of step 302 can refer to the above related description and will not be repeated here.

[0177] 303. The network device deploys a first CSI reconstruction model in a decoder.

[0178] The detailed description of step 303 can refer to the description of step 203 and will not be repeated here.

[0179] 304. The terminal device generates a first CSI using a first CSI generation model.

[0180] The terminal device may monitor the inference performance of the CSI model periodically or regularly, or may monitor the inference performance of the CSI model when a trigger condition is met.

[0181] When the monitoring period is reached, the monitoring time is reached, or the trigger condition is satisfied, the terminal device may generate the first CSI using the first CSI generation model.

[0182] The terminal device may also determine third CSI. The third CSI is the original CSI corresponding to the first CSI, that is, the original CSI corresponding to the first CSI determined by the terminal device, that is, the CSI corresponding to the first CSI not generated using the first CSI generation model, that is, uncompressed CSI.

[0183] 305. The terminal device sends the first CSI to the network device.

[0184] Accordingly, the network device receives the first CSI from the terminal device.

[0185] 306. The network device reconstructs the first CSI using the first CSI reconstruction model to obtain second CSI.

[0186] 307. The network device sends the second CSI to the terminal device.

[0187] Accordingly, the terminal device receives the second CSI from the terminal device.

[0188] After receiving the first CSI, the network device can reconstruct the first CSI using the first CSI reconstruction model to obtain second CSI, and then send the second CSI to the terminal device. The second CSI can be carried in DCI, RRC message or MAC CE.

[0189] 308. The terminal device determines the similarity between the second CSI and the third CSI.

[0190] After receiving the second CSI, the terminal device may determine the similarity between the second CSI and the third CSI. The terminal device may determine the square of the generalized cosine similarity (SGCS) between the second CSI and the third CSI to obtain the similarity between the second CSI and the third CSI, or may determine the normalized mean squared error (NMSE) between the second CSI and the third CSI to obtain the similarity between the second CSI and the third CSI, or may determine the similarity between the second CSI and the third CSI by other means.

[0191] 309. When the similarity is less than the threshold, the terminal device sends a CSI feedback method switching request to the network device.

[0192] Correspondingly, the network device receives a CSI feedback method switching request from the terminal device.

[0193] After determining the similarity between the second CSI and the third CSI, the terminal device may determine whether the similarity is less than a threshold. If the similarity is less than the threshold, it indicates that the inference performance of the CSI model is low, and a CSI feedback method switch request may be sent to the network device. The CSI feedback method switch request is used to request a switch in the CSI feedback method. If the similarity is greater than or equal to the threshold, it indicates that the inference performance of the CSI model is high, and the terminal device may continue to monitor the inference performance of the CSI model.

[0194] 310. The network device sends information about the target CSI feedback mode to the terminal device.

[0195] Correspondingly, the terminal device receives information on the target CSI feedback mode from the network device.

[0196] After receiving the CSI feedback method switching request from the terminal device, the network device may first determine the target CSI feedback method. The target CSI feedback method may include a target CSI feedback mode and a target CSI reconstruction mode. The network device may send information about the target CSI feedback mode to the terminal device. The target CSI feedback method corresponds to the traditional CSI feedback method, or the second CSI generation model and the second CSI reconstruction model, that is, the target CSI feedback mode corresponds to the traditional CSI feedback mode or the second CSI generation model, and the target CSI reconstruction mode corresponds to the traditional CSI reconstruction mode or the second CSI reconstruction model. The first CSI generation model and the second CSI generation model. The first CSI reconstruction model and the second CSI reconstruction model are different. The second CSI generation model corresponds to the second CSI reconstruction model.

[0197] It can be seen that the target CSI feedback method can be an AI / ML CSI feedback method. Accordingly, the target CSI feedback method corresponds to the second CSI generation model and the second CSI reconstruction model.

[0198] It can be seen that the target CSI feedback method may also be a traditional CSI feedback method. Accordingly, the target CSI feedback method corresponds to the traditional CSI feedback method.

[0199] The information of the target CSI feedback mode may be carried in an RRC message, DCI or MAC CE.

[0200] 311. The terminal device switches the CSI feedback mode to the target CSI feedback mode according to information of the target CSI feedback mode.

[0201] After receiving the information of the target CSI feedback mode, the terminal device can switch the CSI feedback mode to the target CSI feedback mode according to the information of the target CSI feedback mode.

[0202] When the target CSI feedback mode corresponds to the second CSI generation model, the terminal device may first delete the first CSI generation model deployed in the encoder, and then deploy the second CSI generation model in the encoder based on information of the second CSI generation model.

[0203] After the terminal device successfully deploys the second CSI generation model, it may send information indicating that the deployment of the second CSI generation model is successful or complete to the network device.

[0204] 312. The network device switches the reconstruction mode to a target CSI reconstruction mode corresponding to the target CSI feedback mode.

[0205] After the network device sends the information of the target CSI feedback mode to the terminal device, the reconstruction mode may be switched to the target CSI reconstruction mode corresponding to the target CSI feedback mode.

[0206] After receiving the information indicating that the second CSI generation model is successfully deployed from the terminal device, the network device may switch the reconstruction mode to the target CSI reconstruction mode corresponding to the target CSI feedback mode.

[0207] exist Figure 3 In the communication method shown, the terminal device monitors the inference performance of the CSI model. When the inference performance of the CSI model is low, the CSI feedback method is switched. This can avoid the situation where the CSI model with low inference performance continues to be used, thereby ensuring the accuracy of CSI reporting.

[0208] Based on the above network architecture, Figure 4 This is a flow chart of another communication method provided by the embodiment of the present application. Figure 4 As shown, the communication method may include the following steps.

[0209] 401. The network device sends first CSI generation model information to the terminal device.

[0210] Correspondingly, the terminal device receives the first CSI generation model information from the network device.

[0211] The detailed description of step 401 can refer to step 201 and will not be repeated here.

[0212] 402. The terminal device deploys the first CSI generation model in the encoder according to the first CSI generation model information.

[0213] The detailed description of step 402 can refer to the above related description and will not be repeated here.

[0214] 403. The network device deploys a first CSI reconstruction model in a decoder.

[0215] The detailed description of step 403 can refer to the description of step 203 and will not be repeated here.

[0216] 404. The terminal device generates a first CSI using a first CSI generation model.

[0217] The network device may monitor the inference performance of the CSI model periodically or regularly, or may monitor the inference performance of the CSI model when a trigger condition is met.

[0218] When the monitoring period is reached, the monitoring time is reached, or the trigger condition is satisfied, the terminal device may generate the first CSI using the first CSI generation model.

[0219] The terminal device may also determine a third CSI, which is the original CSI corresponding to the first CSI.

[0220] 405. The terminal device sends the first CSI to the network device.

[0221] Accordingly, the network device receives the first CSI from the terminal device.

[0222] In one scenario, the terminal device may send the first CSI and the third CSI to the network device. Correspondingly, the network device may receive the first CSI and the third CSI from the terminal device. Because the first CSI and the third CSI are transmitted together, the amount of information transmitted can be reduced, thereby conserving communication resources.

[0223] In another scenario, the terminal device can send the third CSI to the network device. The network device can receive the third CSI from the terminal device. Therefore, the third CSI can be transmitted through dedicated signaling, ensuring the accuracy and security of the third CSI transmission and improving the accuracy of the CSI model's inference performance monitoring.

[0224] 406. The network device reconstructs the first CSI using the first CSI reconstruction model to obtain second CSI.

[0225] 407. The network device determines the similarity between the second CSI and the third CSI.

[0226] The network device determines the similarity between the second CSI and the third CSI in the same manner as the terminal device determines the similarity between the second CSI and the third CSI. For detailed description, please refer to the relevant description below step 308 and will not be repeated here.

[0227] 408. When the similarity is less than the threshold, the network device sends first indication information including information on the target CSI feedback mode to the terminal device.

[0228] Correspondingly, the terminal device receives first indication information including information on the target CSI feedback mode from the network device.

[0229] After determining the similarity between the second CSI and the third CSI, the network device may determine whether the similarity is less than a threshold. If the similarity is less than the threshold, this indicates that the CSI model's inference performance is low. The network device may first determine a target CSI feedback method, and then send first indication information to the terminal device. The first indication information includes information about the target CSI feedback method. The first indication information is used to indicate a switch in the CSI feedback method. If the similarity is greater than or equal to the threshold, this indicates that the CSI model's inference performance is high, and the network device may continue to monitor the CSI model's inference performance.

[0230] The first indication information may be carried in an RRC message, a DCI or a MAC CE.

[0231] For other detailed descriptions of step 408, please refer to the relevant descriptions below step 310, which will not be repeated here.

[0232] 409. The terminal device switches the CSI feedback mode to the target CSI feedback mode according to the first indication information.

[0233] The detailed description of step 409 can refer to the description of step 311 and will not be repeated here.

[0234] 410. The network device switches the reconstruction mode to a target CSI reconstruction mode corresponding to the target CSI feedback mode.

[0235] The detailed description of step 410 can refer to the description of step 312 and will not be repeated here.

[0236] exist Figure 4 In the communication method shown, the network device monitors the inference performance of the CSI model. When the inference performance of the CSI model is low, the CSI feedback method is switched. This can avoid the situation where the CSI model with low inference performance continues to be used, thereby ensuring the accuracy of CSI reporting and reducing the power consumption of the terminal device.

[0237] Based on the above network architecture, Figure 5 This is a flow chart of another communication method provided by the embodiment of the present application. Figure 5 As shown, the communication method may include the following steps.

[0238] 501. The server sends first generation parameter information and first reconstruction parameter information.

[0239] Correspondingly, the terminal device receives the first generation parameter information from the server, and the network device receives the first reconstruction parameter information from the server.

[0240] The first generation parameter information is information about initial parameters of the initial CSI generation model. The first reconstruction parameter information is information about initial parameters of the initial CSI reconstruction model.

[0241] The server may broadcast CSI parameter information. The CSI parameter information may include first generation parameter information and first reconstruction parameter information. The CSI parameter information may include parameter information for multiple CSI models. The parameter information for a CSI model may include parameter information for a CSI generation model and parameter information for a CSI reconstruction model. The parameter information for the CSI generation model may include one or more of: metadata layer information, core data layer information, extended data layer information, and generation parameter information for the CSI generation model. The parameter information for the CSI reconstruction model may include one or more of: metadata layer information, core data layer information, extended data layer information, and reconstruction parameter information for the CSI reconstruction model.

[0242] After the terminal device receives the broadcast CSI parameter information, it can determine whether an initial CSI generation model, that is, the CSI generation model to be trained, is deployed in the encoder of the terminal device. If the initial CSI generation model is deployed, the CSI parameter information can be parsed, and the first generation parameter information can be selected from the parsed information based on the information of the deployed initial CSI generation model.

[0243] After receiving the broadcast CSI parameter information, the network device can determine whether an initial CSI reconstruction model, that is, a CSI generation model to be trained, is deployed in the decoder of the network device. If the initial CSI reconstruction model is deployed, the CSI parameter information can be parsed, and the first reconstruction parameter information can be selected from the parsed information based on the information of the deployed initial CSI reconstruction model.

[0244] 502. The terminal device updates the initial parameters of the initial CSI generation model deployed in the encoder according to the first generation parameter information to obtain a third CSI generation model.

[0245] After the terminal device obtains the first generation parameter information, it can determine the first generation parameter based on the first generation parameter information, and then update the initial parameters of the initial CSI generation model deployed in the encoder based on the first generation parameter, that is, determine the initial parameters of the initial CSI generation model deployed in the encoder as the first generation parameter, and obtain the third CSI generation model.

[0246] 503. The terminal device trains the third CSI generation model to obtain a fourth CSI generation model.

[0247] The fourth CSI generation model may be a CSI generation model obtained after the terminal device trains the third CSI generation model once, or may be a CSI generation model obtained after the terminal device trains the third CSI generation model multiple times.

[0248] 504. The terminal device determines first gradient information according to parameters of the fourth CSI generation model.

[0249] The first gradient information is the first gradient information. The gradient is the partial derivative of the loss function with respect to the model parameters, indicating the direction and magnitude of the parameter adjustment.

[0250] 505. The terminal device sends first gradient information to the server.

[0251] Correspondingly, the server receives the first gradient information from the terminal device.

[0252] 506. The network device updates the initial parameters of the initial CSI reconstruction model deployed in the decoder according to the first reconstruction parameter information to obtain a third CSI reconstruction model.

[0253] After the network device obtains the first reconstruction parameter information, it can determine the first reconstruction parameter based on the first reconstruction parameter information, and then update the initial parameters of the initial CSI reconstruction model deployed in the encoder based on the first reconstruction parameter, that is, determine the initial parameters of the initial CSI reconstruction model deployed in the encoder as the first reconstruction parameter, and obtain the third CSI reconstruction model.

[0254] 507. The network device trains the third CSI reconstruction model to obtain a fourth CSI reconstruction model.

[0255] The fourth CSI reconstruction model may be a CSI reconstruction model obtained after the network device trains the third CSI reconstruction model once, or may be a CSI reconstruction model obtained after the network device trains the third CSI reconstruction model multiple times.

[0256] 508. The network device determines second gradient information according to parameters of the fourth CSI reconstruction model.

[0257] The second gradient information is information of the second gradient.

[0258] 509. The network device sends the second gradient information to the server.

[0259] Accordingly, the server receives the second gradient information from the network device.

[0260] 510. The server determines second generation parameter information according to the first gradient information, and determines second reconstruction parameter information according to the second gradient information.

[0261] After the server receives the first gradient information and the second gradient information, it can determine the second generation parameter information based on the first gradient information, that is, it can determine the first gradient based on the first gradient information, and can aggregate the first gradient to obtain the second generation parameter information, and can determine the second reconstruction parameter based on the second gradient information, that is, it can determine the second gradient based on the second gradient information, and can aggregate the second gradient to obtain the second reconstruction parameter information.

[0262] 511. The server sends second generation parameter information to the terminal device.

[0263] Correspondingly, the terminal device receives the second generation parameter information from the server.

[0264] The server may send the second generated parameter information to the terminal device via an RRC message, DCI, or MAC CE. That is, the second generated parameter information may be carried in the RRC message, DCI, or MAC CE. Thus, the parameter information may be transmitted via dedicated signaling, thereby ensuring the reliability and security of the parameter information.

[0265] 512. The terminal device updates the parameters of the fourth CSI generation model according to the second generation parameter information to obtain a fifth CSI generation model.

[0266] The terminal device can determine the second generation parameter based on the second generation parameter information, and then update the parameters of the fourth CSI generation model based on the second generation parameter to obtain the fifth CSI generation model.

[0267] 513. When the fifth CSI generation model converges, the terminal device determines the first CSI generation model as the fifth CSI generation model.

[0268] After completing a round of CSI model training, i.e., a round of federated learning iterations, the terminal device can determine whether the fifth CSI generation model has converged. If the fifth CSI generation model has converged, the first CSI generation model can be determined as the fifth CSI generation model. If the fifth CSI generation model has not converged, the terminal device can proceed to the next round of CSI model training.

[0269] 514. The server sends second reconstruction parameter information to the network device.

[0270] Correspondingly, the network device receives second reconstruction parameter information from the server.

[0271] The server may send the second reconstruction parameter information to the terminal device via an RRC message, DCI, or MAC CE. That is, the second reconstruction parameter information may be carried in an RRC message, DCI, or MAC CE. Thus, the parameter information may be transmitted via dedicated signaling, thereby ensuring the reliability and security of the parameter information.

[0272] 515. The network device updates the parameters of the fourth CSI reconstruction model according to the second reconstruction parameter information to obtain a fifth CSI reconstruction model.

[0273] The terminal device can determine the second reconstruction parameter based on the second reconstruction parameter information, and then update the parameters of the fourth CSI reconstruction model based on the second reconstruction parameter to obtain the fifth CSI reconstruction model.

[0274] 516. When the fifth CSI network model converges, the network device determines the first CSI reconstruction model as the fifth CSI reconstruction model.

[0275] After completing a round of CSI model training, i.e., a round of federated learning iterations, the network device can determine whether the fifth CSI reconstruction model has converged. If the fifth CSI reconstruction model has converged, CSI model training is complete, and the first CSI network model can be determined to be the fifth CSI reconstruction model. If the fifth CSI reconstruction model has not converged, the network device can proceed to the next round of CSI model training.

[0276] It should be understood that the CSI model training is completed only when the fifth CSI generation model and the fifth CSI reconstruction model converge simultaneously.

[0277] In some embodiments, after completing a round of CSI model training, the server can determine whether the fifth CSI generation model has converged and whether the fifth CSI reconstruction model has converged. If both the fifth CSI generation model and the fifth CSI generation model have converged, the first CSI generation model can be determined to be the fifth CSI generation model, and the first CSI network model can be determined to be the fifth CSI reconstruction model.

[0278] 517. The network device sends first CSI generation model information to the terminal device.

[0279] Correspondingly, the terminal device receives the first CSI generation model information from the network device.

[0280] The detailed description of step 517 can refer to step 201 and will not be repeated here.

[0281] 518. The terminal device deploys the first CSI generation model in the encoder according to the first CSI generation model information.

[0282] The detailed description of step 518 can refer to the above related description and will not be repeated here.

[0283] 519. The network device deploys a first CSI reconstruction model in the decoder.

[0284] The detailed description of step 519 can refer to the description of step 203 and will not be repeated here.

[0285] exist Figure 5 In the communication method shown, federated learning of the CSI model is performed through interaction between the server, network device, and terminal device. This ensures the accuracy and reliability of the trained CSI model, as well as security and communication resource conservation. Furthermore, when CSI needs to be transmitted between the terminal device and the network device, the network device can send CSI generation model information to the terminal device. The terminal device then deploys the CSI generation model in the encoder based on the CSI generation model information, and the network device deploys the CSI reconstruction model in the decoder, thereby implementing the deployment of a bilateral AI model enhanced with CSI feedback.

[0286] Based on the above network architecture, Figure 6 This is a flow chart of another communication method provided by the embodiment of the present application. Figure 6 As shown, the communication method may include the following steps.

[0287] 601. The server sends first generation parameter information and first reconstruction parameter information.

[0288] Correspondingly, the terminal device receives the first generation parameter information from the server, and the network device receives the first reconstruction parameter information from the server.

[0289] The detailed description of step 601 can refer to the description of step 501 and will not be repeated here.

[0290] 602. The network device sends second indication information for indicating a target quantization dictionary to the terminal device.

[0291] Correspondingly, the terminal device receives second indication information for indicating the target quantization dictionary from the network device.

[0292] When training the CSI model is required, the network device may send second indication information to the terminal device. The second indication information is used to indicate a target quantization dictionary. The second indication information may explicitly indicate the target quantization dictionary or implicitly indicate the target quantization dictionary. When the second indication information explicitly indicates the target quantization dictionary, the second indication information may include information about the target quantization dictionary, for example, the second indication information may include an identifier of the target quantization dictionary. When the second indication information implicitly indicates the target quantization dictionary, the second indication information may include associated information of the target quantization dictionary.

[0293] The second indication information may be carried in DCI, RRC message or MAC CE.

[0294] 603. The terminal device determines a target quantization dictionary according to the second indication information.

[0295] After receiving the second indication information, the terminal device may determine the target quantization dictionary according to the second indication information.

[0296] In some embodiments, the second indication information may include an identifier of a target quantization dictionary. The network device may send dictionary library information to the terminal device. Accordingly, the terminal device may receive dictionary library information from the network device and, based on the dictionary library information and the identifier of the target quantization dictionary, obtain the target quantization dictionary from the dictionary library.

[0297] In the case where the quantization dictionary is stored in a dictionary library, the network device may send information about the dictionary library to the terminal device. In the case where the dictionary library is a public dictionary library, the network device may send information about the dictionary library to the terminal device by broadcasting. The information about the dictionary library may be carried in a system information block (SIB) message or in other broadcast messages. In the case where the dictionary library is a non-public dictionary library, the network device may send information about the dictionary library to the terminal device via DCI, RRC message, or MAC CE. The information about the dictionary library may be an index of the dictionary library, an address of the dictionary library, or information about other dictionary libraries.

[0298] The terminal device may obtain the target quantization dictionary from the dictionary library according to the information of the dictionary library and the identifier of the target quantization dictionary, that is, obtain the target quantization dictionary corresponding to the identifier of the target quantization dictionary from the dictionary library corresponding to the information of the dictionary library.

[0299] 604. The terminal device updates the initial parameters of the initial CSI generation model deployed in the encoder according to the first generation parameter information to obtain a third CSI generation model.

[0300] The detailed description of step 604 can refer to the description of step 502 and will not be repeated here.

[0301] 605. The terminal device trains the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model.

[0302] In some embodiments, the terminal device determines a target codebook based on a target quantization dictionary, and may train a third CSI generation model based on the target quantization dictionary and the target codebook to obtain a fourth CSI generation model.

[0303] There is a corresponding relationship between the quantization dictionary and the CSI codebook. For example, when the quantization dictionary identifier is 0X01, the corresponding codebook is [0, 0.2, 0.4, ...].

[0304] In some embodiments, when there is no conflict in the target quantization dictionary, the terminal device may train the third CSI generation model according to the target quantization dictionary to obtain the fourth CSI generation model.

[0305] After the terminal device determines the target quantization dictionary, it can determine whether there is a conflict in the target quantization dictionary. If there is no conflict in the target quantization dictionary, the terminal device can train the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model. If there is a conflict in the target quantization dictionary, the terminal device can send a quantization dictionary reallocation request to the network device via a physical uplink control channel (PUCCH) to request the reallocation of the quantization dictionary. After receiving the quantization dictionary reallocation request, the network device can send fourth indication information to the terminal device. Accordingly, the terminal device can receive fourth indication information from the network device. The fourth indication information can indicate another quantization dictionary.

[0306] The terminal device can determine whether the target quantization dictionary is valid. If the target quantization dictionary is valid, it can be determined that there is no conflict in the target quantization dictionary. If the target quantization dictionary is invalid, it can be determined that there is a conflict in the target quantization dictionary. If the target quantization dictionary exists in the dictionary library, it can be determined that the target quantization dictionary is valid. If the target quantization dictionary does not exist in the dictionary library, it can be determined that the target quantization dictionary is invalid.

[0307] The terminal device may also check whether the target quantization dictionary has been used. If the target quantization dictionary has not been used, it may be determined that there is no conflict in the target quantization dictionary. If the target quantization dictionary has been used, it may be determined that there is a conflict in the target quantization dictionary.

[0308] It can be seen that the target quantization dictionary is used only when there is no conflict in the target quantization dictionary. This ensures that the use of the target quantization dictionary will not affect the training of the CSI model, and can ensure the accuracy and reliability of the CSI model training.

[0309] For other detailed descriptions of step 605 , please refer to the description of step 503 , which will not be repeated here.

[0310] 606. The terminal device determines first gradient information according to parameters of the fourth CSI generation model.

[0311] The detailed description of step 606 can refer to the description of step 504 and will not be repeated here.

[0312] 607. The terminal device sends first gradient information to the server.

[0313] Correspondingly, the server receives the first gradient information from the terminal device.

[0314] 608. The network device updates the initial parameters of the initial CSI reconstruction model deployed in the decoder according to the first reconstruction parameter information to obtain a third CSI reconstruction model.

[0315] The detailed description of step 608 can refer to the description of step 506 and will not be repeated here.

[0316] 609. The network device trains the third CSI reconstruction model according to the target quantization dictionary to obtain a fourth CSI reconstruction model.

[0317] For other detailed descriptions of step 609 , please refer to the description of step 507 , which will not be repeated here.

[0318] 610. The network device determines second gradient information according to parameters of the fourth CSI reconstruction model.

[0319] The detailed description of step 610 can refer to the description of step 508 and will not be repeated here.

[0320] 611. The network device sends second gradient information to the server.

[0321] Accordingly, the server receives the second gradient information from the network device.

[0322] 612. The server determines second generation parameter information according to the first gradient information, and determines second reconstruction parameter information according to the second gradient information.

[0323] The detailed description of step 612 can refer to the description of step 510 and will not be repeated here.

[0324] 613. The server sends second generation parameter information to the terminal device.

[0325] Correspondingly, the terminal device receives the second generation parameter information from the server.

[0326] The detailed description of step 613 can refer to the description of step 511 and will not be repeated here.

[0327] 614. The terminal device updates the parameters of the fourth CSI generation model according to the second generation parameter information to obtain a fifth CSI generation model.

[0328] The detailed description of step 614 can refer to the description of step 512 and will not be repeated here.

[0329] 615. When the fifth CSI generation model converges, the terminal device determines the first CSI generation model as the fifth CSI generation model.

[0330] The detailed description of step 615 can refer to the description of step 513 and will not be repeated here.

[0331] 616. The server sends second reconstruction parameter information to the network device.

[0332] Correspondingly, the network device receives second reconstruction parameter information from the server.

[0333] The detailed description of step 616 can refer to the description of step 514 and will not be repeated here.

[0334] 617. The network device updates the parameters of the fourth CSI reconstruction model according to the second reconstruction parameter information to obtain a fifth CSI reconstruction model.

[0335] The detailed description of step 617 can refer to the description of step 515 and will not be repeated here.

[0336] 618. When the fifth CSI network model converges, the network device determines the first CSI reconstruction model as a sixth CSI reconstruction model.

[0337] The detailed description of step 618 can refer to the description of step 516 and will not be repeated here.

[0338] 619. The network device sends first CSI generation model information to the terminal device.

[0339] Correspondingly, the terminal device receives the first CSI generation model information from the network device.

[0340] The detailed description of step 619 can refer to step 201 and will not be repeated here.

[0341] 620. The terminal device deploys the first CSI generation model in the encoder according to the first CSI generation model information.

[0342] The detailed description of step 620 can refer to the above related description and will not be repeated here.

[0343] 621. The network device deploys a first CSI reconstruction model in a decoder.

[0344] The detailed description of step 621 can refer to the description of step 203 and will not be repeated here.

[0345] exist Figure 6 In the communication method shown, the CSI model is federated through interaction between the server, the network device, and the terminal device, which can ensure the accuracy and reliability of the trained CSI model, as well as the security of the terminal device and save communication resources. In addition, the network device indicates to the terminal device the quantization dictionary used in the process of federated learning, which can ensure the consistency of federated learning. Furthermore, when CSI needs to be transmitted between the terminal device and the network device, the network device can send CSI generation model information to the terminal device so that the terminal device deploys the CSI generation model in the encoder according to the CSI generation model information. In addition, the network device deploys the CSI reconstruction model in the decoder to realize the deployment of a bilateral AI model with CSI feedback enhancement.

[0346] It should be understood that the same or corresponding information in different embodiments may be referenced to each other, and the contents and / or steps in different embodiments may be combined with each other.

[0347] It should be understood that in the above embodiments, some steps may be omitted and some steps may be combined. Figure 5 In the corresponding embodiment, steps 517 to 519 may not be included. Figure 6 In a corresponding embodiment, steps 519 to 521 may not be included.

[0348] It should be understood that the above different embodiments can be combined with each other. Figure 3 and Figure 4 The corresponding embodiment can be Figure 5 Corresponding embodiments are combined. For example, Figure 3 and Figure 4 The corresponding embodiment can be Figure 6 The corresponding embodiments are combined.

[0349] Based on the same inventive concept, embodiments of the present application further provide a communication device for implementing the aforementioned communication method. The implementation solution provided by the communication device is similar to the implementation solution described in the aforementioned method. Therefore, the specific limitations in one or more communication device embodiments provided below can be found in the above-mentioned limitations on the communication method and will not be repeated here.

[0350] Based on the above network structure, Figure 7 This is a schematic diagram of the structure of a communication device provided in an embodiment of the present application. The communication device can be applied to a terminal device, which includes an encoder. The communication device may include:

[0351] A receiving module 701 is configured to receive first CSI generation model information from a network device;

[0352] The deployment module 702 is configured to deploy the first CSI generation model in the encoder according to the first CSI generation model information.

[0353] In some embodiments, the first CSI generation model information is carried in an RRC message, DCI or MAC CE.

[0354] In some embodiments, the deployment module 702 is specifically configured to deploy the first CSI generation model in the encoder according to the first CSI generation model information when the terminal device supports the CSI generation model of a version corresponding to the first CSI generation model information.

[0355] In some embodiments, the communication device may further include:

[0356] A first generating module, configured to generate a first CSI using a first CSI generation model;

[0357] A first sending module, configured to send a first CSI to a network device;

[0358] The receiving module 701 is further configured to receive second CSI from a network device, where the second CSI is reconstructed by using a first CSI reconstruction model corresponding to the first CSI generation model.

[0359] A first determining module is configured to determine a similarity between the second CSI and a third CSI, where the third CSI is an original CSI corresponding to the first CSI;

[0360] The first sending module is further configured to send a CSI feedback method switching request to the network device when the similarity is less than a threshold;

[0361] The receiving module 701 is further configured to receive information about a target CSI feedback mode from a terminal device, where the target CSI feedback mode corresponds to a traditional CSI feedback mode, or a second CSI generation model and a second CSI reconstruction model, where the first CSI generation model is different from the second CSI generation model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

[0362] The first switching module is configured to switch the CSI feedback mode to the target CSI feedback mode according to information of the target CSI feedback mode.

[0363] In some embodiments, the communication device may further include:

[0364] A second generating module, configured to generate a first CSI using a first CSI generation model;

[0365] A second sending module, configured to send the first CSI to the network device;

[0366] The receiving module 701 is further configured to receive first indication information from a network device, where the first indication information is used to instruct a CSI feedback mode to switch, and the first indication information includes information about a target CSI feedback mode, where the target CSI feedback mode corresponds to a traditional CSI feedback mode, or a second CSI generation model and a second CSI reconstruction model, where the first CSI generation model is different from the second CSI generation model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

[0367] The second switching module is configured to switch the CSI feedback mode to a target CSI feedback mode according to the first indication information.

[0368] In some embodiments, the receiving module 701 is further configured to receive first generation parameter information from a server;

[0369] The communication device may further include:

[0370] an updating module, configured to update initial parameters of an initial CSI generation model deployed in the encoder according to the first generation parameter information, to obtain a third CSI generation model;

[0371] a training module, configured to train the third CSI generation model to obtain a fourth CSI generation model;

[0372] a second determining module, configured to determine first gradient information according to parameters of a fourth CSI generation model;

[0373] A third sending module, configured to send the first gradient information to the server;

[0374] The receiving module 701 is further configured to receive second generation parameter information from a server, where the second generation parameter information is determined according to the first gradient information;

[0375] The updating module is further configured to update the parameters of the fourth CSI generation model according to the second generation parameter information to obtain a fifth CSI generation model;

[0376] The second determining module is further configured to determine, when the fifth CSI generation model converges, the first CSI generation model as the fifth CSI generation model.

[0377] In some embodiments, the receiving module 701 receives the first generation parameter information from the server, including:

[0378] The first generation parameter information is received from the server in a broadcast manner.

[0379] In some embodiments, the receiving module 701 is further configured to receive second indication information from a network device, where the second indication information is used to indicate a target quantization dictionary;

[0380] The communication device may further include:

[0381] The second determining module is further configured to determine a target quantization dictionary according to the second indication information;

[0382] The training module is specifically configured to train the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model.

[0383] In some embodiments, the second determining module is further configured to determine a target codebook based on a target quantization dictionary;

[0384] The training module trains the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model, including:

[0385] The third CSI generation model is trained according to the target quantization dictionary and the target codebook to obtain a fourth CSI generation model.

[0386] In some embodiments, the training module trains the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model, including:

[0387] When there is no conflict in the target quantization dictionary, the third CSI generation model is trained according to the target quantization dictionary to obtain a fourth CSI generation model.

[0388] In some embodiments, the second indication information includes an identifier of a target quantization dictionary, and the receiving module 701 is further configured to receive information from a dictionary library;

[0389] The second determining module is specifically configured to obtain the target quantization dictionary from the dictionary library according to the information of the dictionary library and the identifier of the target quantization dictionary.

[0390] In some embodiments, the CSI generation model information includes one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

[0391] Based on the above network structure, Figure 8 This is a structural diagram of another communication device provided in an embodiment of the present application. The communication device can be applied to a network device, which includes a decoder. Figure 8 As shown, the communication device may include:

[0392] A sending module 801 is configured to send first CSI generation model information to a terminal device;

[0393] The deployment module 802 is configured to deploy a first CSI reconstruction model in a decoder, where the first CSI reconstruction model corresponds to the first CSI generation model corresponding to the first CSI generation model information.

[0394] In some embodiments, the first CSI generation model information is carried in an RRC message, DCI or MAC CE.

[0395] In some embodiments, the communication device may further include:

[0396] A first receiving module, configured to receive a first CSI from a terminal device;

[0397] A first reconstruction module, configured to reconstruct the first CSI using a first CSI reconstruction model to obtain a second CSI;

[0398] The sending module 801 is further configured to send the second CSI to the terminal device;

[0399] The first receiving module is further configured to receive a CSI feedback method switching request from a terminal device;

[0400] The sending module 801 is further configured to send information about a target CSI feedback mode to the terminal device, where the target CSI feedback mode corresponds to a traditional CSI feedback mode or a second CSI generation model, and the first CSI generation model is different from the second CSI generation model;

[0401] The first switching module is configured to switch the reconstruction mode to a target CSI reconstruction mode corresponding to the target CSI feedback mode, where the target CSI reconstruction mode corresponds to a traditional CSI reconstruction mode or a second CSI reconstruction model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

[0402] In some embodiments, the communication device may further include:

[0403] A second receiving module, configured to receive a first CSI from a terminal device;

[0404] A second reconstruction module, configured to reconstruct the first CSI using the first CSI reconstruction model to obtain a second CSI;

[0405] A first determining module is configured to determine a similarity between the second CSI and a third CSI, where the third CSI is original CSI corresponding to the first CSI generated by the terminal device;

[0406] The sending module 801 is further configured to send first indication information to the terminal device when the similarity is less than a threshold value, where the first indication information is used to instruct the CSI feedback mode to switch, and the first indication information includes information about a target CSI feedback mode, where the target CSI feedback mode corresponds to a traditional CSI feedback mode or a second CSI generation model, and the first CSI generation model is different from the second CSI generation model.

[0407] The second switching module is used to switch the reconstruction mode to a target CSI reconstruction mode corresponding to the target CSI feedback mode, where the target CSI reconstruction mode corresponds to a traditional CSI reconstruction mode or a second CSI reconstruction model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

[0408] In some embodiments, the third receiving module is configured to receive first reconstruction parameter information from a server;

[0409] The communication device may further include:

[0410] an updating module, configured to update initial parameters of an initial CSI reconstruction model deployed in a decoder according to the first reconstruction parameter information, to obtain a third CSI reconstruction model;

[0411] a training module, configured to train the third CSI reconstruction model to obtain a fourth CSI reconstruction model;

[0412] A second determining module, configured to determine second gradient information according to parameters of a fourth CSI reconstruction model;

[0413] The sending module 801 is further configured to send the second gradient information to the server;

[0414] a third receiving module, further configured to receive second reconstruction parameter information from the server, where the second reconstruction parameter information is determined according to the second gradient information;

[0415] The updating module is further configured to update the parameters of the fourth CSI reconstruction model according to the second reconstruction parameter information to obtain a fifth CSI reconstruction model;

[0416] The second determining module is further configured to determine, when the fifth CSI reconstruction model converges, the first CSI reconstruction model as the fifth CSI reconstruction model.

[0417] In some embodiments, the third receiving module receives the first reconstruction parameter information from the server, including:

[0418] The first reconstruction parameter information is received from the server in a broadcast manner.

[0419] In some embodiments, the sending module 801 is further configured to send second indication information to the terminal device, where the second indication information is used to indicate a target quantization dictionary;

[0420] The training module is specifically configured to train the third CSI reconstruction model according to the target quantization dictionary to obtain a fourth CSI reconstruction model.

[0421] In some embodiments, the second determining module is further configured to determine a target codebook based on a target quantization dictionary;

[0422] The training module trains the third CSI reconstruction model according to the target quantization dictionary to obtain a fourth CSI reconstruction model, including:

[0423] The third CSI reconstruction model is trained according to the target quantization dictionary and the target codebook to obtain a fourth CSI reconstruction model.

[0424] In some embodiments, the second indication information includes an identifier of a target quantization dictionary;

[0425] The sending module 801 is further configured to send information of the dictionary library to the terminal device.

[0426] In some embodiments, the CSI generation model information includes one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

[0427] Each module in the above-mentioned communication device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the communication device in hardware form, or can be stored in the memory of the communication device in software form, so that the processor can call and execute the corresponding operations of each module.

[0428] In an exemplary embodiment, Figure 9A schematic structural diagram of a communication device provided in an embodiment of the present application. The communication device can be a terminal device or a network device. The communication device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the communication device is used to provide computing and control capabilities. The memory of the communication device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the communication device is used to exchange information between the processor and an external device. The communication interface of the communication device is used to communicate with an external device via a network connection. When the computer program is executed by the processor, a communication method is implemented.

[0429] Those skilled in the art will understand that Figure 9 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the communication device to which the scheme of the present application is applied. The specific communication device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0430] In one embodiment, a communication device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0431] In one embodiment, a communication system is further provided, including a terminal device and a network device. Detailed descriptions of the terminal device and the network device can refer to the corresponding steps in the above method embodiment.

[0432] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0433] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0434] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile memory and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a programmable logic unit (PLC), a data processing logic unit based on quantum computing, an artificial intelligence (AI) processor, and the like.

[0435] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0436] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A communication method, characterized in that: Applied to a terminal device, the terminal device includes an encoder, and the method includes: Receiving first channel state information CSI generation model information from a network device; A first CSI generation model is deployed at the encoder according to the first CSI generation model information.

2. The method according to claim 1, characterized in that The first CSI generation model information is carried in a radio resource control RRC message, downlink control information DCI or a media access control MAC control element CE.

3. The method according to claim 1, characterized in that The deploying the first CSI generation model at the encoder according to the first CSI generation model information includes: In a case where the terminal device supports a version of the CSI generation model corresponding to the first CSI generation model information, the first CSI generation model is deployed on the encoder according to the first CSI generation model information.

4. The method according to claim 1, wherein The method further comprises: generating first CSI using the first CSI generation model; sending the first CSI to the network device; receiving second CSI from the network device, where the second CSI is reconstructed by using a first CSI reconstruction model corresponding to the first CSI generation model; determining a similarity between the second CSI and third CSI, where the third CSI is original CSI corresponding to the first CSI; When the similarity is less than a threshold, sending a CSI feedback method switching request to the network device; receiving information of a target CSI feedback mode from the terminal device, where the target CSI feedback mode corresponds to a traditional CSI feedback mode, or a second CSI generation model and a second CSI reconstruction model, where the first CSI generation model is different from the second CSI generation model, and the first CSI reconstruction model is different from the second CSI reconstruction model; According to the information of the target CSI feedback mode, the CSI feedback mode is switched to the target CSI feedback mode.

5. The method according to claim 1, wherein The method further comprises: generating first CSI using the first CSI generation model; sending the first CSI to the network device; receiving first indication information from the network device, where the first indication information is used to instruct a CSI feedback mode switch, the first indication information including information of a target CSI feedback mode, where the target CSI feedback mode corresponds to a traditional CSI feedback mode, or a second CSI generation model and a second CSI reconstruction model, where the first CSI generation model is different from the second CSI generation model, and the first CSI reconstruction model is different from the second CSI reconstruction model; According to the first indication information, the CSI feedback mode is switched to the target CSI feedback mode.

6. The method according to claim 1, wherein The method further comprises: receiving first generation parameter information from a server; updating initial parameters of an initial CSI generation model deployed in the encoder according to the first generation parameter information to obtain a third CSI generation model; training the third CSI generation model to obtain a fourth CSI generation model; determining first gradient information according to parameters of the fourth CSI generation model; sending the first gradient information to the server; receiving second generation parameter information from the server, where the second generation parameter information is determined according to the first gradient information; updating parameters of the fourth CSI generation model according to the second generation parameter information to obtain a fifth CSI generation model; When the fifth CSI generation model converges, the first CSI generation model is determined to be the fifth CSI generation model.

7. The method according to claim 6, characterized in that The receiving first generation parameter information from the server includes: The first generation parameter information is received from the server in a broadcast manner.

8. The method according to claim 6, characterized in that The method further comprises: receiving second indication information from the network device, where the second indication information is used to indicate a target quantization dictionary; determining the target quantization dictionary according to the second indication information; The training of the third CSI generation model to obtain a fourth CSI generation model includes: The third CSI generation model is trained according to the target quantization dictionary to obtain a fourth CSI generation model.

9. The method according to claim 8, characterized in that The method further comprises: determining a target codebook according to the target quantization dictionary; The training of the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model includes: The third CSI generation model is trained according to the target quantization dictionary and the target codebook to obtain a fourth CSI generation model.

10. The method according to claim 8, characterized in that The training of the third CSI generation model according to the target quantization dictionary to obtain a fourth CSI generation model includes: When there is no conflict in the target quantization dictionary, the third CSI generation model is trained according to the target quantization dictionary to obtain a fourth CSI generation model.

11. The method according to claim 8, characterized in that The second indication information includes an identifier of a target quantization dictionary, and the method further includes: Receive information from the dictionary library; The determining the target quantization dictionary according to the second indication information includes: The target quantization dictionary is acquired from the dictionary library according to the information of the dictionary library and the identifier of the target quantization dictionary.

12. The method according to claim 1, characterized in that The CSI generation model information includes one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

13. A communication method, characterized in that: Applied to a network device, the network device including a decoder, the method comprising: Sending first channel state information CSI generation model information to the terminal device; A first CSI reconstruction model is deployed in the decoder, where the first CSI reconstruction model corresponds to the first CSI generation model corresponding to the first CSI generation model information.

14. The method according to claim 13, characterized in that The first CSI generation model information is carried in a radio resource control RRC message, downlink control information DCI or a media access control MAC control element CE.

15. The method according to claim 13, characterized in that The method further comprises: receiving first CSI from the terminal device; reconstructing the first CSI using the first CSI reconstruction model to obtain second CSI; Sending the second CSI to the terminal device; Receiving a CSI feedback method switching request from the terminal device; Sending information of a target CSI feedback mode to the terminal device, where the target CSI feedback mode corresponds to a traditional CSI feedback mode or a second CSI generation model, and the first CSI generation model is different from the second CSI generation model; The reconstruction mode is switched to a target CSI reconstruction mode corresponding to the target CSI feedback mode, where the target CSI reconstruction mode corresponds to a traditional CSI reconstruction mode or a second CSI reconstruction model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

16. The method according to claim 13, characterized in that The method further comprises: receiving first CSI from the terminal device; reconstructing the first CSI using the first CSI reconstruction model to obtain second CSI; Determine a similarity between the second CSI and third CSI, where the third CSI is original CSI corresponding to the first CSI generated by the terminal device; If the similarity is less than a threshold, sending first indication information to the terminal device, where the first indication information is used to instruct a CSI feedback mode to be switched, and the first indication information includes information of a target CSI feedback mode, where the target CSI feedback mode corresponds to a traditional CSI feedback mode or a second CSI generation model, and the first CSI generation model is different from the second CSI generation model; The reconstruction mode is switched to a target CSI reconstruction mode corresponding to the target CSI feedback mode, where the target CSI reconstruction mode corresponds to a traditional CSI reconstruction mode or a second CSI reconstruction model, and the first CSI reconstruction model is different from the second CSI reconstruction model.

17. The method according to claim 13, wherein The method further comprises: receiving first reconstruction parameter information from a server; updating initial parameters of an initial CSI reconstruction model deployed in the decoder according to the first reconstruction parameter information to obtain a third CSI reconstruction model; Training the third CSI reconstruction model to obtain a fourth CSI reconstruction model; determining second gradient information according to parameters of the fourth CSI reconstruction model; sending the second gradient information to the server; receiving second reconstruction parameter information from the server, where the second reconstruction parameter information is determined according to the second gradient information; updating parameters of the fourth CSI reconstruction model according to the second reconstruction parameter information to obtain a fifth CSI reconstruction model; When the fifth CSI reconstruction model converges, the first CSI reconstruction model is determined to be the fifth CSI reconstruction model.

18. The method according to claim 17, characterized in that The receiving first reconstruction parameter information from the server includes: The first reconstruction parameter information is received from the server in a broadcast manner.

19. The method according to claim 17, wherein The method further comprises: Sending second indication information to the terminal device, where the second indication information is used to indicate a target quantization dictionary; The training of the third CSI reconstruction model to obtain a fourth CSI reconstruction model includes: The third CSI reconstruction model is trained according to the target quantization dictionary to obtain a fourth CSI reconstruction model.

20. The method according to claim 19, characterized in that The method further comprises: determining a target codebook according to the target quantization dictionary; The training of the third CSI reconstruction model according to the target quantization dictionary to obtain a fourth CSI reconstruction model includes: The third CSI reconstruction model is trained according to the target quantization dictionary and the target codebook to obtain a fourth CSI reconstruction model.

21. The method according to claim 19, wherein The second indication information includes an identifier of a target quantization dictionary, and the method further includes: The dictionary library information is sent to the terminal device.

22. The method according to claim 13, wherein The CSI generation model information includes one or more of metadata layer information, core parameter layer information, and extended parameter layer information of the CSI generation model.

23. A communication device, characterized in that: Applied to a terminal device, the terminal device includes an encoder, and the apparatus includes: A receiving module, configured to receive first channel state information (CSI) generation model information from a network device; A deployment module is used to deploy a first CSI generation model on the encoder according to the first CSI generation model information.

24. A communication device, characterized in that: Applied to a network device, the network device includes a decoder, and the apparatus includes: A sending module, configured to send first channel state information CSI generation model information to a terminal device; A deployment module is used to deploy a first CSI reconstruction model in the decoder, where the first CSI reconstruction model corresponds to the first CSI generation model corresponding to the first CSI generation model information.

25. A communication device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 22 are implemented.

26. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 22 are implemented.

Images