Communication method and apparatus

Abstract

The present application provides a communication method and apparatus. In the method, a second device sends first indication information to a first device, wherein the first indication information is used for indicating the attribute of an AI model of the second device, and the AI model of the second device can be deployed on a terminal side or on a network side. By means of said method, the attributes of AI models on the terminal side and the network side tend to be consistent, facilitating improvement of the robustness of channel information recovery performance.

Description

Communication method and apparatus TECHNICAL FIELD

[0001] The present disclosure relates to a communication system, and in particular, to a communication method and apparatus for improving the transmission efficiency of feedback information of channel state information-reference signal (CSI-RS). BACKGROUND

[0002] With the continuous development of wireless communication technology, in order to support more services and realize spatial multiplexing between users or data streams, the base station (BS) needs to obtain the channel state information (CSI) of the downlink channel, and then determine the precoding matrix and user scheduling according to the CSI. In the currently widely used frequency division duplex (FDD) based communication system, the uplink and downlink channels are not reciprocal, and the base station needs to obtain the CSI of the downlink channel through the uplink feedback of the user equipment (UE). For example, the base station sends a downlink reference signal to the UE, and the UE receives the downlink reference signal. Since the UE knows the transmission information of the downlink reference signal, the UE can estimate (measure) the downlink channel experienced by the downlink reference signal based on the received downlink reference signal, and then the UE can generate the CSI based on the measured downlink channel matrix, and feed back the CSI to the base station.

[0003] At present, in order to meet the higher requirements of the communication system in system capacity, communication delay and other indicators, the size of the antenna array is continuously increasing, the number of supported antenna ports is increasing, and the dimensions of the corresponding channel matrix and precoding matrix are growing. In this scenario, the overhead of the reference signal issued by the base station increases, and the CSI obtained by the base station is usually compressed to a large extent, and the accuracy of the CSI is low. In order to improve the performance of the communication system, the CSI feedback can be realized based on an artificial intelligence (AI) model. However, how to improve the performance of the AI model is a problem to be solved. SUMMARY

[0004] Embodiments of the present application provide a communication method and apparatus to improve the performance of the AI model.

[0005] In a first aspect, a communication method is provided, which can be executed by a second device or a chip or circuit of the second device.

[0006] Optionally, the second device can be a device at a first AI model side. The device at the first AI model side can be replaced by a device at a terminal side or a device at a network side. The terminal side can include at least one of a terminal device or an AI entity at the terminal side. The AI entity at the terminal side can be the terminal device itself, or an AI entity serving the terminal device, such as a server, e.g., an over the top (OTT) server or a cloud server. The network side can include at least one of a network device or an AI entity at the network side. The AI entity at the network side can be the network device itself, or an AI entity serving the network device, e.g., a radio access network (RAN) intelligent controller (RIC), an operation administration and maintenance (OAM), or a server, e.g., an OTT server or a cloud server.

[0007] For the convenience of description, the second device is described as being at a terminal side in the following.

[0008] The method includes: sending first indication information to a network side, the first indication information being used to indicate an attribute of a first AI model; and the first AI model being deployed at a terminal side.

[0009] Exemplarily, the first AI model attribute can include any one or more of the following: an identification index of the first AI model, a monitoring criterion of the first AI model, a version number of the first AI model, a valid time of the first AI model, a model structure of the first AI model, a capability of the first AI model, and the like.

[0010] The identification index of the AI model can be a model ID (identification / identity) or a sequence number thereof, etc. The monitoring criterion of the AI model can be different indicators for monitoring: an SGCS (squared generalized cosine similarity) or an MSE (Mean Squared Error) or an NMSE (Normalized Mean Squared Error) of reconstructed CSI and true value CSI, etc.; or different threshold settings of the AI monitoring indicators, the threshold corresponding to the indicators is greater than a certain numerical threshold, and the purpose of setting the threshold is to identify a model with strong capability, i.e., a model with high recovery accuracy, i.e., a model greater than or equal to the threshold can be considered as a model with strong capability, such as a model with high CSI recovery accuracy. Optionally, the monitoring criterion of the first AI model can also be replaced by the monitoring criterion of the first AI auto-encoding model (i.e., an auto-encoder) or the second AI model. Optionally, the first AI auto-encoding model includes the first AI model and the second AI model. The version number of the AI model can be one or more of the following: a software version number, a hardware version number, a model number at the time of manufacture, model-associated manufacturer information number, etc. The effective time of the AI model can be directly indicated as the usage period or the expiration date, and the manner is not limited as long as the explicit effective period of the model is clear. The structure of the AI model can be a specific model description, such as at least one of the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the bias in the activation function. The capability of the AI model can be set to strong, medium, weak, etc. according to the recovery accuracy, or different levels of the number of parameters can be set, such as 1K, 1M, 10M, 100M, 1G, etc. The above attributes are all conducive to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0011] In some implementations of the first aspect, the method further includes receiving second indication information from the network side.

[0012] Exemplarily, the second indication information can include any one or more of the following: an identification index of the first data, a version number of the first data, a format of the first data, a value of the first data and / or the first data, an identification index of the second AI model, a version number of the second AI model, a model structure of the second AI model, a parameter value of the second AI model and / or the second AI model, and a file format of the second AI model.

[0013] The identification index of the data can be a data storage ID, or a number of the data according to an order, etc. The version number of the data can be one or more of the following: a software version number, a hardware version number, a number of the data when the data is manufactured, a manufacturer information number associated with the data, etc. The format of the data includes at least one of the following: an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter values, a dimension of input data, a parsing format of output data parameter values, a dimension of output data, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The data can be a single data or a data set. The above parameters are beneficial to realize AI model adjustment or optimization of data values, thereby improving the performance of the AI model.

[0014] The identification index of the AI model can be a model ID, or a number thereof according to an order, etc. The version number of the AI model can be one or more of the following: a software version number, a hardware version number, a number of the model when the model is manufactured, a manufacturer information number associated with the model, etc. The model structure of the AI model can be a specific model description, for example, at least one of the following: a number of neural network layers, a width of the neural network, a connection relationship between layers, a weight of a neuron, an activation function of the neuron, or a bias in the activation function. The AI model parameters obtained through machine learning training can be parameters in the following: a number of neural network layers, a width, a weight of a neuron, or an activation function of the neuron, etc. The file format of the AI model can be.ckpt, SavedModel,.pt or.pth (PyTorch),.onnx (ONNX (Open Neural Network Exchange)),.json,.pb (Protocol Buffers),.h5 or.hdf5, etc. The AI model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning models. The above parameters are beneficial to realize AI model adjustment or optimization, thereby improving the performance of the AI model.

[0015] In some implementations of the first aspect, the method further includes: sending the terminal capability to the network side.

[0016] Exemplarily, the terminal capability can include any one or more of the following: a capability of data that the terminal can support, a capability of a data format that the terminal can support, a capability of a model that the terminal can support, a capability of a model structure that the terminal can support, a capability of a model file format that the terminal can support;

[0017] The data can be single data or a data set. The data format includes at least one of the following: arrangement of input data, arrangement of output data, analysis format of input data parameter value, input data dimension, analysis format of output data parameter value, output data dimension, software update version corresponding to the first data, hardware update version corresponding to the first data, valid time period corresponding to the first data, effective time corresponding to the first data, or invalid time corresponding to the first data. The model herein refers to an AI model, which can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning models. The model structure herein refers to an AI model structure, which can be a specific model description, such as at least one of the following: number of neural network layers, neural network width, connection relationship between layers, neuron weight, neuron activation function, or bias in the activation function. The model file format herein refers to an AI model file format, which can be.ckpt, SavedModel,.pt or.pth (PyTorch),.onnx (ONNX (Open Neural Network Exchange)),.json,.pb (Protocol Buffers),.h5 or.hdf5, etc.

[0018] In some implementations of the first aspect, the method further includes: sending, by the network side, second indication information to the terminal side according to the terminal capability sent by the terminal side; and determining, by the terminal side, the first AI model according to the second indication information. The training in the present application can be initial training, retraining, fine-tuning, or model updating.

[0019] The first AI model can be a device on the network side or a device on the terminal side. For example, the AI model on the terminal side can be used for generating CSI feedback information. The second AI model, which matches the first AI model, such as the AI model on the network side, can be used for recovering channel information corresponding to the CSI feedback information.

[0020] Optionally, the device on the second AI model side can be a device on the network side or a device on the terminal side.

[0021] The first AI model side device obtaining the first indication information from the first device can include a terminal device or a chip for a terminal device obtaining the first indication information from the first device, or an AI entity on the terminal side or a chip for an AI entity on the terminal side obtaining the first indication information from the first device. Optionally, the AI entity on the terminal side or the chip for the AI entity on the terminal side can obtain the first indication information from the first device through forwarding by the terminal device.

[0022] In the scheme of the embodiments of the present application, the consistency of the state information of the first AI model and the state information of the second AI model is ensured through the training reconstruction of the first AI model, thereby the accuracy of the recovered channel information is ensured, and the robustness of the recovery performance of the second AI model on the channel information is improved. For example, in the case of CSI feedback information packet loss, the first AI model is trained and its model attribute is fed back to the first device, and the attributes of the two are consistent when the first device executes the second AI model, thereby the purpose of model monitoring is achieved, and the accuracy of the recovered channel information is improved.

[0023] In a second aspect, a method of communication is provided, which can be executed by the first device or the chip or circuit of the first device.

[0024] The first device can be a device on the side of the second AI model. The device on the side of the second AI model can be replaced by a device on the side of a terminal or a device on the side of a network. The side of the terminal can include at least one of a terminal device or an AI entity on the side of a terminal. The AI entity on the side of the terminal can be the terminal device itself, or an AI entity serving the terminal device, such as a server, for example, an OTT server or a cloud server. The side of the network can include at least one of a network device or an AI entity on the side of a network. The AI entity on the side of the network can be the network device itself, or an AI entity serving the network device, for example, RIC, OAM, or a server, such as an OTT server or a cloud server.

[0025] The following is described for the convenience of description, with the first device being on the side of the network.

[0026] The method includes: receiving, by the network side, first indication information, the first indication information being used to indicate an attribute of a first AI model; and the first AI model being deployed on the side of a terminal.

[0027] Exemplarily, the attribute of the first AI model can include any one or more of the following: an identification index of the first AI model, a monitoring criterion of the first AI model, a version number of the first AI model, a valid time of the first AI model, a model structure of the first AI model, a capability of the first AI model, etc.

[0028] In some implementations of the second aspect, the method further includes: sending second indication information to the side of the terminal.

[0029] Exemplarily, the second indication information can include any one or more of the following: an identification index of the first data, a format of the first data, a value of the first data and / or the first data, an identification index of the second AI model, a model structure of the second AI model, a parameter value of the second AI model and / or the second AI model, a file format of the second AI model.

[0030] In some implementations of the second aspect, the method further includes receiving terminal capability on the terminal side.

[0031] Exemplarily, the terminal capability can include any one or more of the following: capability of data that the terminal can support, capability of data format that the terminal can support, capability of model that the terminal can support, capability of model structure that the terminal can support, capability of model file format that the terminal can support.

[0032] In some implementations of the second aspect, the method further includes: sending, by the network side, second indication information to the terminal side according to the terminal capability sent by the terminal side; and training, by the terminal side, according to the second indication information, and determining the first AI model.

[0033] The first AI model, such as the AI model on the terminal side, can be used for generation of CSI feedback information. The second AI model, such as the AI model on the network side, which matches the first AI model, such as the AI model on the terminal side, can be used for recovery of channel information corresponding to the CSI feedback information.

[0034] Optionally, the device on the first AI model side can be a device on the terminal side or a device on the network side.

[0035] The sending of the first indication information by the device on the second AI model side to the second device can include sending of the first indication information by a network device or a chip for the network device, or sending of the first indication information by an AI entity on the network side or a chip for the AI entity on the network side. Optionally, the AI entity on the network side or the chip for the AI entity on the network side can send the first indication information to the second device through forwarding by the network device.

[0036] In a third aspect, a communication method is provided, which can be executed by a second device, or by a chip or circuit of the second device.

[0037] Optionally, the second device can be a device at a third AI model side. The third AI model side can be replaced by a terminal side or a network side. The terminal side can include at least one of a terminal device or an AI entity at the terminal side. The AI entity at the terminal side can be the terminal device itself, or an AI entity serving the terminal device, such as a server, e.g., an over the top (OTT) server or a cloud server. The network side can include at least one of a network device or an AI entity at the network side. The AI entity at the network side can be the network device itself, or an AI entity serving the network device, e.g., a radio access network (RAN) intelligent controller (RIC), an operation administration and maintenance (OAM), or a server, e.g., an OTT server or a cloud server.

[0038] For convenience of description, the second device is described as a network side in the following.

[0039] The method includes: sending third indication information to the terminal side, the third indication information being used to indicate an attribute of a third AI model; and the third AI model being deployed at the network side.

[0040] The attribute of the third AI model includes any one or more of the following:

[0041] Exemplarily, the attribute of the third AI model can include any one or more of the following: an identification index of the third AI model, a monitoring criterion of the third AI model, a version number of the third AI model, a valid time of the third AI model, a model structure of the third AI model, and a capability of the third AI model.

[0042] In some implementations of the third aspect, the method further includes: receiving fourth indication information from the terminal side.

[0043] Exemplarily, the fourth indication information is used to indicate any one or more of the following: an identification index of the second data, a format of the second data, a value of the second data and / or the second data, an identification index of the fourth AI model, a model structure of the fourth AI model, a parameter value of the fourth AI model and / or the fourth AI model, and a file format of the fourth AI model.

[0044] In some implementations of the third aspect, the method further includes: receiving a terminal capability from the terminal side.

[0045] Exemplarily, the terminal capability can include any one or more of the following: a capability of data that the terminal can support, a capability of data format that the terminal can support, a capability of model that the terminal can support, a capability of model structure that the terminal can support, a capability of model file format that the terminal can support;

[0046] In some implementations of the third aspect, the method further includes: training, by the network side, the third AI model according to the fourth indication information sent by the terminal side.

[0047] In some implementations of the third aspect, the method further includes: training, by the network side, the third AI model according to the fourth indication information sent by the terminal side and / or the terminal capability sent by the terminal side.

[0048] The third AI model can be a device on the network side or a device on the terminal side. Like the AI model on the terminal side, the third AI model can be used for generating CSI feedback information. Like the third AI model, the fourth AI model, i.e., the AI model on the network side, can be used for recovering channel information corresponding to the CSI feedback information.

[0049] Optionally, the device on the third AI model side can be a device on the network side or a device on the terminal side.

[0050] The device on the third AI model side obtaining the first indication information from the first device can include the terminal device or the chip for the terminal device obtaining the third indication information from the third device, or the AI entity on the terminal side or the chip for the AI entity on the terminal side obtaining the third indication information from the third device. Optionally, the AI entity on the terminal side or the chip for the AI entity on the terminal side can obtain the third indication information from the third device through forwarding by the terminal device.

[0051] In a fourth aspect, a method of communication is provided, which can be performed by a first device or a chip or circuit of the first device.

[0052] The first device can be a device on the second AI model side. The device on the second AI model side can be replaced by a device on the terminal side or a device on the network side. The terminal side can include at least one of a terminal device or an AI entity on the terminal side. The AI entity on the terminal side can be the terminal device itself or an AI entity serving the terminal device, such as a server, e.g., an OTT server or a cloud server. The network side can include at least one of a network device or an AI entity on the network side. The AI entity on the network side can be the network device itself or an AI entity serving the network device, such as RIC, OAM, or a server, e.g., an OTT server or a cloud server.

[0053] For the convenience of description, the first device is described as a terminal side.

[0054] The method comprises: receiving, by the terminal side, third indication information, the third indication information being used to indicate an attribute of a third AI model; and deploying the third AI model at a network side.

[0055] The attribute of the third AI model comprises any one or more of the following:

[0056] Exemplarily, the attribute of the third AI model can comprise any one or more of the following: an identification index of the third AI model, a monitoring criterion of the third AI model, a version number of the third AI model, a valid time of the third AI model, a model structure of the third AI model, and a capability of the third AI model.

[0057] In some implementations of the fourth aspect, the method further comprises: sending, to the network side, fourth indication information.

[0058] Exemplarily, the fourth indication information is used to indicate any one or more of the following: an identification index of the second data, a format of the second data, a value of the second data and / or the second data, an identification index of the fourth AI model, a model structure of the fourth AI model, a parameter value of the fourth AI model and / or the fourth AI model, and a file format of the fourth AI model.

[0059] In some implementations of the fourth aspect, the method further comprises: sending, to the network side, a terminal capability.

[0060] Exemplarily, the terminal capability can comprise any one or more of the following: a capability of data that the terminal can support, a capability of a data format that the terminal can support, a capability of a model that the terminal can support, a capability of a model structure that the terminal can support, and a capability of a model file format that the terminal can support.

[0061] In some implementations of the fourth aspect, the method further comprises: training, by the network side, according to the fourth indication information sent by the terminal side, and determining the third AI model.

[0062] In some implementations of the fourth aspect, the method further comprises: training, by the network side, according to the fourth indication information sent by the terminal side and / or the terminal capability sent by the terminal side, and determining the third AI model.

[0063] The third AI model can be a device at the network side or a device at the terminal side. Like the AI model at the network side, the third AI model can be used for recovery of channel information corresponding to CSI feedback information. Like the AI model at the network side, the fourth AI model, which matches the third AI model, can be used for generation of CSI feedback information.

[0064] Optionally, the device on the third AI model side can be a device on the network side or a device on the terminal side.

[0065] The device on the third AI model side obtaining the first indication information from the first device can include that a terminal device or a chip for the terminal device obtains the third indication information from the third device, or an AI entity on the terminal side or a chip for the AI entity on the terminal side obtains the third indication information from the third device. Optionally, the AI entity on the terminal side or the chip for the AI entity on the terminal side can obtain the third indication information from the third device through forwarding of the terminal device.

[0066] In a fifth aspect, a communication apparatus is provided. The communication apparatus can be a terminal device, or a device, module, circuit or chip configured to be arranged in a terminal device, or a device capable of being used in combination with a terminal device. In one design, the communication apparatus can include a module corresponding to each of the steps in the method described in the first aspect. The module can be hardware circuitry, software program or a combination of hardware circuitry and software program. In one design, the communication apparatus can include a processing module and a communication module.

[0067] The sending module is configured to perform the sending actions in the method described in the first aspect, the processing module is configured to perform the processing actions in the method described in the first aspect, and the receiving module is configured to perform the receiving actions in the method described in the first aspect.

[0068] In a sixth aspect, a communication apparatus is provided. The communication apparatus can be a network device, or a device, module, circuit or chip configured to be arranged in a network device, or a device capable of being used in combination with a network device. In one design, the communication apparatus can include a module corresponding to each of the steps in the method described in the second aspect. The module can be hardware circuitry, software program or a combination of hardware circuitry and software program. In one design, the communication apparatus can include a processing module and a communication module.

[0069] The receiving module is configured to perform the receiving actions in the method described in the second aspect, the processing module is configured to perform the processing actions in the method described in the second aspect, and the sending module is configured to perform the sending actions in the method described in the second aspect.

[0070] In a seventh aspect, a communication apparatus is provided. The communication apparatus can include one or more processors coupled with one or more storage media storing instructions that, as a result of execution of the instructions by the one or more processors, cause the method in the first aspect or any possible implementation of the first aspect to be implemented, or cause the method in the second aspect or any possible implementation of the second aspect to be implemented.

[0071] In an eighth aspect, there is provided a communication apparatus comprising one or more processors configured to process data and / or information to cause the method in the first aspect or any possible implementation of the first aspect, or to cause the method in the second aspect or any possible implementation of the second aspect, to be performed.

[0072] Optionally, the communication apparatus can further comprise a communication interface configured to receive data and / or information, and transmit the received data and / or information to the processor. Optionally, the communication interface is further configured to output the data and / or information processed by the processor.

[0073] In a ninth aspect, there is provided a chip comprising a processor configured to execute a program or instructions to cause the method in the first aspect or any possible implementation of the first aspect, or to cause the method in the second aspect or any possible implementation of the second aspect, to be performed.

[0074] Optionally, the chip can further comprise a memory configured to store the program or instructions. Optionally, the chip can further comprise the transceiver.

[0075] In a tenth aspect, there is provided a computer readable storage medium comprising instructions which, when executed by a processor, cause the method in the first aspect or any possible implementation of the first aspect, or the method in the second aspect or any possible implementation of the second aspect, to be performed.

[0076] In an eleventh aspect, there is provided a computer program product comprising computer program code or instructions, which, when executed by a processor, cause the method in the first aspect or any possible implementation of the first aspect, or the method in the second aspect or any possible implementation of the second aspect, to be performed.

[0077] In a twelfth aspect, there is provided a communication system comprising a combination of one or more of the following apparatuses: a communication apparatus performing the method in the first aspect or any possible implementation of the first aspect, a communication apparatus performing the method in the second aspect or any possible implementation of the second aspect. For example, the communication system can comprise the communication apparatus provided in the third aspect, and / or the communication apparatus provided in the fourth aspect. BRIEF DESCRIPTION OF DRAWINGS

[0078] FIG. 1 is a schematic diagram of a communication system suitable for embodiments herein;

[0079] FIG. 2 is a schematic diagram of another communication system suitable for embodiments of the application;

[0080] FIG. 3 is a schematic block diagram of an autoencoder;

[0081] FIG. 4 is a schematic diagram of an AI application framework;

[0082] FIG. 5 is a schematic flow chart of a communication method according to embodiments of the application;

[0083] FIG. 6 is a schematic flow chart of another communication method according to embodiments of the application;

[0084] FIG. 7 is a schematic flow chart of yet another communication method according to embodiments of the application;

[0085] FIG. 8 is a schematic flow chart of yet another communication method according to embodiments of the application;

[0086] FIG. 9 is a schematic block diagram of an apparatus for communication according to embodiments of the application;

[0087] FIG. 10 is a schematic block diagram of another apparatus for communication according to embodiments of the application. DETAILED DESCRIPTION

[0088] The technical solutions in the application will be described below with reference to the accompanying drawings.

[0089] The technical solutions provided by the application can be applied to various communication systems, for example, a 5th generation (5G) or new radio (NR) system, a long term evolution (LTE) system, an LTE frequency division duplex (FDD) system, an LTE time division duplex (TDD) system, a wireless local area network (WLAN) system, a satellite communication system, a future communication system such as a 6th generation (6G) mobile communication system, or a converged system of multiple systems, etc. The technical solutions provided by the application can also be applied to device to device (D2D) communication, vehicle-to-everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and internet of things (IoT) communication system or other communication systems.

[0090] A device in a communication system can send or receive signals to or from another device. The signals can include information, signaling, or data, etc. The device can also be replaced by an entity, network entity, network element, communication device, communication module, node, communication node, etc. The device is taken as an example for description in the disclosure. For example, the communication system can include at least one terminal device and at least one network device. In the communication system, the network device can send a downlink signal to the terminal device, the terminal device can send an uplink signal to the network device, the network device can send a signal to another network device, and the terminal device can send a sidelink signal to another terminal device. It can be understood that the terminal device in the disclosure can be replaced by a second device, and the network device can be replaced by a first device, both of which perform the corresponding communication method in the disclosure.

[0091] In the embodiments of the present application, the terminal device can also be referred to as a user equipment (UE), an access terminal, a user unit, a user station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent, or a user apparatus.

[0092] The terminal device can be a device providing voice / data, for example, a handheld device with wireless connection function, a vehicle-mounted device, etc. At present, some examples of terminals are: mobile phone, tablet computer, notebook computer, palm computer, mobile internet device (MID), wearable device, virtual reality (VR) device, augmented reality (AR) device, wireless terminal in industrial control, wireless terminal in self driving, wireless terminal in telemedicine, wireless terminal in smart grid, wireless terminal in transportation safety, wireless terminal in smart city, wireless terminal in smart home, cellular phone, cordless phone, session initiation protocol (SIP) phone, wireless local loop (WLL) station, personal digital assistant (PDA), handheld device with wireless communication function, computing device or other processing device connected to a wireless modem, wearable device, terminal device in a 5G network, or terminal device in a future evolved public land mobile network (PLMN), etc. The embodiments of the present application are not limited thereto.

[0093] By way of example and not limitation, in the embodiments of the present application, the terminal device can also be a wearable device. The wearable device can also be referred to as a wearable smart device, which is a general term for devices that are designed and developed by applying wearable technology to daily wear, such as glasses, gloves, watches, clothing, and shoes. The wearable device is a portable device that is directly worn on the body or integrated into the user's clothes or accessories. The wearable device is not only a hardware device, but also a device that realizes powerful functions through software support and data interaction and cloud interaction. The general wearable smart device includes a device with full functions and large size, which can realize complete or partial functions without relying on a smart phone, such as a smart watch or smart glasses, and a device that focuses on a certain application function and needs to cooperate with other devices such as a smart phone, such as various smart wristbands and smart jewelry for monitoring vital signs.

[0094] In the embodiments of the present application, the apparatus for implementing the function of the terminal device can be a terminal device, or can be an apparatus capable of supporting the terminal device to implement the function, for example, a chip system, which can be installed in the terminal device or used in matching with the terminal device. In the embodiments of the present application, the chip system can be composed of a chip, or can include the chip and other discrete devices. In the embodiments of the present application, only the apparatus for implementing the function of the terminal device is taken as an example for description, and the scheme of the embodiments of the present application is not limited in this way.

[0095] The network device in the embodiments of the present application can be a device for communicating with a terminal device, and the network device can also be referred to as an access network device or a radio access network device, for example, the network device can be a base station. The network device in the embodiments of the present application can refer to a RAN node (or device) for accessing a terminal device to a wireless network. The base station can broadly cover various names in the following or be replaced by the following names, such as: Node B (NodeB), evolved Node B (eNB), next generation Node B (gNB), relay station, access point, transmitting and receiving point (TRP), transmitting point (TP), primary station, secondary station, multi-standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. The base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The base station can also refer to a communication module, modem or chip used in the foregoing devices or apparatuses. The base station can also be a mobile switching center and a device assuming a base station function in D2D, V2X, M2M communication, a network side device in a 6G network, a device assuming a base station function in a future communication system, etc. The base station can support networks of the same or different access technologies. Optionally, the RAN node can also be a server, a wearable device, a vehicle or a vehicle-mounted device, etc. For example, the access network device in the V2X technology can be a road side unit (RSU). The embodiments of the present application do not limit the specific technology and specific device form of the network device.

[0096] The base station can be fixed or mobile. For example, a helicopter or a drone can be configured to act as a mobile base station, and one or more cells can move according to the location of the mobile base station. In other examples, the helicopter or the drone can be configured to act as a device communicating with another base station.

[0097] In some deployments, the network device mentioned by embodiments of the present application can be a device including a CU, or a DU, or a device including a CU and a DU, or a control plane CU node (central unit-control plane (CU-CP)) and a user plane CU node (central unit-user plane (CU-UP)) and a DU node. For example, the network device can include a gNB-CU-CP, a gNB-CU-UP and a gNB-DU.

[0098] In some deployments, wireless access is assisted for a terminal by multiple RAN nodes cooperating, and different RAN nodes respectively implement part of the functions of a base station. For example, the RAN node can be a CU, a DU, a CU-CP, a CU-UP, or an RU, etc. The CU and the DU can be separately arranged, or can also be included in the same network element, for example, in a BBU. The RU can be included in a radio frequency device or a radio frequency unit, for example, included in an RRU, an AAU or an RRH.

[0099] The RAN node can support one or more types of front-haul interfaces, and different front-haul interfaces respectively correspond to DUs and RUs having different functions.

[0100] If the front-haul interface between the DU and the RU is a common public radio interface (CPRI), the DU is configured to implement one or more of the baseband functions, and the RU is configured to implement one or more of the radio frequency functions.

[0101] If the front-haul interface between the DU and the RU is another interface, compared with the CPRI, part of the baseband functions of the downlink and / or the uplink, such as one or more of precoding, digital beamforming (BF), or inverse fast fourier transform (IFFT) / adding a cyclic prefix (CP) for the downlink, or one or more of digital beamforming (BF), or fast fourier transform (FFT) / removing a cyclic prefix (CP) for the uplink, are moved from the DU to the RU for implementation.

[0102] One possible implementation, the interface can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the split between the DU and the RU is different, corresponding to different categories (Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, F.

[0103] Taking eCPRI Cat A as an example, for downlink transmission, the split is layer mapping, the DU is configured to implement layer mapping and one or more functions before layer mapping (i.e., one or more of encoding, rate matching, scrambling, modulation, layer mapping), and other functions after layer mapping (e.g., one or more of resource element (RE) mapping, digital beamforming (BF), or inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP)) are implemented in the RU. For uplink transmission, the split is de-RE mapping, the DU is configured to implement de-mapping and one or more functions before de-mapping (i.e., one or more of decoding, de-rate matching, de-scrambling, de-modulation, inverse discrete Fourier transform (IDFT), channel equalization, de-RE mapping), and other functions after de-mapping (e.g., one or more of digital BF or fast Fourier transform (FFT) / CP removal) are implemented in the RU. It can be understood that the description of the functions of the DU and the RU corresponding to various types of eCPRI can refer to the eCPRI protocol, which is not described here.

[0104] In one possible design, the processing unit in the BBU for implementing baseband functions is referred to as a base band high (BBH) unit, and the processing unit in the RRU / AAU / RRH for implementing baseband functions is referred to as a base band low (BBL) unit.

[0105] The CU (or CU-CP and CU-UP), DU or RU can also have different names in different systems, but those skilled in the art can understand their meanings. For example, in an open RAN (ORAN) system, the CU can also be referred to as an O-CU (open CU), the DU can also be referred to as an O-DU, the CU-CP can also be referred to as an O-CU-CP, the CU-UP can also be referred to as an O-CU-UP, and the RU can also be referred to as an O-RU. Any of the CUs (or CU-CPs, CU-UPs), DUs and RUs in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0106] In the embodiments of the present application, the apparatus for implementing the function of the network device can be a network device; or can be an apparatus capable of supporting the network device to implement the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. The apparatus can be installed in the network device or used in matching with the network device. In the embodiments of the present application, only the apparatus for implementing the function of the network device is taken as an example for illustration, and the scheme of the embodiments of the present application is not limited.

[0107] The network device and / or the terminal device can be deployed on land, including indoors, outdoors, handheld, and / or vehicle-mounted; can also be deployed on the water surface (such as a ship, etc.); and can also be deployed in the air (such as an airplane, a balloon, and / or a satellite). The scenarios in which the network device and the terminal device are located are not limited in the embodiments of the present application.

[0108] In addition, the terminal device and the network device can be hardware devices, or software functions running on special hardware, software functions running on general hardware, such as virtualized functions instantiated on a platform (for example, a cloud platform), or entities including special or general hardware devices and software functions. The specific forms of the terminal device and the network device are not limited in the present application.

[0109] In a wireless communication network, e.g., in a mobile communication network, the services supported by the network are increasingly diverse, and thus the requirements to be met are increasingly diverse. For example, the network needs to be able to support ultra-high rates, ultra-low latency, and / or ultra-large connections. This feature makes network planning, network configuration, and / or resource scheduling increasingly complex. In addition, as the functions of the network become increasingly powerful, such as supporting increasingly high frequency spectrums, supporting high-order multiple input multiple output (MIMO) technology, supporting beamforming, and / or supporting new technologies such as beam management, network energy saving has become a hot research topic. These new requirements, new scenarios, and new features bring unprecedented challenges to network planning, operation and maintenance, and efficient operation. To meet this challenge, artificial intelligence technology can be introduced into the wireless communication network, thereby realizing network intelligence.

[0110] To support AI technology in a wireless network, AI nodes (also referred to as AI entities) can also be introduced into the network.

[0111] Optionally, the AI entity can be deployed in one or more of the following positions in the communication system: an access network device, a terminal device, or a core network device, etc., or the AI entity can also be deployed separately, e.g., in a position other than any of the above devices, such as a host or a cloud server of an OTT system. The AI entity can communicate with other devices in the communication system, which can be one or more of the following: a network device, a terminal device, or a network element of a core network, etc. Based on the object served by the AI entity, the AI entity can include a network-side AI entity, a terminal-side AI entity, or a core network-side AI entity.

[0112] It can be understood that the present application does not limit the number of AI entities. For example, when there are multiple AI entities, the multiple AI entities can be divided based on functions, e.g., different AI entities are responsible for different functions.

[0113] It can also be understood that the AI entity can be a separate device, can be integrated into the same device to implement different functions, or can be a network element in a hardware device, or a software function running on a dedicated hardware, or a virtualized function instantiated on a platform (e.g., a cloud platform), and the present application does not limit the specific form of the AI entity.

[0114] An AI entity can be an AI network element or an AI module. An AI entity is configured to implement a corresponding AI function. AI modules deployed in different network elements can be the same or different. AI models in an AI entity can implement different functions according to different parameter configurations. An AI model in an AI entity can be configured based on one or more of the following parameters: a structural parameter (e.g., at least one of a number of neural network layers, a width of a neural network, a connection relationship between layers, a weight of a neuron, an activation function of a neuron, or a bias in an activation function), an input parameter (e.g., a type of input parameter and / or a dimension of an input parameter), or an output parameter (e.g., a type of output parameter and / or a dimension of an output parameter). The bias in the activation function can also be referred to as a bias of the neural network.

[0115] One AI entity can have one or more models. The learning process, the training process, or the inference process of different models can be deployed in different entities or devices, or can be deployed in the same entity or device.

[0116] FIG. 1 is a schematic diagram of a communication system applicable to a communication method according to an embodiment of the present application. As shown in FIG. 1, the communication system 100 can include at least one network device, such as the network device 110 shown in FIG. 1, and can include at least one terminal device, such as the terminal device 120 and the terminal device 130 shown in FIG. 1. The network device 110 and the terminal devices (e.g., the terminal device 120 and the terminal device 130) can communicate with each other through wireless links. The communication devices in the communication system, such as the network device 110 and the terminal device 120, can communicate with each other through multi-antenna technology.

[0117] FIG. 2 is a schematic diagram of another communication system applicable to a communication method according to an embodiment of the present application. Compared with the communication system 100 shown in FIG. 1, the communication system 200 shown in FIG. 2 further includes an AI network element 140. The AI network element 140 is configured to perform AI-related operations, such as constructing a training data set or training an AI model.

[0118] In a possible implementation, the network device 110 can send data related to the training of the AI model to the AI network element 140, the AI network element 140 constructs a training data set and trains the AI model. For example, the data related to the training of the AI model can include data reported by the terminal device. The AI network element 140 can send the result of the AI model related operation to the network device 110 and forward the result to the terminal device through the network device 110. For example, the result of the AI model related operation can include at least one of the following: a trained AI model, an evaluation result or a test result of the model, and the like. For example, part of the trained AI model can be deployed on the network device 110, and the other part can be deployed on the terminal device. Alternatively, the trained AI model can be deployed on the network device 110. Alternatively, the trained AI model can be deployed on the terminal device.

[0119] It should be understood that FIG. 2 is only used as an example to illustrate that the AI network element 140 is directly connected to the network device 110, and in other scenarios, the AI network element 140 can also be connected to the terminal device. Alternatively, the AI network element 140 can be connected to both the network device 110 and the terminal device. Alternatively, the AI network element 140 can also be connected to the network device 110 through a third-party network element. The connection relationship between the AI network element and other network elements is not limited in the embodiments of the present application.

[0120] The AI network element 140 can also be arranged as a module in the network device and / or the terminal device, for example, in the network device 110 or the terminal device shown in FIG. 1. One or more AI modules can be deployed in the network device 110. One or more AI modules can be deployed in the terminal device.

[0121] It should be noted that FIGS. 1 and 2 are only simplified schematic diagrams for illustration, for example, the communication system can further include other devices, such as a wireless relay device and / or a wireless backhaul device, which are not shown in FIGS. 1 and 2. In actual application, the communication system can include multiple network devices and multiple terminal devices. The number of network devices and terminal devices included in the communication system is not limited in the embodiments of the present application.

[0122] In order to facilitate understanding of the scheme of the embodiments of the present application, the terms that can be involved in the embodiments of the present application are explained as follows.

[0123] (1) AI model:

[0124] The AI model is an algorithm or computer program that can realize the AI function, and the AI model represents the mapping relationship between the input and the output of the model. The AI model can be understood as a function model that maps a certain dimension of input to a certain dimension of output, and the model parameters are obtained through machine learning training. For example, f(x) = ax2 +b is a quadratic function model, which can be regarded as an AI model, and a and b correspond to parameters of the AI model, which can be obtained by machine learning training. The AI model can also be referred to as a model or an AI function or a function. One AI function can correspond to one or more AI models.

[0125] The type of AI model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q-learning model, or other machine learning (ML) models.

[0126] (2) Two-end model:

[0127] The two-end model can also be referred to as a two-sided model, a collaborative model, a dual model, or a two-side model, etc. The two-end model refers to a model composed of multiple sub-models. The multiple sub-models constituting the model need to match each other. The multiple sub-models can be deployed in different nodes.

[0128] The embodiments of the present application relate to an encoder for compressing channel information and a decoder for restoring channel information. The encoder and the decoder are matched for use, and it can be understood that the encoder and the decoder are matched AI models. One encoder can include one or more AI models, and the decoder matched with the encoder also includes one or more AI models. The number of AI models included in the matched encoder and decoder is the same and one-to-one correspondence. The encoder can also include a quantization module, which can be used to quantize the output of the AI model in the encoder. The decoder can include a dequantization module, which can be used to dequantize the feedback information of the received channel information to obtain the input of the AI model in the decoder. The dequantization processing can also be referred to as dequantization processing.

[0129] In one possible design, a pair of matching encoder and decoder can be two parts of the same auto-encoder (AE). The AE model with the encoder and the decoder deployed at different nodes is a typical bilateral model. The encoder and the decoder of the AE model are usually jointly trained and matched for use. The auto-encoder is a kind of unsupervised learning neural network, which is characterized by taking the input data as the label data, and thus can also be understood as a self-supervised learning neural network. The auto-encoder can be used for data compression and recovery. For example, the encoder in the auto-encoder can compress (encode) the data A to obtain the data B, and the decoder in the auto-encoder can decompress (decode) the data B to recover the data A. Alternatively, it can be understood that the decoder is the inverse operation of the encoder.

[0130] For example, as shown in FIG. 3, the encoder processes the input V to obtain the processed result z, and the decoder can decode the output z of the encoder to the expected output V'.

[0131] The AI model in the embodiments of this application can include an encoder deployed at the terminal side and a decoder deployed at the network side, or an encoder deployed at the terminal side and a decoder deployed at another terminal side, or an encoder deployed at the network side and a decoder deployed at another network side.

[0132] (3) Neural network (NN):

[0133] The neural network is a specific implementation form of AI or machine learning. According to the universal approximation theorem, the neural network can theoretically approximate any continuous function, so that the neural network has the ability to learn any mapping.

[0134] The neural network can be composed of neural units, and a neural unit can refer to an operation unit with x s and intercept 1 as input. The neural network is a network formed by connecting many single neural units described above, that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract the features of the local receptive field, and the local receptive field can be a region composed of several neural units.

[0135] Taking the type of the AI model as a neural network as an example, the AI model involved in the present disclosure can be a deep neural network (DNN). According to the construction mode of the network, the DNN can include a feedforward neural network (FNN), a convolutional neural network (CNN), and a recurrent neural network (RNN), etc.

[0136] The CNN is a neural network specially used for processing data with a similar grid structure. For example, time series data (time axis discrete sampling) and image data (two-dimensional discrete sampling) can be considered as data with a similar grid structure. The CNN does not use all input information for operation at one time, but uses a fixed-size window to intercept part of the information for convolution operation, which greatly reduces the calculation amount of model parameters. In addition, according to the different types of information intercepted by the window (such as people and objects in the same image as different types of information), each window can use different convolution kernel operations, which enables the CNN to better extract the features of the input data.

[0137] The RNN is a type of DNN network that uses feedback time series information. Its input includes a new input value at the current time and an output value of itself at the previous time. The RNN is suitable for obtaining sequence features with correlation in time, and is particularly suitable for speech recognition, channel coding and decoding, etc.

[0138] The FNN network is characterized by complete connection between adjacent layers of neurons, which makes the FNN usually require a large amount of storage space and results in high computational complexity.

[0139] The above-mentioned FNN, CNN, and RNN are all constructed based on neurons. As described above, each neuron performs weighted summation operation on its input value, and the weighted summation result generates an output through a nonlinear function. The weights of the neuron weighted summation operation in the neural network and the nonlinear function are referred to as the parameters of the neural network. The parameters of all neurons of a neural network constitute the parameters of the neural network.

[0140] (4) Dataset:

[0141] The dataset refers to the data used for model training, verification, and testing in machine learning, and the quantity and quality of the data will affect the effect of machine learning.

[0142] In the field of machine learning, the ground truth usually refers to data that is considered to be accurate or real.

[0143] The training data set is used for training of the AI model, and the training data set can include an input of the AI model or an input and a target output of the AI model. The training data set includes one or more training data, and the training data can include a training sample input to the AI model or a target output of the AI model. The target output can also be referred to as a label, a sample label, or a label sample. The label is a true value.

[0144] In the field of communication, the training data set can include simulation data collected through a simulation platform, experimental data collected in an experimental scenario, or measured data collected in an actual communication network. Due to differences in geographical environment and channel conditions, such as differences in indoor, outdoor, mobile speed, frequency band, or antenna configuration, the collected data can be classified when the data is obtained. For example, data with the same channel propagation environment and antenna configuration are classified into one category.

[0145] Model training essentially learns some features from the training data. In the process of training an AI model (such as a neural network model), because the output of the AI model is expected to be as close as possible to the value that is actually intended to be predicted, the weight vector of each layer of the AI model can be updated according to the difference between the predicted value of the current network and the target value that is actually intended to be predicted (of course, before the first update, there is usually an initialization process, that is, the parameters of each layer of the AI model are pre-configured). For example, if the predicted value of the network is too high, the weight vector is adjusted to make it predict lower. The adjustment is continuously made until the AI model can predict the target value that is actually intended to be predicted or a value very close to the target value that is actually intended to be predicted. Therefore, it is necessary to define "how to compare the difference between the predicted value and the target value", which is the loss function or the objective function, which is an important equation for measuring the difference between the predicted value and the target value. Taking the loss function as an example, the higher the output value (loss) of the loss function, the greater the difference. Therefore, the training of the AI model becomes a process of minimizing the loss, so that the value of the loss function is less than a threshold, or so that the value of the loss function meets the target requirement. For example, the AI model is a neural network, and adjusting the model parameters of the neural network includes adjusting at least one of the following parameters: the number of layers of the neural network, the width, the weight of the neuron, or the parameter of the activation function of the neuron.

[0146] The inference data can be used as an input of the AI model that has completed training, for inference of the AI model. In the model inference process, the inference data is input into the AI model, and the corresponding output is obtained as the inference result.

[0147] (5) AI model design:

[0148] The design of the AI model mainly includes a data collection link (e.g., collecting training data and / or inference data), a model training link, and a model inference link. Further, it can also include an inference result application link.

[0149] FIG. 4 shows an AI application framework.

[0150] In the foregoing data collection link, a data source is used to provide a training data set and inference data. In the model training link, an AI model is obtained by analyzing or training the training data provided by the data source. The AI model represents the mapping relationship between the input and output of the model. The AI model is learned through the model training node, which is equivalent to learning the mapping relationship between the input and output of the model using the training data. In the model inference link, the AI model trained through the model training link is used to perform inference based on the inference data provided by the data source, and an inference result is obtained. This link can also be understood as follows: the inference data is input into the AI model, and the output obtained through the AI model is the inference result. The inference result can indicate a configuration parameter used (executed) by an execution object and / or an operation executed by the execution object. In the inference result application link, the inference result is published, for example, the inference result can be uniformly planned by an execution entity, for example, the execution entity can send the inference result to one or more execution objects (e.g., network devices or terminal devices, etc.) to execute. For example, the execution entity can also feed back the performance of the model to the data source, so as to facilitate subsequent implementation of model update training.

[0151] It can be understood that the network element with AI function can be included in the communication system. The above-mentioned AI model design related links can be executed by one or more network elements with artificial intelligence function. In one possible design, the AI function (such as AI module or AI entity) can be configured in the existing network element in the communication system to implement the AI related operation, for example, the training and / or inference of the AI model. For example, the existing network element can be a network device or a terminal device, etc. Or in another possible design, a separate network element can also be introduced in the communication system to execute the AI related operation, such as training the AI model. The separate network element can be referred to as AI network element or AI node, etc., and the name is not limited by the embodiments of the present application. Exemplarily, the AI network element can be directly connected with the network device in the communication system, or can be indirectly connected with the network device through a third party network element. The third party network element can be a core network element such as authentication management function (AMF) network element, user plane function (UPF) network element, OAM, cloud server or other network element, which is not limited. Exemplarily, the separate network element can be deployed in one or more of the network side, the terminal side, or the core network side. Optionally, it can be deployed on the cloud server or OTT or OAM. Exemplarily, as shown in FIG. 2, the AI network element 140 is introduced in the communication system.

[0152] The training processes of different models can be deployed in different devices or nodes, or can be deployed in the same device or node. The inference processes of different models can be deployed in different devices or nodes, or can be deployed in the same device or node. Taking the terminal device to complete the model training link as an example, the terminal device can train the corresponding encoder and decoder, and then send the model parameters of the decoder to the network device. Taking the network device to complete the model training link as an example, the network device can train the corresponding encoder and decoder, and then indicate the model parameters of the encoder to the terminal device. Taking the independent AI network element to complete the model training link as an example, the AI network element can train the corresponding encoder and decoder, and then send the model parameters of the encoder to the terminal device and the model parameters of the decoder to the network device. Further, the model inference link corresponding to the encoder is performed in the terminal device, and the model inference link corresponding to the decoder is performed in the network device.

[0153] The model parameters can include one or more of the following: a structure parameter of a model (e.g., a number of layers of the model, a weight value, and / or the like), an input parameter of the model (e.g., an input dimension, a number of input ports), or an output parameter of the model (e.g., an output dimension, a number of output ports). It can be understood that the input dimension can refer to a size of an input data, for example, when the input data is a sequence, the input dimension corresponding to the sequence can indicate a length of the sequence. The number of input ports can refer to a number of input data. Similarly, the output dimension can refer to a size of an output data, for example, when the output data is a sequence, the output dimension corresponding to the sequence can indicate a length of the sequence. The number of output ports can refer to a number of output data.

[0154] (6) Channel information:

[0155] In a communication system (e.g., an LTE communication system or an NR communication system, etc.), a network device determines one or more of the following configurations of a downlink data channel of a terminal device based on channel information: a resource, a MCS, and a precoding. It can be understood that the channel information can also be referred to as channel state information (CSI) or channel environment information, which is an information capable of reflecting channel characteristics and channel quality.

[0156] Channel information measurement refers to solving channel information by a receiving end according to a reference signal sent by a sending end, i.e., estimating channel information by using a channel estimation method. Exemplarily, the reference signal can include one or more of the following: a channel state information reference signal (CSI-RS), a synchronizing signal / physical broadcast channel block (SSB), a sounding reference signal (SRS), or a demodulation reference signal (DMRS). One or more of the CSI-RS, the SSB, and the DMRS can be used to measure downlink channel information. The SRS and / or the DMRS can be used to measure uplink channel information.

[0157] The channel information can be determined based on a channel measurement result of a reference signal. Alternatively, the channel information can be the channel measurement result of the reference signal. In the embodiments of the present application, the channel measurement result of the reference signal can also be replaced by the channel information.

[0158] Taking a FDD communication scenario as an example, in the FDD communication scenario, since the uplink and downlink channels are not reciprocal or the reciprocity of the uplink and downlink channels cannot be guaranteed, the network device needs to obtain the downlink CSI through the uplink feedback of the terminal device. The network device usually sends a downlink reference signal to the terminal device, and the terminal device receives the downlink reference signal. Since the terminal device knows the transmission information of the downlink reference signal, the terminal device can perform channel measurement and interference measurement estimation (measurement) on the downlink channel experienced by the downlink reference signal according to the received downlink reference signal. The terminal device generates the downlink CSI based on the downlink channel matrix obtained by the measurement. The terminal device generates a CSI report according to a protocol pre-defined manner or a network device configured manner, and feeds back to the network device, so that the network device obtains the downlink CSI.

[0159] In this application, the meaning of CSI is broader than that in the traditional scheme, and is not limited to channel quality indication (CQI), precoding matrix indicator (PMI), rank indicator (RI), or CSI-RS resource indicator (CRI). It can also be one or more of channel response information (such as channel response matrix, frequency domain channel response information, and time domain channel response information), weight information corresponding to the channel response, reference signal receiving power (RSRP), or signal to interference plus noise ratio (SINR).

[0160] Wherein, RI is used to indicate the receiving end of the reference signal, such as the terminal device, the number of layers of the recommended downlink transmission, CQI is used to indicate the modulation and coding mode that the receiving end of the reference signal, such as the terminal device, judges that the current channel condition can support, and PMI is used to indicate the precoding recommended by the receiving end of the reference signal, such as the terminal device. The number of layers of the precoding indicated by PMI corresponds to RI.

[0161] As described previously, the measurement on the reference signal can obtain channel information. The compression and / or quantization operation on the channel information can obtain feedback information. The feedback information can be reported through a channel information report. The decompression and / or dequantization operation on the feedback information can recover the channel information.

[0162] The feedback information can also be referred to as feedback information of channel information, feedback information of CSI, CSI feedback information, compressed information, compressed information of channel information, compressed information of CSI, compressed channel information, or compressed CSI, and the like.

[0163] The recovered channel information can also be referred to as CSI recovery information.

[0164] With the increasing size of the MIMO system antenna array, the number of supported antenna ports increases, and the dimensions of the corresponding channel matrix and precoding matrix grow. The error of approximating a large-scale channel matrix and precoding matrix with a limited number of predefined code words increases. One method to improve the accuracy of channel recovery is to increase the number of code words in the codebook, but this will also increase the overhead of CSI feedback (including one or more of the corresponding number of code words and weighting coefficients), thereby reducing the available resources for data transmission and causing a loss of system capacity.

[0165] The introduction of AI technology into wireless communication networks has produced a CSI feedback method based on AI models, namely AI-CSI feedback. The terminal device uses an AI model to compress and feedback the CSI, and the network device uses an AI model to decompress and recover the compressed CSI. In AI-based CSI feedback, a sequence (such as a bit sequence) is transmitted, and the overhead is lower than that of traditional CSI feedback. Moreover, AI models have stronger non-linear feature extraction capabilities, and can more effectively compress and represent channel information and more effectively recover the channel based on the feedback information compared to traditional schemes.

[0166] CSI feedback can be implemented based on an AE AI model. For example, the encoder in FIG. 3 can be a CSI generator, and the decoder can be a CSI reconstructor. For example, the encoder can be deployed in the terminal device, and the decoder can be deployed in the network device. The channel information V is generated into CSI feedback information z by the encoder. The channel information is reconstructed by the decoder, i.e., the recovered channel information V’ is obtained.

[0167] The channel information V can be obtained by measuring the channel information. For example, the channel information V can include a feature vector matrix of a downlink channel (a matrix composed of feature vectors). The encoder processes the feature vector matrix of the downlink channel to obtain the CSI feedback information z. In other words, the compression and / or quantization operation of the feature matrix according to the codebook in the related scheme is replaced by the operation of processing the feature matrix by the encoder to obtain the CSI feedback information z. The recovered channel information V’ can be obtained by processing the CSI feedback information z by the decoder.

[0168] The training process and inference process of the AI model in the embodiments of the present application are further exemplarily described below.

[0169] The training data for training the AI model includes training samples and sample labels. Exemplarily, the training samples are channel information measured by a terminal device, and the sample labels are real channel information, such as true CSI. For the case where the encoder and the decoder belong to the same autoencoder, the training data can only include the training samples, or in other words, the training samples are the sample labels.

[0170] In the field of wireless communication, the true CSI can be high-precision CSI.

[0171] The specific training process is as follows: the model training node processes the channel information, i.e., the training samples, using the encoder to obtain CSI feedback information, and processes the feedback information using the decoder to obtain recovered channel information, i.e., CSI recovery information. Then, the difference between the CSI recovery information and the corresponding sample label, i.e., the value of the loss function, is calculated, and the parameters of the encoder and the decoder are updated according to the value of the loss function, so that the difference between the recovered channel information and the corresponding sample label is minimized, i.e., the loss function is minimized. Exemplarily, the loss function can be mean square error (MSE) or cosine similarity. By repeating the above operation, the encoder and the decoder that meet the target requirements can be obtained. The above model training node can be a terminal device, a network device, or other network elements with AI functions in a communication system. CSI feedback information packet loss will affect the feedback performance, i.e., the accuracy of network-side channel information recovery. The following takes CSI feedback based on time-domain correlation as an example for illustration.

[0172] In a wireless communication link, the channel of a low-to-medium speed user changes continuously in time, and the feedback performance can be improved by exploiting the time-domain correlation of the channel. For example, by exploiting the time-domain correlation between historical channel measurement results and current channel measurement results to achieve channel information compression, it is beneficial to reduce the information loss in the compression process while reducing the overhead of feedback channel information. The terminal device and the network device can respectively use the time-domain correlation to compress and feedback channel information and to recover the channel information.

[0173] In order to improve the performance of the communication system, the CSI feedback can be implemented based on an artificial intelligence (AI) model. The AI model is composed of an encoder and a decoder, and the encoder and the decoder of the AI model are matched with each other. In the existing communication system, there is no protocol for the encoder and the decoder, i.e., if the two ends of the AI model are trained separately, the encoder and the decoder may not understand each other, which affects the CSI recovery performance of the decoder. Similar situations also exist for other double-end models. In view of this, the present application improves the performance of the AI model through the following embodiments.

[0174] FIG. 5 shows a schematic flowchart of a communication method. The communication method comprises but is not limited to the following steps:

[0175] S501: The second device sends first indication information to the first device.

[0176] The first indication information is used to indicate the first AI model attribute of the second device. The first AI model of the second device can be deployed on the terminal side or on the network side. The first AI model can be an encoder or a decoder. When the first AI model is deployed on the terminal side, the first AI model is an encoder. Correspondingly, when the first AI model is deployed on the network side, the first AI model is a decoder.

[0177] The AI model attribute includes the identification index of the first AI model, the monitoring criterion of the first AI model, the version number of the first AI model, the valid time of the first AI model, the model structure of the first AI model, the capability of the first AI model, etc. The identification index of the AI model can be a model ID or a sorting number, etc. The monitoring criterion of the AI model can be different indicators for monitoring: SGCS or MSE or NMSE, etc. for reconstructing the CSI and the true value CSI; or different threshold settings of the AI monitoring indicators, the threshold corresponding to the indicator is greater than a certain numerical threshold, and the purpose of setting the threshold is to identify the model with strong capability, the model with strong capability has high recovery accuracy, that is, the model greater than or equal to the threshold can be considered as the model with strong capability, such as the model with high CSI recovery accuracy. Optionally, the monitoring criterion of the first AI model can be replaced by the monitoring criterion of the first AI self-encoding model (i.e. the self-encoder) or the second AI model. Optionally, the first AI self-encoding model includes the first AI model and the second AI model. The version number of the AI model can be one or more of the following: software version number, hardware version number, model factory number, model associated manufacturer information number, etc. The valid time of the AI model can be directly indicated as the use period or the expiration date, the manner is not limited, as long as the explicit valid period of the model is clear. The structure of the AI model can be a specific model description, for example, at least one of the number of neural network layers, the width of neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the bias in the activation function. The capability of the AI model can be set to strong, medium, weak, etc. according to the recovery accuracy, or different quantity levels can be set according to the model parameters, for example: 1K order of magnitude, 1M order of magnitude, 10M order of magnitude, 100M order of magnitude, 1G order of magnitude, etc. The above attributes are all conducive to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0178] The first indication information can be carried and transmitted by signaling, and the first indication information can also be predefined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0179] The first device can be a terminal side or a network side. If the first device is a terminal side, the corresponding second device is a network side. If the first device is a network side, the corresponding second device is a terminal side.

[0180] FIG. 6 shows a schematic flowchart of a communication method. In this scenario, the first device is a network side device, and the second device is a terminal side device. The communication method includes but is not limited to the following steps:

[0181] S610: The terminal side sends terminal capability to the network side.

[0182] The terminal capability can include any one or more of the following: a capability of data that the terminal can support, a capability of data format that the terminal can support, a capability of model that the terminal can support, a capability of model structure that the terminal can support, and a capability of model file format that the terminal can support.

[0183] Specifically, the data can be a single data or a data set. The data format includes at least one of the following: an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter value, a dimension of input data, a parsing format of output data parameter value, a dimension of output data, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The model refers to an AI model, which can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning models. The model structure refers to an AI model structure, which can be a specific model description, such as at least one of the following: a number of neural network layers, a neural network width, a connection relationship between layers, a neuron weight, an activation function of a neuron, or a bias in the activation function. The model file format refers to an AI model file format, which can be.pb (Protocol Buffers),.ONNX (Open Neural Network Exchange), etc.

[0184] The terminal capability is carried and transmitted by signaling, and the terminal capability can also be predefined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0185] Optionally, the network side can select the AI model or the AI model attribute based on the terminal capability.

[0186] S620: The network side sends second indication information to the terminal side. The second indication information is used to indicate second AI model information. The second AI model is deployed at the network side. The second AI model can be a decoder or an encoder.

[0187] The second indication information can include any one or more of the following: an identification index of the first data, a version number of the first data, a format of the first data, a value of the first data and / or the first data, an identification index of the second AI model, a version number of the second AI model, a model structure of the second AI model, a parameter value of the second AI model and / or the second AI model, and a file format of the second AI model.

[0188] The identification index of the data can be a data storage ID or a number in order, etc. The version number of the data can be one or more of the following: a software version number, a hardware version number, a number when the data is manufactured, a manufacturer information number associated with the data, etc. The format of the data includes at least one of the following: an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter value, an input data dimension, a parsing format of output data parameter value, an output data dimension, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The data can be a single data or a data set. The above parameters are beneficial to realize AI model adjustment or optimize data value, thereby improving the performance of the AI model.

[0189] The identification index of the AI model can be a model ID or a sequence number thereof. The version number of the AI model can be one or more of a software version number, a hardware version number, a number when the model is manufactured, manufacturer information associated with the model, and the like. The model structure of the AI model can be a specific model description, such as at least one of a number of neural network layers, a neural network width, a connection relationship between layers, a neuron weight, a neuron activation function, or a bias in the activation function. The AI model parameters obtained through machine learning training can be parameters such as a number of neural network layers, a width, a neuron weight, or a neuron activation function. The file format of the AI model can be.ckpt, SavedModel,.pt or.pth (PyTorch),.onnx (ONNX (Open Neural Network Exchange)),.json,.pb (Protocol Buffers),.h5 or.hdf5, and the like. The AI model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q-learning model, or other machine learning models. The above parameters are beneficial to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0190] The second indication information can be carried and transmitted by signaling, or can be predefined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0191] Optionally, the second indication information involves multiple times, and specific information can be carried and transmitted by multiple signaling.

[0192] Correspondingly, the terminal side receives the second indication information sent by the network side.

[0193] S630: The terminal side trains the first AI model according to the second indication information, and obtains the first AI model and its attributes.

[0194] When the second indication information is the identification index of the first data, the terminal side obtains the data according to the identification index of the first data according to a protocol, and trains the first AI model and its attributes using the data. The data used for training the first AI model and its attributes is obtained by training the first AI model and its attributes according to the data. The data can be data required by the terminal side for training, or data required by the network side for training.

[0195] When the second indication information is a version number of the first data, the terminal side acquires data according to the version number of the first data, and trains the first AI model and its attributes according to the data. The data can be data required by the terminal side for training, or data required by the network side for training.

[0196] When the second indication information is a format of the first data, the first data, and / or a value of the first data, the terminal side first parses the data according to the data format, acquires data required by the terminal side for training, and then trains the first AI model and its attributes.

[0197] When the second indication information is an identification index of the second AI model, the terminal side acquires a specific AI model according to the index. The AI model can be a model required by the terminal side, or a network side model. The terminal side can train the model, or can not train the model.

[0198] When the second indication information is a version number of the second AI model, the terminal side acquires a specific AI model according to the version number. The AI model can be a model required by the terminal side, or a network side model. The terminal side can train the model, or can not train the model.

[0199] When the second indication information is a model structure of the second AI model, the second AI model, and / or a parameter value of the second AI model, the terminal parses the second AI model according to the second AI model structure, and then trains the first AI model according to the second AI model.

[0200] When the second indication information is the second AI model and / or a parameter value of the second AI model, and a file format of the second AI model, the terminal parses the second AI model according to the file format of the second AI model, and then trains the first AI model according to the second AI model.

[0201] Optionally, for the case that the AI model can be directly parsed, the model can be retrained, or the model can be directly used. At this time, the obtained first AI model can be the same as the second AI model deployed on the network side, or can not be the same.

[0202] Optionally, when the second indication information includes multiple signals, the terminal side can receive one or more signals.

[0203] Optionally, when the second indication information includes multiple signals, the terminal side receives the second indication information, and can use all or part of the signals for training the first AI model. All or part of the signals can be used to train the first AI model, that is, all or part of the signals can be used to train the first AI model.

[0204] S640: The terminal side sends first indication information to the network side. The first indication information is used to indicate the first AI model attribute, the first AI model is deployed on the terminal side, and the first AI model can be an encoder.

[0205] The AI model attribute includes an identification index of the first AI model, a monitoring criterion of the first AI model, a version number of the first AI model, a valid time of the first AI model, a model structure of the first AI model, and a capability of the first AI model. The identification index of the AI model can be a model ID or a sequence number thereof. The monitoring criterion of the AI model can be different indicators for monitoring, such as SGCS or MSE or NMSE of reconstructed CSI and true value CSI, or different threshold settings of AI monitoring indicators, which correspond to a threshold greater than a certain numerical threshold. The purpose of setting the threshold is to identify a model with strong capability, that is, a model with high recovery accuracy, that is, a model greater than or equal to the threshold can be considered as a model with strong capability, such as a model with high CSI recovery accuracy. Optionally, the monitoring criterion of the first AI model can be replaced by the monitoring criterion of the first AI auto-encoding model (i.e., the auto-encoder) or the second AI model. Optionally, the first AI auto-encoding model includes the first AI model and the second AI model. The version number of the AI model can be one or more of the following: a software version number, a hardware version number, a model number at the time of manufacture, and a manufacturer information number associated with the model. The AI model valid time can be directly indicated as the use-by date or the expiration date, and the manner is not limited as long as the explicit validity period of the model is clear. The structure of the AI model can be a specific model description, such as at least one of the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the bias in the activation function. The capability of the AI model can be set to strong, medium, and weak levels according to the recovery accuracy, or different levels of parameters can be set, such as 1K, 1M, 10M, 100M, and 1G. The above attributes are beneficial to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0206] The first indication information can be carried and transmitted by signaling, and the first indication information can be pre-defined by a protocol. The signaling can be a physical layer signaling or a high layer signaling, which is not limited in the present application.

[0207] FIG. 7 shows a schematic flowchart of a communication method. In this scenario, the first device is a terminal side device, and the second device is a network side device. The communication method includes but is not limited to the following steps:

[0208] S710: The terminal side sends fourth indication information to the network side. The fourth indication information is used to indicate fourth AI model information, the fourth AI model is deployed on the terminal side, and the fourth AI model can be an encoder or a decoder.

[0209] The fourth indication information is used to indicate any one or more of the following: an identification index of the second data, a version number of the second data, a format of the second data, a value of the second data and / or the second data, an identification index of the fourth AI model, a version number of the fourth AI model, a model structure of the fourth AI model, a parameter value of the fourth AI model and / or the fourth AI model, and a file format of the fourth AI model.

[0210] The identification index of the data can be a data storage ID or a number in order. The version number of the data can be one or more of the following: a software version number, a hardware version number, a number when the data is manufactured, and a manufacturer information number associated with the data. The format of the data includes at least one of the following: an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter values, an input data dimension, a parsing format of output data parameter values, an output data dimension, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The data can be a single data or a data set. The above parameters are beneficial to adjust or optimize the data value of the AI model, thereby improving the performance of the AI model.

[0211] The identification index of the AI model can be a model ID or a sequence number thereof, etc. The version number of the AI model can be one or more of the following: a software version number, a hardware version number, a number when the model is manufactured, a manufacturer information number associated with the model, etc. The model structure of the AI model can be a specific model description, such as at least one of a number of neural network layers, a neural network width, a connection relationship between layers, a neuron weight, a neuron activation function, or a bias in the activation function. The AI model parameters obtained through machine learning training can be parameters such as a number of neural network layers, a width, a neuron weight, or a neuron activation function. The file format of the AI model can be.ckpt, SavedModel,.pt or.pth (PyTorch),.onnx (ONNX (Open Neural Network Exchange)),.json,.pb (Protocol Buffers),.h5 or.hdf5, etc. The AI model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q-learning model, or other machine learning models. The above parameters are beneficial to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0212] The fourth indication information can be carried and transmitted by signaling, or can be predefined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0213] Optionally, the fourth indication information involves multiple times, and specific information can be carried and transmitted by multiple signaling.

[0214] Correspondingly, the network side receives the fourth indication information sent by the terminal side.

[0215] S720: The network side trains the third AI model using the fourth indication information to obtain the third AI model and its attributes. That is, the third AI model is trained according to the fourth indication information to obtain the third AI model and its attributes.

[0216] When the fourth indication information is the identification index of the second data, the network side obtains the data according to the protocol according to the identification index of the second data, and trains the third AI model and its attributes using the data. That is, the third AI model and its attributes are obtained by training according to the data. The data can be data required by the network side for training, or data required by the terminal side for training.

[0217] When the fourth indication information is a version number of the second data, the network side obtains data according to the version number of the second data, and obtains the third AI model and its attribute by training the data. The data can be data required by the network side for training, or data required by the terminal side for training. When the fourth indication information is a format of the second data, the second data, and / or a value of the second data, the network side first parses the data according to the data format, reconstructs / retrieves the data of the terminal side, and then trains the data to obtain the third AI model and its attribute.

[0218] When the fourth indication information is an identification index of the fourth AI model, the network side obtains a specific AI model according to the index. The AI model can be a model required by the terminal side, or a network side model. The network side can train the model, or can not train the model.

[0219] When the fourth indication information is a version number of the fourth AI model, the network side obtains a specific AI model according to the version number. The AI model can be a model required by the terminal side, or a network side model. The terminal side can train the model, or can not train the model. When the fourth indication information is a model structure of the fourth AI model, the fourth AI model, and / or a parameter value of the fourth AI model, the network side parses the fourth AI model according to the fourth AI model structure, and then trains the fourth AI model to obtain the third AI model.

[0220] When the fourth indication information is the fourth AI model and / or a parameter value of the fourth AI model, or a file format of the fourth AI model, the network side parses the fourth AI model according to the file format of the fourth AI model, and then trains the fourth AI model to obtain the third AI model.

[0221] Optionally, for a case that the AI model can be directly parsed, the model can be retrained, or the model can be directly used. At this time, the obtained third AI model can be the same as the fourth AI model deployed on the terminal side, or can not be the same.

[0222] Optionally, when the fourth indication information includes multiple signals, the network side can receive one or more signals.

[0223] Optionally, when the fourth indication information includes multiple signals, the network side receives the fourth indication information, and can use all or part of the signal information for training the third AI model. All or part of the signal information can be used to train the third AI model, that is, all or part of the signal information can be used to train the third AI model.

[0224] S730: The network side sends the third indication information to the terminal side.

[0225] The third indication information is used for indicating a third AI model attribute, the third AI model is deployed at a network side, and the third AI model can be a decoder.

[0226] The AI model attribute includes an identification index of the third AI model, a monitoring criterion of the third AI model, a version number of the third AI model, a valid time of the third AI model, a model structure of the third AI model, and a capability of the third AI model. The identification index of the AI model can be a model ID or a sequence number, etc. The monitoring criterion of the AI model can be different indicators for monitoring, such as SGCS or MSE or NMSE of reconstructed CSI and true value CSI, or different threshold settings of AI monitoring indicators, the threshold corresponding to the indicator is greater than a certain numerical threshold, and the purpose of setting the threshold is to identify a model with strong capability, the model with strong capability has high recovery accuracy, that is, a model greater than or equal to the threshold can be considered as a model with strong capability, such as a model with high CSI recovery accuracy. Optionally, the monitoring criterion of the first AI model can be replaced by the monitoring criterion of the first AI auto-encoding model (i.e., an auto-encoder) or the second AI model. Optionally, the first AI auto-encoding model includes the first AI model and the second AI model. The version number of the AI model can be one or more of the following: a software version number, a hardware version number, a model factory number, a model associated manufacturer information number, etc. The AI model valid time can be directly indicated as a use period or an expiration date, the manner is not limited, as long as the explicit validity period of the model is clear. The structure of the AI model can be a specific model description, such as at least one of the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the bias in the activation function. The capability of the AI model can be set to strong, medium, and weak levels according to the recovery accuracy, or different quantity levels can be set according to the model parameters, such as 1K order, 1M order, 10M order, 100M order, 1G order, etc. The above attributes are all conducive to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0227] The third indication information can be carried and transmitted by signaling, and the third indication information can also be pre-defined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited by the present application. FIG. 8 shows a schematic flowchart of a communication method. In this scenario, the first device is a terminal side device, and the second device is a network side device. The communication method includes but is not limited to the following steps:

[0228] S810: The terminal side sends a terminal capability to the network side.

[0229] The terminal capability can include any one or more of the following: a capability of data that the terminal can support, a capability of a data format that the terminal can support, a capability of a model that the terminal can support, a capability of a model structure that the terminal can support, and a capability of a model file format that the terminal can support.

[0230] Specifically, the data can be single data or a data set. The data format includes at least one of the following: an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter values, a dimension of input data, a parsing format of output data parameter values, a dimension of output data, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The model herein refers to an AI model, which can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning models. The model structure herein refers to an AI model structure, which can be a specific model description, such as at least one of the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the bias in the activation function. The model file format herein refers to an AI model file format, which can be.pb (Protocol Buffers),.ONNX (Open Neural Network Exchange), or the like.

[0231] The terminal capability is carried and transmitted through signaling, and the terminal capability can also be predefined by a protocol. The signaling herein can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0232] Correspondingly, the network side receives the terminal capability sent by the terminal side. The terminal side sends fourth indication information to the network side in S820. The fourth indication information is used to indicate fourth AI model information, and the fourth AI model is deployed at the terminal side. The fourth AI model can be an encoder.

[0233] The fourth indication information is used to indicate any one or more of the following: an identification index of the second data, a version number of the second data, a format of the second data, the second data and / or a value of the second data, an identification index of the fourth AI model, a version number of the fourth AI model, a model structure of the fourth AI model, the fourth AI model and / or a parameter value of the fourth AI model, and a file format of the fourth AI model.

[0234] The identification index of the data can be a data storage ID, or a number thereof in order. The version number of the data can be one or more of a software version number, a hardware version number, a number of the data when it is manufactured, a manufacturer information number associated with the data, and the like. The format of the data includes at least one of an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter values, a dimension of input data, a parsing format of output data parameter values, a dimension of output data, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The data can be a single data or a data set. The above parameters are beneficial to realize AI model adjustment or optimization of data values, thereby improving the performance of the AI model.

[0235] The identification index of the AI model can be a model ID, or a number thereof in order. The version number of the AI model can be one or more of a software version number, a hardware version number, a number of the model when it is manufactured, a manufacturer information number associated with the model, and the like. The model structure of the AI model can be a specific model description, for example, at least one of a number of neural network layers, a width of the neural network, a connection relationship between layers, a weight of a neuron, an activation function of the neuron, or a bias in the activation function. The AI model parameter obtained through machine learning training can be a parameter in a number of neural network layers, a width, a weight of a neuron, or an activation function of the neuron, and the like. The file format of the AI model can be.ckpt, SavedModel,.pt or.pth (PyTorch),.onnx (ONNX (Open Neural Network Exchange)),.json,.pb (Protocol Buffers),.h5 or.hdf5, and the like. The AI model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning models. The above parameters are beneficial to realize AI model adjustment or optimization, thereby improving the performance of the AI model.

[0236] The fourth indication information can be carried and transmitted through signaling, or the fourth indication information can be pre-defined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0237] Optionally, the fourth indication information involves multiple times, and specific information can be carried and transmitted through multiple signaling.

[0238] Correspondingly, the network side receives the fourth indication information sent by the terminal side. S830: The network side trains a third AI model according to the received terminal capability and / or fourth indication information, and obtains the third AI model and its attributes.

[0239] When the terminal capability is a data supported by the terminal, the data here can be a single data or a data set, and the network side trains a third AI model and its attributes according to the data. The data can be data required by the network side for training, or data required by the terminal side for training.

[0240] When the terminal capability is a data format supported by the terminal, the data format here includes at least one of the following: an arrangement manner of input data, an arrangement manner of output data, a parsing format of input data parameter values, an input data dimension, a parsing format of output data parameter values, an output data dimension, a software update version corresponding to the first data, a hardware update version corresponding to the first data, a valid time period corresponding to the first data, an effective time corresponding to the first data, or an invalid time corresponding to the first data. The network side first parses the data according to the data format, reconstructs / resumes the data of the terminal side, and then trains a third AI model and its attributes according to the reconstructed / resumed data.

[0241] When the terminal capability is a model supported by the terminal, the model here refers to an AI model, which can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model, or other machine learning models. The AI model can be a model of the terminal side or a model of the network side. The network side can train the model or can not train the model.

[0242] When the terminal capability is a model structure supported by the terminal, the model structure here refers to an AI model structure, which can be a specific model description, such as at least one of the following: a number of neural network layers, a neural network width, a connection relationship between layers, a neuron weight, an activation function of a neuron, or a bias in the activation function. The network side parses an AI model according to the AI model structure, and then trains a third AI model according to the parsed AI model.

[0243] The terminal capability is a model file format supported by the terminal, where the model file format refers to an AI model file format, which can be.ckpt, SavedModel,.pt or.pth (PyTorch),.onnx (ONNX (Open Neural Network Exchange)),.json,.pb (Protocol Buffers),.h5 or.hdf5, etc. The network side parses the AI model according to the file format of the AI model, and then trains the parsed AI model to obtain a third AI model.

[0244] When the fourth indication information is an identification index of the second data, the network side acquires data according to the identification index of the second data according to a protocol, and trains the data to obtain a third AI model and its attributes. The data can be data required by the network side for training, or data required by the terminal side for training.

[0245] When the fourth indication information is a version number of the second data, the network side acquires data according to the version number of the second data, and trains the data to obtain a third AI model and its attributes. The data can be data required by the network side for training, or data required by the terminal side for training.

[0246] When the fourth indication information is a format of the second data, the second data, and / or a value of the second data, the network side first parses the data according to the data format, reconstructs / restores the data of the terminal side, and then trains the data to obtain a third AI model and its attributes.

[0247] When the fourth indication information is an identification index of the fourth AI model, the network side acquires a specific AI model according to the index. The AI model can be a model required by the terminal side, or a network side model. The network side can train the model, or can not train the model.

[0248] When the fourth indication information is a version number of the fourth AI model, the network side acquires a specific AI model according to the version number. The AI model can be a model required by the terminal side, or a network side model. The terminal side can train the model, or can not train the model.

[0249] When the fourth indication information is a model structure of the fourth AI model, the fourth AI model, and / or a parameter value of the fourth AI model, the network side parses the fourth AI model according to the fourth AI model structure, and then trains the fourth AI model to obtain a third AI model.

[0250] The fourth indication information is the fourth AI model and / or parameter value of the fourth AI model, file format of the fourth AI model, the network side parses the fourth AI model according to the file format of the fourth AI model, and then trains the fourth AI model to obtain the third AI model.

[0251] Optionally, for the case that the AI model can be directly parsed, the model can be retrained, or the model can be directly used. At this time, the obtained third AI model can be the same as or different from the fourth AI model deployed on the terminal side.

[0252] Optionally, when the fourth indication information contains multiple signaling, the network side can receive one or more pieces of signaling information.

[0253] Optionally, when the fourth indication information contains multiple signaling, the network side receives the fourth indication information, and can use all or part of the signaling information for training the third AI model. Wherein, all or part of the signaling information can be used to train the third AI model, that is, the third AI model can be trained according to all or part of the signaling information.

[0254] Optionally, the network side can train the third AI model according to all or part of the information of the terminal capability.

[0255] Optionally, the network side can train the third AI model according to part or all of the information of the terminal capability and the fourth indication information.

[0256] S840: The network side sends third indication information to the terminal side. The third indication information is used to indicate the attribute of the third AI model, the third AI model is deployed on the network side, and the third AI model can be a decoder.

[0257] The AI model attributes include an identification index of the third AI model, a monitoring criterion of the third AI model, a version number of the third AI model, a valid time of the third AI model, a model structure of the third AI model, and a capability of the third AI model. The identification index of the AI model can be a model ID or a sequence number, etc. The monitoring criterion of the AI model can be different indicators for monitoring, such as an SGCS or an MSE or an NMSE of reconstructed CSI and true value CSI, or different threshold settings of an AI monitoring indicator, where the threshold corresponding to the indicator is greater than a certain numerical threshold, and the purpose of setting the threshold is to identify a model with strong capability, i.e., a model with high recovery accuracy, that is, a model greater than or equal to the threshold can be considered as a model with strong capability, such as a model with high CSI recovery accuracy. Optionally, the monitoring criterion of the first AI model can be replaced by the monitoring criterion of the first AI auto-encoding model (i.e., an auto-encoder) or the second AI model. Optionally, the first AI auto-encoding model includes the first AI model and the second AI model. The version number of the AI model can be one or more of the following: a software version number, a hardware version number, a model number at the time of manufacture, a manufacturer information number associated with the model, etc. The AI model valid time can be directly indicated as a usage period or an expiration date, and the manner is not limited as long as the explicit validity period of the model is clear. The structure of the AI model can be a specific model description, such as at least one of the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of neurons, the activation function of neurons, or the bias in the activation function. The capability of the AI model can be set to strong, medium, and weak levels according to the recovery accuracy, or different quantity levels can be set according to the model parameters, such as 1K, 1M, 10M, 100M, 1G, etc. The above attributes are beneficial to the adjustment or optimization of the AI model, thereby improving the performance of the AI model.

[0258] The third indication information can be carried and transmitted by signaling, and the third indication information can also be pre-defined by a protocol. The signaling can be physical layer signaling or high layer signaling, which is not limited in the present application.

[0259] It can also be understood that some optional features in the embodiments of the present application can not depend on other features in some scenarios, or can be combined with other features in some scenarios, which are not limited.

[0260] It can also be understood that the schemes in the embodiments of the present application can be reasonably combined, and the explanations or descriptions of various terms appearing in the embodiments can be mutually referenced or explained in various embodiments, which are not limited.

[0261] It can also be understood that the sizes of various digital serial numbers in the embodiments of the present application do not mean the order of execution, but are only distinguished for the convenience of description, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0262] It can also be understood that the methods and operations implemented by the device in the various method embodiments described above can also be implemented by the constituent components of the device, such as chips or circuits.

[0263] Corresponding to the methods given in the above method embodiments, the embodiments of the present application also provide corresponding devices, which include modules for executing the corresponding modules of the above various method embodiments. The module can be software, hardware, or a combination of software and hardware. It can be understood that the technical features described in the above method embodiments are also applicable to the following device embodiments.

[0264] FIG. 9 is a schematic diagram of a communication device 900 provided by an embodiment of the present application. The device 900 includes a transceiver unit 910 and a processing unit 920. The transceiver unit 910 can be used to implement the corresponding communication function. The transceiver unit 910 can also be referred to as a communication interface or a communication unit, etc. The processing unit 920 can be used to implement the corresponding processing function, such as configuring resources.

[0265] Optionally, the device 900 further includes a storage unit, which can be used to store instructions and / or data. The processing unit 920 can read the instructions and / or data in the storage unit, so that the device implements the actions of the device or network element in the above various method embodiments.

[0266] The device 900 can be a second device, or can be applied to the second device or matched with the second device for use, and can implement the communication device of the communication method executed by the second device; or the device 900 can be a first device, or can be applied to the first device or matched with the first device for use, and can implement the communication device of the communication method executed by the first device.

[0267] When the device 900 is applied to the second device, the device 900 can implement the steps or processes corresponding to the execution of the second device in the above method embodiments. Among them, the transceiver unit 910 can be used to execute the transceiver-related operations of the second device in the above method embodiments, and the processing unit 920 can be used to execute the processing-related operations of the second device in the above method embodiments.

[0268] When the device 900 is applied to the first device, the device 900 can implement the steps or processes corresponding to the execution of the first device in the above method embodiments, wherein the transceiver unit 910 can be used to execute the transceiver-related operations of the first device in the above method embodiments, and the processing unit 920 can be used to execute the processing-related operations of the first device in the above method embodiments.

[0269] It should be understood that the specific process of each unit performing the corresponding steps described above has been described in detail in the above method embodiments, and for the sake of brevity, will not be repeated here.

[0270] It should also be understood that the apparatus 900 herein is embodied in the form of functional units. The term "unit" herein can refer to an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a dedicated processor, or a group processor, etc.) and a memory for executing one or more software or firmware programs, a combination of logic circuitry and / or other suitable components supporting the described functions. In an optional example, those skilled in the art can understand that the apparatus 900 can be embodied as the first device in the above embodiments, and can be used to perform the processes and / or steps corresponding to the first device in the above method embodiments; or the apparatus 900 can be embodied as the second device in the above embodiments, and can be used to perform the processes and / or steps corresponding to the second device in the above method embodiments, and for the sake of brevity, will not be repeated here.

[0271] The apparatus 900 of each of the above schemes has the function of implementing the corresponding steps performed by the device (such as the first device, or the second device) in the above method. The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions; for example, the transceiver unit can be replaced by a transceiver (for example, the transmitting unit in the transceiver unit can be replaced by a transmitter, and the receiving unit in the transceiver unit can be replaced by a receiver), and other units, such as the processing unit, can be replaced by a processor, which respectively performs the transceiving operations and related processing operations in each method embodiment.

[0272] In addition, the transceiver unit 910 described above can also be a transceiver circuit (for example, it can include a receiving circuit and a transmitting circuit), and the processing unit 920 can be a processing circuit. The processing circuit can include one or more processors, or a circuit for processing functions in one or more processors, etc.

[0273] It should be noted that the apparatus in FIG. 9 can be a network element or device in the above embodiments, or a chip or chip system, such as a system on chip (SoC). Among them, the transceiver unit can be an input / output circuit, a communication interface; the processing unit is a processor or microprocessor integrated on the chip or an integrated circuit. Not limited here.

[0274] Figure 10 is a schematic diagram of another apparatus 1000 for communication. The apparatus 1000 includes a processor 1010 configured to execute instructions stored in a memory 1020, or to read data / signaling stored in the memory 1020, to perform the methods in the above method embodiments. The processor 1010 can be one or more.

[0275] Optionally, the apparatus 1000 further includes the memory 1020 configured to store the computer programs or instructions and / or data. The memory 1020 can be integrated in the processor 1010, or can be separate from the processor 1010. The memory 1020 can be one or more.

[0276] Optionally, the apparatus 1000 further includes a transceiver circuit 1030 configured to receive and / or transmit signals. For example, the processor 1010 is configured to control the transceiver circuit 1030 to receive and / or transmit signals. The processor 1010 can also be replaced by a processing circuit.

[0277] The apparatus 1000 can be a network element or a device in the above embodiments, or can be a chip or a chip system. When the apparatus 1000 is a network element or a device in the above embodiments, the transceiver circuit 1030 can be a transceiver. When the apparatus 1000 is a chip or a chip system, the transceiver circuit 1030 can be an interface circuit or an input / output interface.

[0278] As an option, the apparatus 1000 can be applied to the second device, and specifically the apparatus 1000 can be the second device, or can be an apparatus capable of supporting the second device and implementing the functions of the second device in any of the above examples. The apparatus 1000 is configured to implement the operations performed by the second device in the above method embodiments.

[0279] For example, the processor 1010 is configured to execute the computer programs or instructions stored in the memory 1020 to implement the related operations of the second device in the above method embodiments.

[0280] As another option, the apparatus 1000 can be applied to the first device, and specifically the apparatus 1000 can be the first device, or can be an apparatus capable of supporting the first device and implementing the functions of the first device in any of the above examples. The apparatus 1000 is configured to implement the operations performed by the first device in the above method embodiments.

[0281] For example, the processor 1010 is configured to execute the computer programs or instructions stored in the memory 1020 to implement the related operations of the first device in the above method embodiments.

[0282] It should be appreciated that a processor referenced in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), ASICs, field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0283] It should also be understood that the memory referenced in the embodiments of the present application can be a volatile memory and / or a non-volatile memory. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically EPROM (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM). For example, the RAM can be used as an external cache. As an example but not limitation, the RAM includes the following various forms: static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM) and direct memory bus random access memory (DRAM) (DRAM).

[0284] It should be noted that when the processor is a general-purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, a discrete gate or transistor logic component, a discrete hardware component, the memory (storage module) can be integrated in the processor.

[0285] It should also be noted that the memory described herein is intended to include, but not limited to, these and any other suitable types of memory.

[0286] The embodiment of the present application further provides a computer readable storage medium, which stores computer instructions for implementing the method executed by the communication device in each method embodiment.

[0287] For example, the computer program is executed by a computer, so that the computer can implement the method executed by the first device in each method embodiment.

[0288] For another example, the computer program is executed by a computer, so that the computer can implement the method executed by the second device in each method embodiment.

[0289] The embodiment of the present application further provides a computer program product, which contains instructions, and the instructions are executed by a computer to implement the method executed by the device (such as the first device, or the second device) in each method embodiment.

[0290] The embodiment of the present application further provides a communication system, which comprises the first device and the second device.

[0291] Optionally, the system further comprises a device in communication with the first device and / or the second device.

[0292] The explanation and beneficial effects of the related content in any of the above provided devices can refer to the corresponding method embodiments provided above, and will not be repeated here.

[0293] In several embodiments provided by the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the device embodiments described above are only schematic, and the division of the units is only a logical function division, and there can be another division way in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0294] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. For example, the computer can be a personal computer, a server, a network device, or the like. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD), etc. For example, the foregoing available media includes but is not limited to: a variety of media that can store program codes such as a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk, etc.

[0295] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A communication method characterized by comprising: Applied to the terminal side, the method comprises: sending first indication information to the network side, the first indication information being used to indicate the attribute of the first AI model; the first AI model being deployed on the terminal side; the attribute of the first AI model comprises any one or more of the following: the identification index of the first AI model; the monitoring criterion of the first AI model; the version number of the first AI model; the valid time of the first AI model; the model structure of the first AI model; the capability of the first AI model.

2. The method of claim 1, wherein, The method further comprises: receiving second indication information from the network side, the second indication information being used to indicate any one or more of the following: the identification index of the first data; the version number of the first data; the format of the first data; the value of the first data and / or the first data; the identification index of the second AI model; the version number of the second AI model; the model structure of the second AI model; the parameter value of the second AI model and / or the second AI model; the file format of the second AI model.

3. The method of claim 2, wherein, The method further comprises: the terminal side sending terminal capability to the network side, the terminal capability being used to indicate any one or more of the following: the capability of the data that the terminal can support; the capability of the data format that the terminal can support; the capability of the model that the terminal can support; the capability of the model structure that the terminal can support; the capability of the model file format that the terminal can support.

4. The method of claim 3, wherein, The method further comprises: receiving the second indication information from the network side and training the first AI model according to the second indication information.

5. A communication method characterized by comprising: Applied to the network side, the method comprises: receiving first indication information sent by the terminal side, the first indication information being used to indicate the attribute of the first AI model; the first AI model being deployed on the terminal side; the attribute of the first AI model comprises any one or more of the following: the identification index of the first AI model; the monitoring criterion of the first AI model; the version number of the first AI model; the valid time of the first AI model; the model structure of the first AI model; the capability of the first AI model.

6. The method of claim 5, wherein, The method further comprises: sending second indication information to the terminal side, the second indication information being used to indicate any one or more of the following: the identification index of the first data; the version number of the first data; the format of the first data; the value of the first data and / or the first data; the identification index of the second AI model; the version number of the second AI model; the model structure of the second AI model; the parameter value of the second AI model and / or the second AI model; the file format of the second AI model.

7. The method of claim 6, wherein, The method further comprises: the network side receiving terminal capability sent by the terminal side, the terminal capability being used to indicate any one or more of the following: the capability of the data that the terminal can support; the capability of the data format that the terminal can support; the capability of the model that the terminal can support; the capability of the model structure that the terminal can support; the capability of the model file format that the terminal can support.

8. The method of claim 7, wherein, The method further comprises: the network side sending second indication information to the terminal side according to the terminal capability.

9. A communication method characterized by comprising: Applied to the network side, the method comprises: sending third indication information to a terminal side, the third indication information being used to indicate attributes of a third AI model; the third AI model being deployed at a network side; the attributes of the third AI model include any one or more of the following: an identification index of the third AI model; a monitoring criterion of the third AI model; a version number of the third AI model; a valid time of the third AI model; a model structure of the third AI model; a capability of the third AI model.

10. The method of claim 9, wherein, The method further includes: receiving fourth indication information from the terminal side, the fourth indication information being used to indicate any one or more of the following: an identification index of the second data; a version number of the second data; a format of the second data; a value of the second data and / or the second data; an identification index of the fourth AI model; a version number of the fourth AI model; a model structure of the fourth AI model; a parameter value of the fourth AI model and / or the fourth AI model; a file format of the fourth AI model.

11. The method of claim 10, wherein, The method further includes: the terminal side sending terminal capability to the network side, the terminal capability being used to indicate any one or more of the following: a capability of data that the terminal can support; a capability of a data format that the terminal can support; a capability of a model that the terminal can support; a capability of a model structure that the terminal can support; a capability of a model file format that the terminal can support.

12. The method of claim 10, wherein, The method further includes: the network side determining the third AI model according to the fourth indication information sent by the terminal side.

13. The method of claim 11, wherein, The method further includes: the network side determining the third AI model according to the fourth indication information sent by the terminal side and / or the terminal capability sent by the terminal side.

14. A communication method, comprising: Applied to a terminal side, the method includes: sending third indication information to a network side, the third indication information being used to indicate attributes of a third AI model; the third AI model being deployed at a network side; the attributes of the third AI model include any one or more of the following: an identification index of the third AI model; a monitoring criterion of the third AI model; a version number of the third AI model; a valid time of the third AI model; a model structure of the third AI model; a capability of the third AI model.

15. The method of claim 14, wherein, The method further includes: the terminal side sending fourth indication information, the fourth indication information being used to indicate any one or more of the following: an identification index of the second data; a version number of the second data; a format of the second data; a value of the second data and / or the second data; an identification index of the fourth AI model; a version number of the fourth AI model; a model structure of the fourth AI model; a parameter value of the fourth AI model and / or the fourth AI model; a file format of the fourth AI model.

16. The method of claim 15, wherein, The method further includes: the terminal side sending terminal capability to the network side, the terminal capability being used to indicate any one or more of the following: a capability of data that the terminal can support; a capability of a data format that the terminal can support; a capability of a model that the terminal can support; a capability of a model structure that the terminal can support; a capability of a model file format that the terminal can support.

17. The method of claim 15, wherein, The fourth indication information is used to determine the third AI model.

18. The method of claim 16, wherein, The fourth indication information and / or the terminal capability sent by the terminal side are used to determine the third AI model.

19. A computer-readable storage medium, characterized in that, The computer readable storage medium is for storing instructions which, when executed by a processor, cause the method of any one of claims 1 to 18 to be implemented.

20. A communications device, characterized by The communication device comprises a processor coupled with a storage medium storing instructions which, when executed by the processor, cause the communication device to perform the method of any one of claims 1 to 18.

21. A computer program, characterized in that, An article of manufacture comprising instructions which, when executed by a processor, cause the method of any one of claims 1 to 18 to be implemented.

22. A communication system, characterized by An article of manufacture comprising means for implementing the method of any one of claims 1 to 4 and means for implementing the method of any one of claims 5 to 8, or, an article of manufacture comprising means for implementing the method of any one of claims 9 to 13 and means for implementing the method of any one of claims 14 to 18.

Images