Method, device and apparatus for updating wireless channel model, and storage medium

Abstract

The application discloses a wireless channel model updating method and device, equipment and a storage medium, and relates to the field of communication. The method comprises the following steps: obtaining first channel information through a first wireless channel model, the first wireless channel model being a generator (410) in a first generative adversarial network; and in the case that the first wireless channel model satisfies an updating condition based on the first channel information through a first discriminator (420), which is a discriminator in the first generative adversarial network, the first wireless channel model is updated. The method can perceive the deterioration of a neural network model in real time and update the neural network model in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication, and in particular to a wireless channel model updating method and device, equipment and a storage medium. BACKGROUND

[0002] In the application process of a neural network model, the actual business data will change to varying degrees over time, and the neural network model cannot be applied to new business data. Therefore, the neural network model needs to be updated in a timely manner.

[0003] In the related art, in the application process of a neural network model, a computer device sends data output by the neural network model to a business module to perform related businesses, and the business module indirectly feeds back the performance of the neural network model according to business performance indicators generated in the process of performing the business, and updates the neural network model when the performance of the neural network model is fed back as being poor.

[0004] The above method can only indirectly judge the reliability of the neural network model through the business performance indicators fed back by the business module, and cannot perceive the degradation of the neural network model in real time. SUMMARY

[0005] Embodiments of the present application provide a wireless channel model updating method, device, equipment and storage medium, which can perceive the degradation of the neural network model in real time and update the neural network model in a timely manner. The technical solution is as follows.

[0006] According to an aspect of the present application, a wireless channel model updating method is provided, which is performed by a first communication device, and the method comprises:

[0007] obtaining first channel information through a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network;

[0008] in a case where it is determined by a first discriminator based on the first channel information that the first wireless channel model meets an updating condition, triggering updating of the first wireless channel model, the first discriminator being a discriminator in the first generative adversarial network.

[0009] According to an aspect of the present application, a wireless channel model updating method is provided, which is performed by a second communication device, and the method comprises:

[0010] In a case where the first wireless channel model meets an updating condition determined by a first discriminator based on first channel information output by the first wireless channel model, the first wireless channel model is updated, the first wireless channel model being a generator in a first generative adversarial network, and the first discriminator being a discriminator in the first generative adversarial network.

[0011] According to an aspect of the present application, a first communication device is provided, the device comprising:

[0012] a first model module configured to obtain first channel information by a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network;

[0013] a first updating module configured to trigger updating of the first wireless channel model in a case where the first wireless channel model meets an updating condition determined by a first discriminator based on the first channel information, the first discriminator being a discriminator in the first generative adversarial network.

[0014] According to an aspect of the present application, a second communication device is provided, the device comprising:

[0015] a second updating module configured to update the first wireless channel model in a case where the first wireless channel model meets an updating condition determined by a first discriminator based on first channel information output by the first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network, and the first discriminator being a discriminator in the first generative adversarial network.

[0016] According to an aspect of the present application, a first communication device is provided, the device comprising: a processor; wherein

[0017] the processor is configured to obtain first channel information by a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network;

[0018] the processor is configured to trigger updating of the first wireless channel model in a case where the first wireless channel model meets an updating condition determined by a first discriminator based on the first channel information, the first discriminator being a discriminator in the first generative adversarial network.

[0019] According to an aspect of the present application, a second communication device is provided, the device comprising: a processor; wherein

[0020] The processor is configured to update the first wireless channel model based on first channel information output by the first wireless channel model, in a case where the first wireless channel model satisfies an update condition, the first wireless channel model being a generator in a first generative adversarial network, and the first channel information being used to determine whether the first wireless channel model satisfies the update condition.

[0021] According to an aspect of the present application, a first communication device is provided, comprising: a processor; a transceiver connected to the processor; a memory for storing executable instructions of the processor; wherein the processor is configured to load and execute the executable instructions to implement the wireless channel model updating method performed by the first communication device as described in the above aspect.

[0022] According to an aspect of the present application, a second communication device is provided, comprising: a processor; a transceiver connected to the processor; a memory for storing executable instructions of the processor; wherein the processor is configured to load and execute the executable instructions to implement the wireless channel model updating method performed by the second communication device as described in the above aspect.

[0023] According to an aspect of the present application, a computer readable storage medium is provided, which stores executable instructions, the executable instructions being loaded and executed by the processor to implement the wireless channel model updating method as described in the above aspect.

[0024] According to an aspect of the present application, a computer program product or computer program is provided, which comprises computer instructions stored in a computer readable storage medium, the computer instructions being read by the processor of the computer device from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to perform the wireless channel model updating method as described in the above aspect.

[0025] According to an aspect of the present application, a chip is provided, which comprises programmable logic circuit or program, and is used to implement the wireless channel model updating method as described in the above aspect.

[0026] The technical solutions provided by the embodiments of the present application have at least the following beneficial effects:

[0027] In training the wireless channel model, the wireless channel model is taken as a generator and a discriminator to form a generative adversarial network, and the generative adversarial network is subjected to adversarial training to obtain a trained wireless channel model and the discriminator. In the application stage, the wireless channel model is used to perform wireless channel related services, and the discriminator is used to evaluate the data output by the wireless channel model, and the advantages and disadvantages of the wireless channel model are determined according to the evaluation result of the discriminator. When it is determined according to the evaluation result that the wireless channel model is deteriorated, the wireless channel model is updated. The method can use the discriminator trained in the training process to evaluate the output data of the model in the application stage, so as to realize real-time sensing of the reliability of the output data of the model, and update the model in time when the output data of the model is unreliable. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0029] Figure 1 is a structural diagram of a neural network model provided by an example embodiment of the present application;

[0030] Figure 2 is a structural diagram of a neural network model provided by an example embodiment of the present application;

[0031] Figure 3 is a flowchart of an updating method of a wireless channel model provided by an example embodiment of the present application;

[0032] Figure 4 is a flowchart of an updating method of a wireless channel model provided by an example embodiment of the present application;

[0033] Figure 5 is a flowchart of an updating method of a wireless channel model provided by an example embodiment of the present application;

[0034] Figure 6 is a flowchart of an updating method of a wireless channel model provided by an example embodiment of the present application;

[0035] Figure 7 is a schematic diagram of an implementation environment of an updating method of a wireless channel model provided by an example embodiment of the present application;

[0036] Figure 8 is a method flowchart of an updating method of a wireless channel model provided by an example embodiment of the present application;

[0037] Figure 9Method flow chart of the updating method of the wireless channel model provided by an example embodiment of the present application;

[0038] Figure 10 Flow chart of the updating method of the wireless channel model provided by an example embodiment of the present application;

[0039] Figure 11 Method flow chart of the updating method of the wireless channel model provided by an example embodiment of the present application;

[0040] Figure 12 Method flow chart of the updating method of the wireless channel model provided by an example embodiment of the present application;

[0041] Figure 13 Flow chart of the updating method provided by an example embodiment of the present application;

[0042] Figure 14 Flow chart of the updating method provided by an example embodiment of the present application;

[0043] Figure 15 Method flow chart of the updating method of the wireless channel model provided by an example embodiment of the present application;

[0044] Figure 16 Flow chart of the updating method provided by an example embodiment of the present application;

[0045] Figure 17 Flow chart of the updating method provided by an example embodiment of the present application;

[0046] Figure 18 Method flow chart of the updating method of the wireless channel model provided by an example embodiment of the present application;

[0047] Figure 19 Flow chart of the updating method provided by an example embodiment of the present application;

[0048] Figure 20 Flow chart of the updating method provided by an example embodiment of the present application;

[0049] Figure 21 Structure block diagram of the first communication device provided by an example embodiment of the present application;

[0050] Figure 22 Structure block diagram of the second communication device provided by an example embodiment of the present application;

[0051] Figure 23 Structure diagram of the communication device provided by an example embodiment of the present application. DETAILED DESCRIPTION

[0052] For the purpose, technical solutions and advantages of the present application to be more clear, the following will be further described in detail with the embodiments of the present application in conjunction with the drawings.

[0053] Neural network: Artificial Neural Networks (ANN) is also referred to as neural network (NN) or connection model, which is a kind of algorithm mathematical model that simulates the behavior characteristics of animal neural network and carries out distributed parallel information processing. The basic structure of a simple neural network is shown in Figure 1 , which includes input layer 101, hidden layer 102 and output layer 103. The input layer 101 is responsible for receiving data, the hidden layer 102 processes the data, and the final result is generated in the output layer 103. Among them, each node represents a processing unit, which can also be called a neuron, and multiple neurons form a layer of neural network. Multiple layers of information transmission and processing construct a whole neural network.

[0054] With the continuous development of neural network research, in recent years, neural network deep learning algorithm has been proposed, more hidden layers are introduced, and feature learning is carried out through multi-hidden layer neural network layer by layer training, which greatly improves the learning and processing ability of neural network, and is widely applied in pattern recognition, signal processing, optimization combination, anomaly detection and other aspects.

[0055] With the development of deep learning, convolutional neural network (CNN) has been further studied. As shown in Figure 2 , the basic structure of a convolutional neural network includes input layer 101, multiple convolutional layers 104, multiple pooling layers 105, fully connected layer 106 and output layer 103. The introduction of convolutional layer 104 and pooling layer 105 effectively controls the dramatic increase of network parameters, limits the number of parameters and excavates the characteristics of local structure, and improves the robustness of algorithm.

[0056] The neural network model trained by using a limited data set will not be applicable to new business data when performing actual business, so as to cause the model to fail. In order to ensure the normal operation of the business, the neural network model needs to be updated in time. As shown in Figure 3As shown, the model training module 201 uses the training data collected by the data collection module 204 to train the model, deploys the trained model to the model inference module 202 for application. The model inference module 202 outputs output data according to the inference data in the data collection module 204, sends the output data to the business application module 203 for business execution. At the same time, the model inference module 202 collects model performance feedback, evaluates whether the model needs to be updated based on historical performance, feeds back the model performance to the model training module 201, so that the training module 201 updates the model, and sends the updated model to the model inference module 202.

[0057] With the development of neural networks, neural networks are increasingly used in wireless communication systems, such as neural network-based CSI (Channel State Information) feedback, channel estimation, CSI prediction, etc.

[0058] Channel estimation: Due to the complexity and time-varying nature of the wireless channel environment, the estimation and recovery of the wireless channel by the receiver directly affects the recovery performance of the received data. The channel estimation and recovery process in the related art is as shown in Figure 4 As shown, in addition to transmitting the sending data symbol 301, the transmitter also transmits a series of pilot symbols (sending reference signal symbol 302) known to the receiver, such as CSI-RS (Reference Signal), DMRS (Demodulation Reference Signal), etc. The receiver receives the received data symbol 303 corresponding to the sending data symbol 301, and the received reference signal symbol 304 corresponding to the sending reference signal symbol 302, and performs channel estimation 311: the receiver estimates the channel information at the reference signal position according to the real pilot (sending reference signal symbol 302) and the received pilot (received reference signal symbol 304) using the LS algorithm or other methods. Then, based on the estimated channel information, channel recovery 312 is performed: the receiver recovers the channel information on the full time-frequency resource according to the estimated channel information at the pilot position (reference signal position) using the interpolation algorithm, which is used for subsequent channel information feedback or data recovery, etc.

[0059] CSI Feedback: Given the tremendous success of AI (Artificial Intelligence) technology, especially deep learning, in computer vision and natural language processing, the communications field has begun to explore its application to solve technical challenges that traditional communication methods struggle with. The neural network architectures commonly used in deep learning are non-linear and data-driven, enabling feature extraction from actual channel matrix data and, at the base station level, reconstructing the compressed channel matrix information from the UE (User Equipment) side as accurately as possible. This not only ensures accurate channel information reconstruction but also reduces CSI feedback overhead at the UE side. Deep learning-based CSI feedback treats channel information as an image to be compressed, using a deep learning autoencoder to compress and feed back the channel information. At the receiver, the compressed channel image is reconstructed, thus preserving channel information to a greater extent.

[0060] An autoencoder comprises an encoder and a decoder. The encoder is deployed at the transmitting end (UE side), and the decoder is deployed at the receiving end (base station / access network equipment side). Specifically, after obtaining channel information through channel estimation, the transmitting end uses the encoder to compress and encode the channel information to obtain an encoded result. The transmitting end then sends the encoded result to the receiving end, i.e., feeding back the compressed bitstream to the receiving end via the air interface feedback link. The receiving end receives the encoded result and uses the decoder to decode it to obtain the channel information. In other words, the decoder reconstructs the channel information based on the feedback bitstream to obtain complete CSI feedback. For example, the encoder can employ a stack of multiple fully connected layers, and the decoder can employ a design of convolutional layers and residual structures. For example, without changing this encoding / decoding framework, the network model structure within the encoder and decoder can be flexibly designed.

[0061] CSI prediction: By inputting the periodic CSI feedback information received at the receiver into a neural network model, it is possible to predict the CSI at non-measurement times, thereby obtaining a more complete CSI. For example... Figure 5 As shown, at the time corresponding to the shaded box 305, the receiver receives periodic CSI feedback, while at the times corresponding to the white box 306 and the black box 307, there is no CSI feedback, where the black box 307 represents several future times. The CSI feedback received at the time corresponding to the shaded box 305 is input into the neural network model, thereby outputting the predicted CSI at the time corresponding to the black box 307.

[0062] Generative Adversarial Networks (GANs) are trained adversarially to make the samples generated by the generator conform to the distribution of real data.

[0063] In the generative adversarial network, there are two networks for adversarial training. One is the discriminator network (discriminator), and the goal of the discriminator is to accurately distinguish whether the input data is real data or data generated by the generator; the other is the generation network (generator), and the goal of the generator is to generate samples that the discriminator cannot distinguish. The two networks with opposite goals are alternately trained. When the model converges, if the discriminator can no longer determine whether the data output by the generator is real data, the generator can generate samples conforming to the real data distribution.

[0064] The generative adversarial network is a training method of a neural network model, based on Figure 6 A simple description of an adversarial training method for a generative adversarial network is given. The generative adversarial network is composed of a generator 308 and a discriminator 309.

[0065] In the training phase, first fix the model parameters in the generator, and train the discriminator: input the input data into the generator to obtain the generated data. The generated data output by the generator and the real data in the sample data set are respectively input into the discriminator as inputs, and the prediction labels are respectively input into the discriminator to obtain the prediction labels. The loss value is calculated based on the prediction label and the actual label. Among them, when the data input into the discriminator is generated data, the actual label is a generated label, and when the data input into the discriminator is real data, the actual label is a real label. The model parameters in the discriminator are adjusted based on the error (loss value) between the prediction label and the actual label, and the discriminator is trained so that the prediction label output by the discriminator is close to the actual label, that is, the discriminator can accurately distinguish whether the input data is real data or generated data.

[0066] After the discriminator converges, fix the model parameters in the discriminator, and train the generator. When training the generator, the generator and the discriminator are regarded as a whole, the input data is input into the generator to obtain the generated data, the generated data is input into the discriminator to obtain the prediction label, and then the loss value is calculated based on the prediction label and the real label. The model parameters in the generator are adjusted based on the error (loss value) between the prediction label and the real label, and the generator is trained, that is, the data output by the generator can be distinguished as real data by the discriminator, which means that the generated data generated by the generator is close to the real data.

[0067] For example, according to the above method, the model parameters of one network are fixed, the other network is trained, and the generator is iteratively trained to converge, and finally a trained generator is obtained. The trained generator can output generated data close to real data based on input data, and the discriminator has fully learned the characteristics of real data.

[0068] In the related art, the updating mode of the neural network wireless channel model is mostly based on the model performance feedback on the inference side, and whether the current neural network model needs to be updated is determined after the performance index is quantized. That is, the model updating in the related art is driven by feedback data.

[0069] The resource overhead of air interface transmission is large when the model performance data is fed back in real time by using the method in the related art. Periodic feedback of performance data can reduce the overhead caused by model performance feedback, but at the cost of real-time awareness of model degradation.

[0070] In addition, the feedback of model performance needs to go through multiple modules, from model output to specific business modules to guide business decisions, and the business modules bear the role of quantifying the reliability of the model, and indirectly feedback the performance of the model through business indicators. This method of model performance feedback involves multiple network element interactions, and the feedback process is complicated, and it only indirectly reflects the performance of the neural network model, which is a relatively coarse feedback.

[0071] Moreover, the above-mentioned model performance feedback scheme needs to continuously observe performance feedback for a period of time to roughly quantify the degradation degree of the neural network model and qualitatively determine whether the model is invalid. Such a feedback scheme does not guide how the model is updated. The feedback form is single and does not have diversity. In the face of complex scenarios, the model may not be able to update in time, and it is impossible to dynamically adjust according to the change of data distribution, which makes the model fall into the dilemma of updating unable to catch up with data changes, affecting the long-term performance of the model.

[0072] Further, the neural network model is mostly a single neural network model or a fusion of multiple neural network models with the same function, and the neural network model is only responsible for predicting the output and does not output a reliability indication, so it is impossible to evaluate in real time whether the current output result is reliable, thereby causing the output data based on the neural network model to guide the decision of the business module to have a certain blindness. The above-mentioned model performance feedback scheme also belongs to post-feedback and is a remedial measure. That is, the above-mentioned model updating scheme is a remedial updating scheme under the premise of sacrificing business performance.

[0073] The model performance based on deep learning is strongly related to the data distribution. Due to the instability of the wireless environment, the data distribution is inevitably affected by factors such as time, environment, and system strategy. Therefore, the neural network model cannot avoid invalidation.

[0074] In summary, how to effectively and quickly update the neural network model to cope with the change of data distribution caused by various uncontrollable factors, and how to evolve the post-remedial model updating scheme to a timely awareness of model degradation and timely proactive updating scheme, are the model generalization problems that need to be solved.

[0075] To solve the above problems, the embodiment of the present application provides a wireless channel model updating method, which can timely perceive model degradation and actively update the model.

[0076] The network architecture and service scenarios described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of network architecture and the emergence of new service scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0077] Please refer to Figure 7 which shows a schematic diagram of a network architecture 100 provided by an embodiment of the present application. The network architecture 100 can include a terminal device 10 and a network device, wherein the network device can include at least one of an access network device 20 and a core network device 30.

[0078] The terminal device 10 can refer to 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 wireless communication device, a user agent or a user apparatus. Optionally, the terminal device 10 can also be a cellular phone, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a wearable device, a terminal device in a 5th generation system (5GS) or a terminal device in a future evolved public land mobile network (PLMN), etc., and the embodiments of the present application are not limited thereto. For the convenience of description, the above-mentioned devices are collectively referred to as terminal devices. The number of terminal devices 10 is usually multiple, and one or more terminal devices 10 can be distributed in a cell managed by each access network device 20.

[0079] The access network device 20 is a device deployed in an access network to provide wireless communication functions for the terminal device 10. The access network device 20 can include various forms of macro base stations, micro base stations, relay stations, access points, and the like. In systems using different wireless access technologies, the names of devices with access network device functions can be different, for example, in a 5G NR system, it is called gNodeB or gNB. As communication technology evolves, the name of the "access network device" may change. For ease of description, in the embodiments of the present application, the above-mentioned devices that provide wireless communication functions for the terminal device 10 are collectively referred to as access network devices. Optionally, through the access network device 20, a communication relationship can be established between the terminal device 10 and the core network device 30. Illustratively, in a long term evolution (Long Term Evolution, LTE) system, the access network device 20 can be an evolved universal terrestrial radio access network (Evolved Universal Terrestrial Radio Access Network, EUTRAN) or one or more eNodeBs in the EUTRAN; in a 5G NR system, the access network device 20 can be a radio access network (Radio Access Network, RAN) or one or more gNBs in the RAN. In the embodiments of the present application, the network device refers to the access network device 20, such as a base station, unless otherwise specified.

[0080] The core network device 30 is a device deployed in the core network, and the main functions of the core network device 30 are to provide user connection, manage users, and complete bearer for services, and provide an interface to external networks as a bearer network. For example, the core network device in the 5G NR system can include an access and mobility management function (Access and Mobility Management Function, AMF) network element, a user plane function (User Plane Function, UPF) network element, and a session management function (Session Management Function, SMF) network element.

[0081] In one example, the access network device 20 and the core network device 30 communicate with each other through some air interface technology, such as the NG interface in the 5G NR system. The access network device 20 and the terminal device 10 communicate with each other through some air interface technology, such as the Uu interface.

[0082] The "5G NR system" in the embodiments of the present disclosure can also be referred to as a 5G system or an NR system, but those skilled in the art can understand its meaning. The technical solutions described in the embodiments of the present disclosure can be applicable to the 5G NR system, and can also be applicable to the subsequent evolution system of the 5G NR system.

[0083] Figure 8 A flow chart of a method for updating a wireless channel model is shown. The embodiment is exemplified by the method being applied in a first communication device, for example, the first communication device can be a terminal 10 or a network device 20. The method comprises: Figure 1

[0084] Step 410: obtaining first channel information by a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network.

[0085] The first wireless channel model is a neural network model obtained by training the generator and a first discriminator.

[0086] The embodiment of the application is exemplified by the method for updating a wireless channel model being applied to a wireless channel model. The method can also be applied to other neural network models in the field of communication, or neural network models in other fields (for example, image processing, natural language processing). When applied to other neural network models, the name of the other neural network model can be used to replace the "wireless channel model" in the embodiment of the application.

[0087] For example, step 410 is a step of applying the trained first wireless channel model. In the model application process, the first wireless channel model is used to obtain first channel information based on second channel information. The first channel information is used to perform a wireless channel related service, or guide the decision of a wireless channel related service.

[0088] The first wireless channel model and the second wireless channel model are only used to distinguish the wireless channel model before updating and the wireless channel model after updating. Both of them can be called a wireless channel model.

[0089] The network structure of the wireless channel model can be arbitrary, and the embodiment of the application does not limit it.

[0090] For example, the first channel information is obtained by the first wireless channel model based on the second channel information, which means that the second channel information is input data, the first channel information is output data, the second channel information is input into the first wireless channel model, and the first wireless channel model outputs the first channel information based on the input second channel information.

[0091] The second channel information is channel information input into the wireless channel model. The first channel information is channel information output by the wireless channel model based on the second channel information. For example, the second channel information and the first channel information are channel information of the same type, or channel information of different types.

[0092] ​The channel information (second channel information or first channel information) is channel-related data, for example, the channel information can include at least one of the following parameters: a transmitting antenna, a receiving antenna, a time delay, a number of FDM (Frequency Division Multiplexing) symbols, a number of subcarriers, a characteristic vector obtained after eigenvalue decomposition of a channel, and the like.

[0093] In the training phase, the first wireless channel model and the first discriminator form a first generative adversarial network, wherein the first wireless channel model serves as a generator in the first generative adversarial network, and the first wireless channel model and the first discriminator are trained by using an adversarial training method of the generative adversarial network.

[0094] Step 420: In a case where it is determined by the first discriminator based on the first channel information that the first wireless channel model meets an updating condition, updating the first wireless channel model is triggered, and the first discriminator is a discriminator in the first generative adversarial network.

[0095] For example, the first discriminator is used to determine whether the first wireless channel model is deteriorated. The first discriminator is used to evaluate the first channel information output by the first wireless channel model to obtain a first probability value. If the first probability value meets an updating condition, it is indicated that the first wireless channel model meets the updating condition, and updating the first wireless channel model is triggered.

[0096] The first discriminator is used to evaluate the first channel information to obtain a first probability value (a first evaluation result).

[0097] The first channel information is input into the first discriminator to obtain the first probability value. For example, the first probability value is a value between 0 and 1, or the first probability value is 0 or 1.

[0098] In the training phase, the first discriminator is trained to be used to determine whether input data is real data. The first discriminator is used to adversarially train the generator (the first wireless channel model) so that the generator can output channel information close to real channel information.

[0099] In the model application phase in step 420, the first discriminator is continuously used to determine the reliability of the first channel information. The closer the first probability value is to real data, the more reliable the first channel information is, and vice versa.

[0100] It should be noted that, in the related art, the discriminator is only used in the training phase for adversarial training with the generator, and in the application phase of the model, the discriminator is discarded and only the generator is used to perform services. However, in the method provided in the embodiments of the present application, the discriminator is creatively used in the application phase to evaluate the output result of the generator, so that whether the generator is deteriorated can be observed in real time based on the evaluation result output by the discriminator.

[0101] The first discriminator and the second discriminator are only used to distinguish the discriminator before the update and the discriminator after the update, and both can be referred to as discriminators. Among them, the first discriminator is a discriminator obtained by adversarial training with the first wireless channel model, and the second discriminator is a discriminator used for adversarial training with the second wireless channel model, or the second discriminator is a discriminator obtained by adversarial training with the second wireless channel model.

[0102] Illustratively, the discriminator is a binary classification model. For example, the output of the discriminator is 0 or 1, where 0 represents one class and 1 represents another class. For example, 0 represents non-real data and 1 represents real data. Alternatively, the output of the discriminator is a value from 0 to 1, where the closer the value is to 0, the closer the data input to the discriminator is to non-real data, and the closer the value is to 1, the closer the data input to the discriminator is to real data. 0.5 means that it cannot be determined whether the data input to the discriminator is real data or non-real data.

[0103] In a case where it is determined based on the first probability value that the first wireless channel model satisfies the update condition, the first wireless channel model is updated.

[0104] The degradation of the first wireless channel model is determined based on the first probability value, and in a case where it is determined based on the first probability value that the first wireless channel model satisfies the update condition, i.e., the performance of the first wireless channel model is poor, the first wireless channel model is updated to obtain the second wireless channel model.

[0105] Illustratively, the update condition is used to determine whether the first wireless channel model is degraded. When the first channel information output by the first wireless channel model is more likely to be judged by the first discriminator as non-real data (i.e., the probability value is 0 or the probability value is lower than 0.5), it means that the first wireless channel model has been degraded and cannot output more real channel information, so the first wireless channel model needs to be updated. That is, degradation means that the channel information output by the first wireless channel model loses authenticity / reliability.

[0106] Due to the change in the data layout of the channel information (second channel information) input to the first wireless channel model, the first wireless channel model can not be able to accurately extract features in the new second channel information, and can not be able to output more real channel information. At this time, the first wireless channel model needs to be updated so that it can relearn the features in the second channel information after the change in the data layout and output more real channel data.

[0107] The updating the first wireless channel model refers to continuously training the first wireless channel model to obtain a second wireless channel model. That is, the first wireless channel model is trained using new training samples to obtain the second wireless channel model. Illustratively, the second channel information is included in the training samples, that is, the first wireless channel model is trained using the real channel information input into the first wireless channel model in the application process, so that the first wireless channel model learns the characteristics of the second channel information and adapts to data changes.

[0108] To sum up, the method provided in the embodiment trains the wireless channel model as a generator and a discriminator to form a generative adversarial network for adversarial training, to obtain a trained wireless channel model and a discriminator. In the application stage, the wireless channel model is used to perform wireless channel related services, and the discriminator is used to evaluate the data output by the wireless channel model, and the advantages and disadvantages of the wireless channel model are determined according to the evaluation result of the discriminator. When it is determined according to the evaluation result that the wireless channel model is deteriorated, the wireless channel model is updated. The method can use the discriminator trained in the training process to evaluate the output data of the model in the application stage, so as to realize real-time perception of the reliability of the output data of the model, and update the model in time when the output data of the model is unreliable.

[0109] Figure 9 A flowchart of a wireless channel model updating method provided by an example embodiment of the present application is shown. The embodiment takes the method applied in the second communication device as an example, for example, the second communication device can be the terminal 10 or the network device 20 shown. The method includes: Figure 1

[0110] Step 310: In the case where the first wireless channel model meets the updating condition determined by the first discriminator based on the first channel information, the first wireless channel model is updated, the first channel information is output by the first wireless channel model, the first wireless channel model is a generator in the first generative adversarial network, and the first discriminator is a discriminator in the first generative adversarial network.

[0111] Illustratively, the second communication device cooperates with the first communication device to perform wireless channel services.

[0112] 1. In one case, the second communication device sends the second channel information to the first communication device, the first communication device inputs the second channel information into the first wireless channel model to obtain the first channel information, and applies the first channel information to wireless channel service decision. The second channel information is the input information when the first wireless channel model generates the first channel information.

[0113] At this time, the first communication device can use the first discriminator to determine whether the first wireless channel model needs to be updated, and the first communication device can update the first wireless channel model. ​

[0114] The first communication device can also determine, by using the first discriminator, whether the first wireless channel model needs to be updated, and send an update indication to the second communication device to instruct the second communication device to update the first wireless channel model to obtain a second wireless channel model, and send the second wireless channel model to the first communication device by the second communication device.

[0115] 2. In another case, the first wireless channel model includes both an encoder and a decoder. The encoder and the decoder are respectively deployed in the first communication device and the second communication device.

[0116] At this time, the first communication device can determine, by using the first discriminator, whether the first wireless channel model needs to be updated, and update the first wireless channel model by the first communication device, and send the second decoder and the second discriminator obtained after the update to the second communication device.

[0117] The second communication device can also determine, by using the first discriminator, whether the first wireless channel model needs to be updated, and send an update indication to the first communication device to instruct the first communication device to update the first wireless channel model, and send the second decoder and the second discriminator obtained after the update to the second communication device.

[0118] The first communication device can also determine, by using the first discriminator, whether the first wireless channel model needs to be updated, and send an update indication to the second communication device to instruct the second communication device to update the first wireless channel model, and send the second encoder, the second decoder and the second discriminator obtained after the update to the first communication device.

[0119] The second communication device can also determine, by using the first discriminator, whether the first wireless channel model needs to be updated, and update the first wireless channel model by the second communication device, and send the second encoder obtained after the update to the first communication device.

[0120] For example, when the first wireless channel model is only deployed in a single communication device, four cases of division of labor between the first communication device and the second communication device are given.

[0121] Case 1: The first wireless channel model is deployed in the first communication device, and the first communication device determines that the model is degraded and performs update of the wireless channel model.

[0122] The second communication device sends second channel information to the first communication device, and the first communication device receives the second channel information sent by the second communication device. The first communication device obtains first channel information based on the second channel information by using the first wireless channel model. The first communication device updates the first wireless channel model when it is determined by the first discriminator based on the first channel information that the first wireless channel model meets the update condition.

[0123] Case 2: The first wireless channel model is deployed at the side of the first communication device, the first communication device determines that the model is deteriorated, and instructs the second communication device to perform the update of the wireless channel model.

[0124] The second communication device sends second channel information to the first communication device, and the first communication device receives the second channel information sent by the second communication device. The first communication device obtains first channel information based on the second channel information through the first wireless channel model. In the case where the first communication device determines that the first wireless channel model meets the update condition based on the first channel information through the first discriminator, the first communication device sends an update instruction to the second communication device. The second communication device receives the update instruction, updates the first wireless channel model to obtain a second wireless channel model, and sends the second wireless channel model and a second discriminator to the first communication device. The first communication device receives and deploys the second wireless channel model and the second discriminator.

[0125] Case 3: The first wireless channel model is deployed at the side of the first communication device, the second communication device determines that the model is deteriorated, and performs the update of the wireless channel model.

[0126] The second communication device sends second channel information to the first communication device, and the first communication device receives the second channel information sent by the second communication device. The first communication device obtains first channel information based on the second channel information through the first wireless channel model. The second communication device also obtains the first channel information based on the second channel information through the first wireless channel model. In the case where the second communication device determines that the first wireless channel model meets the update condition based on the first channel information through the first discriminator, the second communication device updates the first wireless channel model. The second communication device sends the second wireless channel model and the second discriminator to the first communication device. The first communication device receives and deploys the second wireless channel model and the second discriminator.

[0127] Case 4: The first wireless channel model is deployed at the side of the first communication device, the second communication device determines that the model is deteriorated, and instructs the first communication device to perform the update of the wireless channel model.

[0128] The second communication device sends second channel information to the first communication device, and the first communication device receives the second channel information sent by the second communication device. The first communication device obtains first channel information based on the second channel information through the first wireless channel model. The second communication device also obtains the first channel information based on the second channel information through the first wireless channel model. In the case where the second communication device determines that the first wireless channel model meets the update condition based on the first channel information through the first discriminator, the second communication device sends an update instruction to the first communication device. The first communication device receives the update instruction, updates the first wireless channel model to obtain a second wireless channel model, and sends the second wireless channel model and a second discriminator to the second communication device.

[0129] Exemplarily, four enumerations of the division of the first communication device and the second communication device are given when the first wireless channel model includes the first encoder and the first decoder.

[0130] Case 5: The first decoder is deployed at the first communication device side, and the first encoder is deployed at the second communication device side. The first communication device discriminates that the model is deteriorated, and performs the update of the wireless channel model.

[0131] The second communication device obtains the first encoding result based on the second channel information through the first encoder, and sends the first encoding result to the first communication device. The first communication device receives the first encoding result sent by the second communication device. The first communication device obtains the first channel information based on the first encoding result through the first decoder. The first communication device updates the first wireless channel model to obtain the second wireless channel model in the case that the first communication device determines that the first wireless channel model meets the update condition based on the first channel information through the first discriminator. The first communication device sends the second encoder to the second communication device. The second communication device receives the second encoder sent by the first communication device.

[0132] Case 6: The first decoder is deployed at the first communication device side, and the first encoder is deployed at the second communication device side. The first communication device discriminates that the model is deteriorated, and instructs the second communication device to perform the update of the wireless channel model.

[0133] The second communication device obtains the first encoding result based on the second channel information through the first encoder, and sends the first encoding result to the first communication device. The first communication device receives the first encoding result sent by the second communication device. The first communication device obtains the first channel information based on the first encoding result through the first decoder. The first communication device sends the update instruction to the second communication device in the case that the first communication device determines that the first wireless channel model meets the update condition based on the first channel information through the first discriminator. The second communication device receives the update instruction, and updates the first wireless channel model to obtain the second wireless channel model. The second communication device sends the second decoder and the second discriminator to the first communication device. The first communication device receives the second decoder and the second discriminator sent by the second communication device.

[0134] Case 7: The first decoder is deployed at the first communication device side, and the first encoder is deployed at the second communication device side. The second communication device discriminates that the model is deteriorated, and performs the update of the wireless channel model.

[0135] The second communication device obtains a first encoding result based on the second channel information through a first encoder, and sends the first encoding result to the first communication device. The first communication device receives the first encoding result sent by the second communication device. The first communication device obtains the first channel information based on the first encoding result through a first decoder. The second communication device also obtains the first channel information based on the first encoding result through the first decoder. The second communication device updates the first wireless channel model to obtain a second wireless channel model in a case where the first wireless channel model is determined to meet an update condition based on the first channel information through a first discriminator. The second communication device sends a second decoder and a second discriminator to the first communication device. The first communication device receives the second decoder and the second discriminator sent by the second communication device.

[0136] Case 8: The first decoder is deployed on the first communication device side, and the first encoder is deployed on the second communication device side. The second communication device discriminates that the model is deteriorated, and instructs the first communication device to perform the update of the wireless channel model.

[0137] The second communication device obtains a first encoding result based on the second channel information through a first encoder, and sends the first encoding result to the first communication device. The first communication device receives the first encoding result sent by the second communication device. The first communication device obtains the first channel information based on the first encoding result through a first decoder. The second communication device also obtains the first channel information based on the first encoding result through the first decoder. The second communication device sends an update instruction to the first communication device in a case where the first wireless channel model is determined to meet an update condition based on the first channel information through a first discriminator. The first communication device receives the update instruction, and updates the first wireless channel model to obtain a second wireless channel model. The first communication device sends a second encoder, a second decoder and a second discriminator to the second communication device. The second communication device receives the second encoder, the second decoder and the second discriminator sent by the second communication device.

[0138] In the cases 5 to 8, the first communication device can be replaced by the second communication device, and the second communication device is replaced by the first communication device, that is, the first decoder is deployed on the second communication device side, and the first encoder is deployed on the first communication device side. In this way, four cases can also be obtained, and the four cases will not be described in detail in the embodiment.

[0139] In summary, the method provided in the embodiment trains the wireless channel model as a generator and a discriminator to form a generative adversarial network, performs adversarial training, and obtains a trained wireless channel model and a discriminator. In the application stage, the wireless channel model is used to perform wireless channel related services, and the discriminator is used to evaluate the data output by the wireless channel model, and the advantages and disadvantages of the wireless channel model are determined according to the evaluation result of the discriminator. When it is determined according to the evaluation result that the wireless channel model is deteriorated, the wireless channel model is updated. The method can use the discriminator trained by the training settlement to evaluate the output data of the model in the application stage, so as to realize real-time perception of the reliability of the model output data, and update the model in time when the model output data is unreliable.

[0140] Figure 10 An architecture diagram of the wireless channel model updating method provided in an example embodiment of the application is shown. The embodiment uses the architecture to Figure 1 The terminal 10 or the network device 20 shown is used as an example.

[0141] The architecture is composed of five main working modules: a data collection module 501, an offline joint training module 502, an online model inference module 503, an online discriminator inference module 504, and a service application module 505.

[0142] The data collection module 501 is a data platform that realizes data filtering, data structuring, and other data preprocessing work, and provides training data and inference data to the offline joint training module 502 and the online model inference module 503.

[0143] The offline joint training module 502 jointly trains the first wireless channel model and the first discriminator under the driving of the training data, and learns from the training method of the generative adversarial network. By alternately training the model parameters of the two models, the model converges until the model training is completed, the trained first wireless channel model is deployed to the online model inference module 503, and the trained first discriminator is deployed to the online discriminator inference module 504. The first discriminator supervises the output of the first wireless channel model, for example, whether the output of the first wireless channel model conforms to the characteristics of the real data distribution, if it conforms, the first discriminator outputs 1, and if it does not conform, the first discriminator outputs 0. Exemplarily, according to different rules of different services, the output (evaluation result) of the discriminator can also be a discrimination index between 0 and 1.

[0144] The online model inference module 503 receives the first wireless channel model sent by the offline joint training module 502, and completes model deployment. The offline joint training module 502 starts to intervene in the service, inputs inference data (second channel information) into the first wireless channel model, and outputs model output (first channel information) required by the service application module 505. The model output is provided to the service application module 505 to execute service decision.

[0145] The online discriminator inference module 504 receives the first discriminator sent by the offline joint training module 502, and completes first discriminator deployment. The online discriminator inference module 504 starts to intervene in the service, inputs the model output output by the first wireless channel model into the first discriminator, the first discriminator performs discriminant inference on the model output, and obtains discriminator output (evaluation result). The discriminator output is sent to the service application module 505 to assist in executing service decision.

[0146] Exemplarily, the architecture can further include a discriminator model online training module: when the online discriminator inference module 504 starts to intervene in the service, the discriminator model online training module simultaneously pre-updates the model parameters of the first discriminator online, and caches the second discriminator obtained after updating. The second discriminator is not used to execute the discriminant task (not used to discriminate the first channel information output by the first wireless channel model). The second discriminator is only used to memorize the distribution of new inference data (second channel information), to learn the characteristics of the distribution of new inference data in real time, and to always save the model parameters most suitable for the current data distribution. When the evaluation result meets the update condition, the discriminator model online training module sends the second discriminator to the offline joint training module 502, to drive the offline joint training module 502 to train the first wireless channel model based on the second discriminator, and obtain the second wireless channel model more suitable for the current data distribution.

[0147] The service application module 505 receives the model output sent by the online model inference module 503, and receives the discriminator output output by the online discriminator inference module 504. If the discriminator output meets the preset condition, the service application module 505 will directly use the output of the first wireless channel model as the basis for service decision. If the discriminator output does not meet the preset condition, the service application module 505 will not use the output of the first wireless channel model, and exemplarily, in this case, other strategy processes (strategy processes not using neural network models) can be used to execute the service.

[0148] Exemplarily, the present application embodiment exemplarily provides a variety of update condition setting schemes, and the present application embodiment further provides a variety of update modes.

[0149] Figure 11A flowchart of a method for updating a wireless channel model is shown. The method is applied to a first communication device, which can be the terminal 10 or the network device 20 shown in FIG. 1. The method includes the following steps: Figure 1

[0150] Step 401: obtaining a first wireless channel model and a first discriminator by adversarial training of an initial wireless channel model and an initial discriminator using training samples.

[0151] The first wireless channel model and the first discriminator are obtained by training before the first wireless channel model and the first discriminator are applied.

[0152] For example, a model structure of the wireless channel model and the discriminator is determined, and model parameters are initialized to obtain the initial wireless channel model and the initial discriminator. The initial wireless channel model and the initial discriminator are adversarially trained using training samples to obtain the first wireless channel model and the first discriminator.

[0153] For example, step 401 can also be performed by a second communication device. When step 401 is performed by the second communication device, the first communication device receives the first wireless channel model and the first discriminator sent by the second communication device.

[0154] Step 410: obtaining first channel information by the first wireless channel model, which is a neural network model obtained by adversarial training of a generator and the first discriminator.

[0155] For example, the first channel information is obtained by the first wireless channel model based on second channel information. One second channel information is input into the first wireless channel model to obtain one channel data. The first channel information in step 410 can refer to one channel data, or can refer to multiple channel data output by the first wireless channel model based on multiple second channel information.

[0156] Step 420: obtaining a first probability value by evaluating the first channel information by the first discriminator.

[0157] For example, the first discriminator outputs one probability value according to input of one channel data. For example, the first wireless channel model outputs one channel data, which is input into the first discriminator to output one probability value. The first probability value includes at least one probability value corresponding to each channel data in the first channel information.

[0158] ​The probability value is used to represent whether the first channel information is real data or non-real data. Alternatively, the probability value includes 0 or 1, 0 representing non-real data and 1 representing real data. Alternatively, the probability value is a numerical value (fraction or integer) between 0 and 1, and a value close to 0 represents a value close to non-real data and a value close to 1 represents a value close to real data.

[0159] The real data refers to data obtained by actual acquisition or other data acquisition methods. The non-real data (simulated data / generated data) refers to data output by the wireless channel model.

[0160] For example, the first discriminator can include multiple discriminators, and the multiple discriminators are used to determine the authenticity (whether it is real channel data) of the channel information from multiple dimensions. The first probability value includes evaluation results output by the multiple discriminators based on the first channel information.

[0161] Step 421: performing a wireless channel service based on the first channel information and the first probability value.

[0162] For example, the first channel information and the first probability value can both be applied to perform a wireless channel-related service. In related technologies, the wireless channel model only applies the first channel information to perform service decision-making, and the method provided in the embodiments of the present application further provides the first probability value to assist in performing service decision-making.

[0163] The first probability value can evaluate the reliability of the first channel information. For example, if the first channel information is one channel information, if the first probability value is non-real data, it means that the first channel information is unreliable and cannot be used to perform service decision-making; if the first probability value is real data, it means that the first channel information is reliable and can be used to perform service decision-making.

[0164] Further, if the first probability value is a numerical value between 0 and 1, the first probability value can also be used as a weight coefficient of the first channel information when the first channel information is applied to perform channel services, so as to reflect the reliability of the first channel information when the first channel information is applied to actual services.

[0165] For example, the method provided in the embodiments of the present application can use the discriminator to perceive whether the first wireless channel model is deteriorated in real time before the channel information output by the first wireless channel model is applied to the channel service, without needing to obtain performance indicators from service feedback after the channel information is applied to the channel service. The method can perceive the deterioration of the first wireless channel model in time before the channel information is applied, so as to update the wireless channel model in time and prevent the model deterioration from affecting service processing.

[0166] Step 430: in the case that the first wireless channel model is determined to satisfy the update condition based on the first probability value, updating the first wireless channel model to obtain a second wireless channel model.

[0167] For example, the determination of whether the update condition is satisfied can be a real-time determination, a determination after outputting a specified number of channel information, or a periodic determination. That is, the update condition can be determined based on the probability value of each channel information output by the first wireless channel model, so as to update the wireless channel model in real time. The update condition can also be determined based on the probability value of a specified number of channel information output by the first wireless channel model. The update condition can also be determined periodically according to the probability value obtained in the period.

[0168] For example, the first channel information includes a plurality of channel information, and the first probability value includes a plurality of probability values corresponding to the plurality of channel information; and the discriminator is configured to discriminate whether the input data is real channel data (real data) or non-real channel data (non-real data).

[0169] The update condition is a criterion for evaluating whether the wireless channel model is deteriorated, which is set based on the first probability value. The setting of the update condition can be arbitrary, and any way of representing the deterioration degree of the wireless channel model based on the first probability value can be used to implement the setting of the update condition. The embodiments of the present application do not limit the setting of the update condition.

[0170] For example, the embodiments of the present application provide several optional update conditions. The update condition includes at least one of the following four conditions:

[0171] Condition 1: the proportion of probability values in the first probability value that are lower than a first threshold is higher than a preset value.

[0172] That is, the proportion of probability values in the first probability value that are non-real channel data (i.e., the probability value is lower than the first threshold, for example, lower than 0.5) is higher than the preset value.

[0173] For example, the deterioration degree of the wireless channel model is represented by the proportion of non-real channel data, and the higher the proportion, the worse the deterioration degree of the wireless channel model; the lower the proportion, the better the wireless channel model.

[0174] For example, the proportion can be calculated once for each channel information output by the first wireless channel model. That is, the first probability value includes a plurality of probability values recently output by the first discriminator, and the proportion of the plurality of probability values recently output by the first discriminator that represent non-real channel data is calculated. If the proportion is high, the model is deteriorated and needs to be updated.

[0175] Exemplarily, the ratio can also be calculated once the first wireless channel model outputs a specified number of channel information, i.e., the first probability values include a specified number of probability values. For example, the ratio of the non-real channel data is calculated based on 100 probability values of 100 channel information output by the first wireless channel model each time the first wireless channel model outputs 100 channel information.

[0176] Exemplarily, the ratio can also be calculated periodically, i.e., the ratio is calculated based on multiple probability values output by the first discriminator within a period of time each time the period of time elapses. That is, the first probability values include multiple probability values output by the first discriminator within a first period, and the period of time of the first period is a preset time length. For example, the ratio of the non-real channel data is calculated based on 58 probability values of 58 channel information output by the first wireless channel model within 1 hour each time 1 hour elapses.

[0177] Condition 2: x consecutive probability values in the first probability values are lower than the second threshold.

[0178] That is, x consecutive probability values in the first probability values are non-real channel data (i.e., the x consecutive probability values are lower than the second threshold, for example, lower than 0.5), and x is a positive integer.

[0179] Condition 2 can also be expressed as: x probability values output by the first discriminator recently are non-real channel data. Or, there are x consecutive probability values in the multiple probability values that are non-real channel data, wherein the multiple probability values are output by the first discriminator within a first period, or the multiple probability values are a specified number of probability values output by the first discriminator.

[0180] Exemplarily, the first probability values include x probability values output by the first discriminator recently. Each time the first discriminator outputs a probability value, it is determined whether x probability values output by the first discriminator recently are all non-real channel data. If they are all non-real channel data, the first probability values meet the updating condition, and the first wireless channel model needs to be updated.

[0181] Exemplarily, the first probability values can also be N probability values output by the first discriminator within a first period, and N is an integer greater than x. It is determined whether there are x consecutive probability values in the first probability values that are all non-real channel data.

[0182] Exemplarily, the first probability values can also be M probability values output by the first discriminator consecutively, and M is a preset value. M is an integer greater than x. It is determined whether there are x consecutive probability values in the first probability values that are all non-real channel data.

[0183] Condition 3: the probability value distribution obtained according to the first probability values meets the first condition.

[0184] Exemplarily, the probability value can be a value from 0 to 1 (including decimal and integer), and whether the first wireless channel model is deteriorated can be determined based on a parameter on a probability value distribution graph / curve according to a plurality of probability value distribution graphs / curves. For example, the reliability of the channel information output by the first wireless channel model is determined according to the curvature, slope, rate of change, clustering situation and the like in the probability value distribution graph / curve, and then whether the first wireless channel model is deteriorated is determined.

[0185] The first condition satisfied by the probability value distribution refers to that the probability value distribution shows that the channel data is not reliable / non-real channel data.

[0186] Condition 4: The evaluation value calculated based on the first probability value reaches a third threshold value.

[0187] Exemplarily, other evaluation values for evaluating the deterioration degree of the model can also be calculated based on the first probability value, and whether the update condition is satisfied can be determined by whether the evaluation value reaches the third threshold value.

[0188] The above four conditions can be independently set as the update condition, or can be set as the update condition in any combination.

[0189] For example, condition 1 and condition 2 can be set together as the update condition. That is, when the proportion of the probability values in the first probability value that are lower than the first threshold value is higher than a preset value, and x consecutive probability values in the first probability value are lower than the second threshold value, it is determined that the first probability value satisfies the update condition, and the first wireless channel model is triggered to be updated.

[0190] Exemplarily, the first discriminator can further include a plurality of discriminators, for example, the first discriminator includes two discriminators. Different discriminators evaluate the authenticity of the first channel information from different dimensions. Exemplarily, the first discriminator includes at least two discriminators, and the first probability value includes at least two sub-probability values output by the at least two discriminators. When it is determined based on at least one of the at least two sub-probability values that the first wireless channel model satisfies the update condition, the first wireless channel model is triggered to be updated.

[0191] After it is determined based on the first probability value that the first wireless channel model satisfies the update condition, the first wireless channel model is triggered to be updated. Embodiments of the present application provide two ways to update the model:

[0192] Method 1: obtaining second channel data, taking the second channel data (positive samples) as training samples, and training and updating the first discriminator to obtain a second discriminator; after it is determined based on the first probability value that the first wireless channel model satisfies the update condition, the second discriminator and the first wireless channel model are used to form a generative adversarial network, and the first wireless channel model is adversarially trained to obtain a second wireless channel model.

[0193] After obtaining the first channel information by applying the first wireless channel model, the first probability value is obtained by using the first discriminator to evaluate the first channel information. At the same time, the second discriminator is obtained by training the first discriminator using the second channel information. That is, the second discriminator is always able to learn the latest data distribution characteristics of the service data by continuously training the first discriminator using the second channel information.

[0194] It should be noted that the second discriminator obtained by training is not used to evaluate the first channel information. After obtaining the second discriminator by training, the first discriminator is still used to evaluate the first channel information, that is, the first discriminator is used to evaluate the first channel information, and the second discriminator is used to update the first wireless channel model. The second discriminator is used to form a generative adversarial network with the first wireless channel model to adversarially train the first wireless channel model to obtain the second wireless channel model. In other words, the first discriminator is updated in real time, and the first wireless channel model is updated based on the first probability value to determine that the first wireless channel model meets the update condition. When the first wireless channel model is updated, the model parameters of the second discriminator are fixed and unchanged, that is, the second discriminator does not need to be trained again, and the second discriminator is directly used to adversarially train the first wireless channel model.

[0195] That is, the first discriminator is iteratively trained using the second channel information to obtain the second discriminator, and the second discriminator is used to update the first wireless channel model.

[0196] For example, the second channel information includes m channel information input into the first wireless channel model, m is a positive integer; the first discriminator is trained using the jth training sample, the jth training sample includes the jth channel information and the jth label, the jth label labels the jth channel information as real channel data, j is a positive integer not greater than m; j is set to j+1, and the above steps are repeatedly executed to iteratively train the first discriminator to obtain the second discriminator.

[0197] The first wireless channel model is trained in the following manner: for the second generative adversarial network composed of the first wireless channel model and the second discriminator, the parameters of the second discriminator are kept unchanged, and the first wireless channel model is adversarially trained using the training sample to obtain the second wireless channel model.

[0198] For example, the real-time iterative training of the first discriminator not only uses the second channel information (positive sample), but also uses the first channel information output by the first wireless channel model as a negative sample, and the second discriminator is obtained by real-time iterative training of the first discriminator.

[0199] Manner 2: After determining that the first wireless channel model satisfies the update condition based on the first probability value, the second channel information is obtained; for the generative adversarial network composed of the first discriminator and the first wireless channel model, the second channel information and the first channel information are taken as training samples to perform adversarial training, and the second wireless channel model and the second discriminator are obtained.

[0200] Manner 2 does not need to train the first discriminator in real time, but only needs to uniformly perform alternating adversarial training based on the first wireless channel model and the first discriminator after satisfying the update condition, to obtain the trained second wireless channel model and the second discriminator.

[0201] Step 440: Obtain third channel information through the second wireless channel model.

[0202] Exemplarily, after updating the first wireless channel model to obtain the second wireless channel model and the second discriminator corresponding to the second wireless channel model, the second wireless channel model is continuously applied to obtain third channel information based on fourth channel information, the second discriminator is used to evaluate the third channel information to obtain a second probability value, and a wireless channel related service is executed based on the third channel information and the third probability value.

[0203] Step 450: Obtain a second probability value by evaluating the third channel information through the second discriminator.

[0204] Step 460: Update the second wireless channel model in a case where it is determined that the second wireless channel model satisfies the update condition based on the second probability value.

[0205] The second probability value output by the second discriminator is continuously used to monitor whether the second wireless channel model needs to be updated.

[0206] In summary, the method provided in the embodiment provides multiple methods for setting update conditions, ensuring the real-time nature of model updating. Moreover, for the update of the wireless channel model, the real-time updating of the discriminator can be used to enable the discriminator to learn the data distribution change in real time, and when the wireless channel model is updated, the discriminator obtained by real-time updating is used to adversarially train the wireless channel model, so that the wireless channel model can also learn the data distribution change, and further output more real channel information. This method can feed back the data layout change trend to the model when triggering the model update, thereby guiding the model update direction and improving the model update efficiency.

[0207] Exemplarily, an example embodiment of updating a channel estimation model using the wireless channel model provided in the present application is given.

[0208] Figure 12A flow chart of the method for updating the wireless channel model is shown. The embodiment is exemplified by the method being applied in the first communication device, which can be the terminal 10 or the network device 20. The method comprises: Figure 1

[0209] Step 510: obtaining a first channel estimation result by a first channel estimation model, the first channel estimation model being a neural network model obtained by adversarial training of a generator and a first discriminator.

[0210] Exemplarily, the method is executed by an online environment module of the terminal device or the network device. The first channel estimation model and the first discriminator are obtained by adversarial training in an offline environment.

[0211] The second communication device can be a terminal device or a network device, and the second communication device sends second channel information to the first communication device, and the first communication device obtains a first channel estimation result based on the second channel information by the first channel estimation model. The second channel information can be a received reference signal and a real reference signal. Alternatively, the second channel information can be estimated channel data.

[0212] Exemplarily, the second channel information is data input to the first channel estimation model, and the first channel estimation result is data output by the first channel estimation model. The second channel information is input to the first channel estimation model, and the first channel estimation model outputs the first channel estimation result based on the input second channel information.

[0213] Exemplarily, as shown in Figure 13 Before step 510 is executed, step 601 is executed, and the first channel estimation model and the first discriminator are jointly trained. Exemplarily, as shown in Figure 14 The channel estimation model 701 and the discriminator 702 form a generative adversarial network. In the training phase, the input of the channel estimation model 701 is a pilot data set (received reference signal + real reference signal), or the input is estimated channel data (for example, the estimated channel data obtained in the channel estimation step shown in Figure 4 ), and the output is a channel data set (channel estimation result). The NMSE function is used as a loss function, the error (loss value) between the channel estimation result and the real channel data is measured, and the model parameters of the channel estimation model 501 are updated based on the loss value.

[0214] ​The first discriminator 702 takes as input the channel dataset (channel estimation results), where the channel estimation results output by the channel estimation model are labeled 0, and the actual real channel data are labeled 1. The discriminator can be understood as a binary classifier, and its loss function can be the cross-entropy loss function. The discriminator outputs a number between 0 and 1. If the discriminator cannot distinguish whether the input is a channel estimation result or real channel data, then it outputs 0.5.

[0215] The training process for the first channel estimation model and the first discriminator can be as follows: First, initialize the channel estimation model and the discriminator; input training samples into the channel estimation model to obtain the channel estimation result (labeled 0); fix the model parameters of the channel estimation model, and train the discriminator using the channel estimation result (negative sample) and real channel data (positive sample) to enable the discriminator to clearly distinguish between real channel data and channel estimation result (binary classification), thus obtaining a trained discriminator that can distinguish between real channel data and channel estimation result; then, release the fixation of the model parameters of the channel estimation model, fix the model parameters of the discriminator, and train the channel estimation model using training samples until the model converges. At this point, the channel estimation result generated by the channel estimation model is more realistic and conforms to the potential characteristics of real channel data.

[0216] It should be noted that the above training steps for generative adversarial networks are only a relatively basic and commonly used training method. Other generative adversarial network training methods can also be used for adversarial training.

[0217] After training and obtaining the first channel estimation model and the first discriminator, step 602 is executed, whereby the offline environment module sends the first channel estimation model and the first discriminator to the online environment module. Step 603 involves the online environment module deploying the first channel estimation model and the first discriminator. Subsequently, step 604 is executed, using the first channel estimation model and the first discriminator for online inference (outputting channel estimation results and evaluation results).

[0218] Step 520: Evaluate the first channel estimation result using the first discriminator to obtain the first probability value.

[0219] like Figure 13 As shown, during the inference process, step 605 is executed, and the first discriminator continuously learns new data distributions online and updates the model parameters of the first discriminator to obtain the second discriminator.

[0220] Step 530: If the first channel estimation model satisfies the update condition based on the first probability value, update the first channel estimation model to obtain the second channel estimation model.

[0221] If the first wireless channel model satisfies the update condition based on the first probability value, a first update instruction is sent to the offline environment module; a second channel estimation model is received from the offline environment module, which is obtained by updating the first channel estimation model offline.

[0222] The first update instruction also includes a second discriminator, which is obtained by training the first discriminator online in real time using the second channel information; the second channel estimation model is obtained by training the first channel estimation model offline adversarially based on the second discriminator.

[0223] like Figure 13 As shown, in step 606, when environmental changes or other factors cause changes in data distribution, the degradation level of the current first channel estimation model can be directly perceived using the evaluation results output by the first discriminator. For example, the negative discrimination results output by the discriminator are counted (e.g., the number of times the discriminator output is less than 0.8). When the probability of negative discrimination results exceeds a certain threshold (e.g., 30%) within a recent time window, it means that the current channel estimation model has degraded. To avoid continued model failure and further performance degradation, a model update needs to be triggered. Step 607: When the degradation level exceeds the budget threshold, a model update is triggered. Step 608: The second discriminator is sent to the offline environment module. Step 609: The offline environment module uses the second discriminator to update the first channel estimation model to obtain the second channel estimation model. At this time, the channel estimation results generated by the second channel estimation model are more consistent with the potential distribution characteristics of the recent real channel data. Step 610: The offline environment module sends the second channel estimation model to the online environment module. The online environment module removes the first discriminator, replaces it with the second discriminator, and replaces the first channel estimation model with the second channel estimation model, then continues the inference process.

[0224] In summary, the method provided in this embodiment, during channel estimation model training, uses the channel estimation model as a generator and discriminator to form a generative adversarial network (GAN) for adversarial training, resulting in a trained channel estimation model and discriminator. In the application phase, the channel estimation model is used to execute channel estimation-related services, and the discriminator evaluates the data output by the channel estimation model. The quality of the channel estimation model is determined based on the discriminator's evaluation results. When the evaluation results indicate that the channel estimation model has degraded, it is updated. This method can utilize the discriminator trained during the training phase to evaluate the model's output data, thereby real-time sensing of the reliability of the model's output data and timely updating the model when the output data becomes unreliable.

[0225] An exemplary embodiment of updating a CSI autoencoder model using the wireless channel model provided in this application is given.

[0226] Figure 15 A flow chart of a method for updating a wireless channel model is shown. The method is applied in a first communication device, which can be a terminal device or a network device (an access network device) as shown. Figure 1 The method comprises the following steps:

[0227] At step 710, a first CSI recovery result is obtained by a first CSI auto-encoding model, which is a neural network model trained as a generator and a first discriminator.

[0228] The first CSI auto-encoding model comprises a first CSI encoder and a first CSI decoder. The first CSI auto-encoding model is used to encode and compress the CSI feedback information at the terminal device side, and the compressed encoding result is sent to the access network device side for decoding to obtain the recovered CSI feedback information. The first CSI auto-encoding model is used to encode and recover the CSI feedback information, i.e., in an ideal case, the input CSI feedback information and the output CSI feedback information (the first CSI recovery result) of the first CSI auto-encoding model are the same.

[0229] That is, the second channel information is the CSI feedback information / CSI data. The second channel information is the data input into the first CSI auto-encoding model, and the first CSI recovery result is the output data of the first CSI auto-encoding model. The second channel information is input into the first CSI encoder to obtain a first encoding result, and the first encoding result is input into the first CSI decoder to obtain the first CSI recovery result.

[0230] Before step 510, the first CSI decoder and the first discriminator trained by the terminal device are received. The first encoding result encoded by the first CSI encoder from the real CSI data (the second channel information) is received. Step 510 comprises decoding the first encoding result by the first CSI decoder to obtain the first CSI recovery result (the first channel information).

[0231] As shown in FIG. 6, the method comprises the following steps: Figure 16As shown, the CSI self-encoding model includes a CSI encoder 703 and a CSI decoder 704. The CSI encoder is deployed at the terminal device side, and the CSI decoder is deployed at the access network device side. The CSI encoder encodes the input CSI data to obtain a first encoding result. The terminal device sends the first encoding result to the access network device. The access network device decodes the first encoding result by using the CSI decoder to obtain a first CSI recovery result. The first CSI recovery result is input into the first discriminator 705 to obtain an evaluation result. The discriminator is used to distinguish the CSI recovery result and the real CSI data.

[0232] Step 720: evaluating the first CSI recovery result by the first discriminator to obtain a first probability value.

[0233] Step 730: in a case where it is determined based on the first probability value that the first CSI self-encoding model meets an updating condition, updating the first CSI self-encoding model to obtain a second CSI self-encoding model.

[0234] The second CSI self-encoding model includes a second CSI encoder and a second CSI decoder.

[0235] In a case where it is determined based on the first probability value that the first CSI self-encoding model meets the updating condition, a second updating indication is sent to the terminal device. The second updating indication is used to instruct the terminal device to update the first CSI self-encoding model and the first discriminator. The second CSI decoder and the second discriminator sent by the terminal device are obtained by adversarial training of the first CSI self-encoding model and the first discriminator.

[0236] As Figure 17As shown, in step 801, the terminal device trains the first CSI self-encoding model and the first discriminator. In step 802, the terminal device sends the first CSI decoder and the first discriminator to the access network device. In step 803, the access network device deploys the first CSI decoder and the first discriminator. In step 804, the terminal device collects and saves the CSI data in the recent period of time. In step 805, the access network device uses the first CSI decoder and the first discriminator to output the first CSI recovery result and the first probability value in online inference. In step 806, the access network device calculates the degradation degree based on the probability value output by the first discriminator. In step 807, when the degradation degree exceeds the budget threshold, the model update is triggered. When the output of the first CSI decoder is not like the real CSI data, the first discriminator can make instant feedback. In combination with the discrimination result of the first discriminator, the access network device can immediately determine whether the CSI recovery result parsed by the first CSI decoder is qualified CSI data. This provides a reference for the next decision. When the data distribution changes due to environmental changes and other factors, the access network device can directly perceive the degradation degree of the CSI self-encoding model by using the first discriminator, and trigger the model update. In step 808, the access network device sends a second update indication to the terminal device. In step 809, the terminal device updates the first CSI self-encoding model and the first discriminator by using the CSI data saved in step 804 to obtain a second CSI self-encoding model and a second discriminator. In step 810, the terminal device sends the second CSI decoder and the second discriminator to the access network device.

[0237] In summary, the method provided in the embodiment trains the CSI self-encoding model as a generator and a discriminator to form a generative adversarial network for adversarial training, to obtain a trained CSI self-encoding model and a discriminator. In the application stage, the CSI self-encoding model is used to perform channel estimation related services, and the discriminator is used to evaluate the data output by the CSI self-encoding model, and the evaluation result of the discriminator is used to determine the pros and cons of the CSI self-encoding model. When it is determined according to the evaluation result that the CSI self-encoding model is degraded, the CSI self-encoding model is updated. The method can use the discriminator trained in the training stage to evaluate the output data of the model in the application stage, so as to perceive the reliability of the output data of the model in real time, and update the model in time when the output data of the model is unreliable.

[0238] For example, an example embodiment of an updating method of a wireless channel model provided in the application is given.

[0239] Figure 18 A flowchart of an updating method of a wireless channel model provided in an example embodiment of the application is shown. The embodiment is exemplified by the method being applied in a first communication device, which can be an access network device or a terminal device. Figure 1The illustrated network device 20 (access network device). The method comprises:

[0240] Step 910: obtaining a first CSI prediction result by a first CSI prediction auto-encoding model, the first CSI prediction auto-encoding model being a neural network model obtained by training a generator and a first discriminator in pairs.

[0241] The first CSI prediction auto-encoding model comprises a first CSI prediction encoder and a first CSI prediction decoder. Illustratively, the first CSI prediction auto-encoding model is used to, at the terminal device side, encode a second encoding result by the CSI prediction encoder based on N consecutive historical period CSIs, send the second encoding result to the access network device side, and decode the second encoding result by the CSI prediction decoder at the access network device side to obtain the first CSI prediction result. The first CSI prediction result is a predicted CSI sequence based on the N consecutive historical period CSIs. That is, referring to the illustration in Figure 5 , the first CSI prediction auto-encoding model is used to predict a CSI sequence on a future time-frequency based on N consecutive historical period CSIs.

[0242] The first CSI prediction result comprises a CSI sequence.

[0243] The second channel information comprises N consecutive historical period CSIs. The second channel information is input into the first CSI prediction encoder to obtain a second encoding result, and the second encoding result is input into the first CSI prediction decoder to obtain the first CSI prediction result (first channel information).

[0244] Before step 910, the access network device receives a first CSI prediction encoder and a first discriminator sent by the terminal device, the first CSI prediction encoder and the first discriminator being obtained by training of the terminal device. The access network device receives a second encoding result sent by the terminal device, the second encoding result being obtained by encoding a CSI sequence by the first CSI prediction encoder of the terminal device. Step 910 comprises decoding the second encoding result by the first CSI prediction decoder to obtain the first CSI prediction result.

[0245] As Figure 19As shown, the CSI prediction auto-encoding model includes a CSI prediction encoder 706 and a CSI prediction decoder 707. The CSI prediction encoder is deployed at the terminal device side, and the CSI prediction decoder is deployed at the access network device side. The CSI prediction encoder encodes the input continuous N historical period CSI to obtain a second encoding result, the terminal device sends the second encoding result to the access network device, and the access network device decodes the second encoding result by using the CSI prediction decoder to obtain a first CSI prediction result (predicted CSI sequence). The first CSI prediction result is input into the space-time relationship discriminator 705 and the time sequence relationship discriminator 709 to obtain an evaluation result. The space-time relationship discriminator 705 and the time sequence relationship discriminator 709 are used to evaluate the authenticity of the CSI prediction result in the space-time and time sequence dimensions, respectively.

[0246] Step 920: evaluating the first CSI prediction result by the first discriminator to obtain a first probability value.

[0247] For example, the first discriminator includes a first time sequence relationship discriminator and a first space-time relationship discriminator, and the first probability value includes a first time sequence probability value and a first space-time probability value.

[0248] The time sequence relationship discriminator is used to evaluate the authenticity of the CSI prediction result from the time dimension of the CSI prediction result. The space-time relationship discriminator is used to evaluate the authenticity of the CSI prediction result from the space dimension of the CSI prediction result.

[0249] The time sequence relationship discriminator learns the time correlation characteristics of the actual CSI sequence, and discriminates whether the time relationship characteristics of the predicted CSI sequence (CSI prediction result) output by the CSI prediction auto-encoding model conform to the time correlation characteristics of the actual CSI sequence. The space-time relationship discriminator learns the spatial characteristics of the CSI, and focuses on the extraction of spatial features to discriminate whether the spatial distribution characteristics of the CSI (CSI prediction result) output by the CSI prediction auto-encoding model conform to the spatial distribution characteristics of the actual CSI data.

[0250] Step 930: in the case where it is determined based on the first probability value that the first CSI prediction auto-encoding model satisfies an update condition, updating the first CSI prediction auto-encoding model to obtain a second CSI prediction auto-encoding model.

[0251] The second CSI prediction auto-encoding model includes a second CSI prediction encoder and a second CSI prediction decoder.

[0252] In the case where at least one of the first time sequence probability value and the first space-time probability value satisfies the update condition, the first CSI prediction auto-encoding model is updated to obtain a second CSI prediction auto-encoding model.

[0253] For example, the access network device receives a second CSI prediction decoder and a second discriminator sent by the terminal device; wherein the second CSI prediction decoder and the second discriminator are obtained by the terminal device updating the first CSI prediction autoencoder model and the first discriminator when at least one of the first temporal probability value and the first spatiotemporal probability value satisfies the update condition.

[0254] like Figure 20 As shown, in step 901, the terminal device trains a first CSI prediction autoencoder model and a first discriminator. In step 902, the terminal device sends a first CSI prediction decoder and a first discriminator to the access network device. In step 903, the access network device deploys the first CSI prediction decoder and the first discriminator. In step 904, the terminal device learns online, continuously updating the model parameters of the first discriminator using the CSI prediction results generated during the application process (inference process) and the real CSI sequences in the database to obtain a second discriminator. In step 905, the access network device uses the first CSI prediction decoder and the first discriminator for online inference, outputting a first CSI prediction result and a first probability value. In step 906, the terminal device calculates the degree of degradation based on the probability value output by the first discriminator. In step 907, when the degree of degradation exceeds a budget threshold, a model update is triggered. When the predicted CSI sequence output by the CSI prediction autoencoder model does not resemble a real CSI sequence, the first discriminator can provide immediate feedback. Combining the probability value of the first discriminator, the access network device can immediately determine whether the CSI prediction result parsed by the CSI prediction decoder is reliable. This provides a reference for the next step of decision-making. The first discriminator can continuously learn new data distributions on the terminal device side and pre-update model parameters. When environmental changes or other factors cause changes in the data distribution, the terminal device can directly perceive the degradation degree of the CSI prediction autoencoder model using the first discriminator and trigger an update. The terminal device uses the second discriminator obtained in real time to adversarially train the first CSI prediction autoencoder model to obtain the second CSI prediction autoencoder model. In step 908, the terminal device sends the second CSI prediction decoder and the second discriminator to the access network device.

[0255] In summary, the method provided in the embodiment trains the CSI prediction auto-encoding model as a generator and a discriminator to form a generative adversarial network, performs adversarial training, and obtains a trained CSI prediction auto-encoding model and a discriminator. In the application stage, the CSI prediction auto-encoding model is used to perform channel estimation related services, and the discriminator is used to evaluate the data output by the CSI prediction auto-encoding model, and the probability value of the discriminator is used to determine the advantages and disadvantages of the CSI prediction auto-encoding model. When the CSI prediction auto-encoding model is determined to be degraded according to the probability value, the CSI prediction auto-encoding model is updated. The method can use the discriminator trained in the training stage to evaluate the output data of the model in the application stage, so as to realize real-time sensing of the reliability of the output data of the model, and update the model in time when the output data of the model is unreliable.

[0256] Figure 22 A block diagram of a first communication device provided by an example embodiment of the present application is shown, and the device comprises:

[0257] The first model module 21 is configured to obtain first channel information by using a first wireless channel model, and the first wireless channel model is a generator in a first generative adversarial network.

[0258] The first update module 22 is configured to trigger updating of the first wireless channel model when it is determined by a first discriminator that the first wireless channel model meets an update condition based on the first channel information, and the first discriminator is a discriminator in the first generative adversarial network.

[0259] In an optional embodiment, the first update module 22 comprises:

[0260] The first sending sub-module 24 is configured to send a first update indication to a second communication device, and the first update indication is used to trigger the second communication device to update the first wireless channel model to obtain a second wireless channel model.

[0261] The first receiving sub-module 23 is configured to receive the second wireless channel model sent by the second communication device.

[0262] In an optional embodiment, the first model module 21 is configured to obtain the first channel information based on second channel information by using the first wireless channel model.

[0263] The device further comprises:

[0264] The first real-time training module 28 is configured to use the second channel information as a training sample to update the first discriminator in real time to obtain a second discriminator.

[0265] The first update module 22 is configured to update the first wireless channel model to obtain a second wireless channel model based on the second discriminator and the first wireless channel model.

[0266] In an optional embodiment, the first update module 22 is configured to update the first wireless channel model to obtain a second wireless channel model.

[0267] In an optional embodiment, the first model module 21 is configured to obtain the first channel information based on the second channel information by using the first wireless channel model.

[0268] The apparatus further includes:

[0269] The first real-time training module 28 is configured to update the first discriminator to obtain a second discriminator based on the second channel information as a training sample.

[0270] The first update module 22 is configured to update the first wireless channel model to obtain the second wireless channel model based on a second generative adversarial network composed of the second discriminator and the first wireless channel model.

[0271] In an optional embodiment, the first update module 22 is configured to update the first discriminator and the first wireless channel model to obtain a second discriminator and the second wireless channel model based on the first generative adversarial network composed of the first discriminator and the first wireless channel model.

[0272] In an optional embodiment, the apparatus further includes:

[0273] The first receiving module 27 is configured to receive second channel information sent by a second communication apparatus.

[0274] The first model module 21 is configured to obtain the first channel information based on the second channel information by using the first wireless channel model.

[0275] In an optional embodiment, the first wireless channel model includes a first encoder and a first decoder, the first decoder is deployed on the side of the first communication apparatus, and the first encoder is deployed on the side of the second communication apparatus; the second wireless channel model includes a second encoder and a second decoder.

[0276] The first receiving submodule 23 is configured to receive the second decoder and the second discriminator sent by the second communication apparatus.

[0277] In an optional embodiment, the apparatus further includes:

[0278] The first receiving module 27 is configured to receive a first encoding result sent by the second communication device, the first encoding result being obtained by the first encoder based on second channel information;

[0279] The first model module 21 is configured to obtain the first channel information based on the first encoding result by the first decoder.

[0280] In an optional embodiment, the first wireless channel model comprises a first encoder and a first decoder, the first encoder being deployed at the first communication device side, and the first decoder being deployed at the second communication device side, and the first communication device side storing the first decoder; and the second wireless channel model comprises a second encoder and a second decoder.

[0281] The first updating module 22 comprises:

[0282] The first sending sub-module 24 is configured to send the second decoder and the second discriminator to the second communication device.

[0283] In an optional embodiment,

[0284] The first wireless channel model comprises a first channel estimation model.

[0285] The first channel information comprises a first channel estimation result.

[0286] The second channel information comprises a reference signal.

[0287] In an optional embodiment,

[0288] The first wireless channel model comprises a first CSI self-encoding model, and the second wireless channel model comprises a second CSI self-encoding model.

[0289] The first channel information comprises a first CSI recovery result.

[0290] In an optional embodiment, the first communication device comprises a terminal device, and the second communication device comprises a network device.

[0291] The first wireless channel model comprises a first CSI prediction self-encoding model, and the second wireless channel model comprises a second CSI prediction self-encoding model.

[0292] The first channel information comprises a first CSI prediction result.

[0293] In an optional embodiment, the updating module comprises:

[0294] The first evaluation submodule 26 is configured to evaluate the first channel information by using the first discriminator to obtain a first probability value.

[0295] The first updating submodule 25 is configured to trigger updating the first wireless channel model when it is determined that the first wireless channel model satisfies an updating condition based on the first probability value.

[0296] In an optional embodiment, the first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding to the plurality of channel information.

[0297] The updating condition comprises at least one of the following conditions:

[0298] The proportion of probability values in the first probability value that are lower than a first threshold is higher than a preset value.

[0299] x consecutive probability values in the first probability value are lower than a second threshold, and x is a positive integer.

[0300] A probability value distribution obtained based on the first probability value satisfies a first condition.

[0301] An evaluation value calculated based on the first probability value reaches a third threshold.

[0302] In an optional embodiment, the first discriminator comprises at least two sub discriminators, and the first probability value comprises at least two sub probability values.

[0303] In an optional embodiment, the first updating submodule 25 is configured to trigger updating the first wireless channel model when it is determined that the first wireless channel model satisfies an updating condition based on at least one of the at least two sub probability values.

[0304] Figure 23 A block diagram of an updating device of a wireless channel model is shown, which can be implemented as the second communication device described above. The device comprises:

[0305] The second updating module 32 is configured to update the first wireless channel model when it is determined that the first wireless channel model satisfies an updating condition based on first channel information by using a first discriminator, the first channel information is output by the first wireless channel model, the first wireless channel model is a generator in a first generative adversarial network, and the first discriminator is a discriminator in the first generative adversarial network.

[0306] In an optional embodiment, the second updating module 32 comprises:

[0307] The second receiving submodule 33 is configured to receive a first update indication sent by the first communication device, wherein the first update indication is sent by the first communication device in a case where the first wireless channel model satisfies an update condition based on the first channel information determined by the first discriminator.

[0308] The second updating submodule 34 is configured to update the first wireless channel model to obtain a second wireless channel model.

[0309] The second sending submodule 35 is configured to send the second wireless channel model to the first communication device.

[0310] In an optional embodiment, the first update indication comprises a second discriminator, wherein the second discriminator is obtained by updating the first discriminator in real time using second channel information, and the second channel information is input information when the first wireless channel model generates the first channel information.

[0311] The second updating submodule 34 is configured to update the first wireless channel model to obtain the second wireless channel model in a second generative adversarial network composed of the second discriminator and the first wireless channel model.

[0312] In an optional embodiment, the apparatus further comprises:

[0313] The second sending module 36 is configured to send second channel information to the first communication device, wherein the second channel information is input information when the first wireless channel model generates the first channel information.

[0314] In an optional embodiment, the first wireless channel model comprises a first encoder and a first decoder, the first decoder is deployed on the first communication device side, and the first encoder is deployed on the second communication device side; and the second wireless channel model comprises a second encoder and a second decoder.

[0315] The second sending submodule 35 is configured to send the second decoder and a second discriminator to the first communication device.

[0316] In an optional embodiment, the apparatus further comprises:

[0317] The second model module 31 is configured to obtain a first encoding result based on second channel information by the first encoder, wherein the second channel information is input information when the first wireless channel model generates the first channel information.

[0318] The second sending module 36 is configured to send the first encoding result to the first communication device, and the first channel information is obtained by the first communication device based on the first encoding result by using the first decoder.

[0319] In an optional embodiment,

[0320] The first wireless channel model comprises a first channel estimation model.

[0321] The first channel information comprises a first channel estimation result.

[0322] The second channel information comprises a reference signal.

[0323] In an optional embodiment,

[0324] The first wireless channel model comprises a first CSI self-encoding model, and the second wireless channel model comprises a second CSI self-encoding model.

[0325] The first channel information comprises a first CSI recovery result.

[0326] In an optional embodiment, the second updating module 32 is configured to update the first wireless channel model when it is determined that the first wireless channel model satisfies an updating condition based on a first probability value, and the first probability value is obtained by the first discriminator based on the first channel information.

[0327] In an optional embodiment, the first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding to the plurality of channel information.

[0328] The updating condition comprises at least one of the following conditions:

[0329] The proportion of probability values in the first probability value that are lower than a first threshold value is higher than a preset value.

[0330] x consecutive probability values in the first probability value are lower than a second threshold value, and x is a positive integer.

[0331] A probability value distribution obtained based on the first probability value satisfies a first condition.

[0332] An evaluation value calculated based on the first probability value reaches a third threshold value.

[0333] In an optional embodiment, the first discriminator comprises at least two sub discriminators, and the first probability value comprises at least two sub probability values.

[0334] In an optional embodiment, the second updating module 32 is configured to update the first wireless channel model when it is determined that the first wireless channel model satisfies an updating condition based on at least one of the at least two sub-probability values.

[0335] Figure 23 A structure diagram of a communication device (terminal or network device) provided by an example embodiment of the present application is shown, which includes a processor 1001, a receiver 1002, a transmitter 1003, and a memory 1004.

[0336] The processor 1001 includes one or more processing cores, and the processor 1001 performs various functional applications and information processing by running software programs and modules.

[0337] The receiver 1002 and the transmitter 1003 can be implemented as a communication component, which can be a communication chip.

[0338] The memory 1004 is connected to the processor 1001.

[0339] The memory 1004 can be used to store at least one instruction, and the processor 1001 is configured to execute the at least one instruction to implement each step in the above method embodiments.

[0340] In addition, the memory 1004 can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, including but not limited to: a magnetic or optical disk, an Electrically-Erasable Programmable Read Only Memory (EEPROM), an Erasable Programmable Read Only Memory (EPROM), a Static Random Access Memory (SRAM), a Read-Only Memory (ROM), a magnetic storage, a flash memory, and a Programmable Read-Only Memory (PROM).

[0341] In an example embodiment, a computer readable storage medium is also provided, and the computer readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, which are loaded and executed by the processor to implement the wireless channel model updating method performed by the terminal or the network device provided by each of the above method embodiments.

[0342] In an example embodiment, a computer program product or computer program is also provided, the computer program product or computer program comprising computer instructions stored in a computer-readable storage medium, the computer instructions being readable by a processor of a communication device, the processor executing the computer instructions causing the communication device to perform the method for updating a wireless channel model performed by a terminal or a network device according to the above aspects.

[0343] The above only describes optional embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of updating a wireless channel model, the method comprising: The method is performed by a first communication device, and the method comprises: obtaining first channel information based on second channel information through a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network, the first wireless channel model comprising a first encoder and a first decoder, the first decoder being deployed at a side of the first communication device, and the first encoder being deployed at a side of a second communication device; evaluating the first channel information through a first discriminator to obtain a first probability value, the first discriminator being a discriminator in the first generative adversarial network, the first discriminator comprising a first time sequence relationship discriminator and a first space-time relationship discriminator; updating the first discriminator in real time by taking the second channel information as a training sample to obtain a second discriminator, the second discriminator being used to form a second generative adversarial network with the first wireless channel model to update the first wireless channel model; in a case where it is determined that the first wireless channel model meets an update condition based on the first probability value, sending a first update indication to the second communication device, the first update indication being used to trigger the second communication device to update the first wireless channel model to obtain a second wireless channel model, the first update indication comprising the second discriminator, and the second wireless channel model comprising a second encoder and a second decoder; receiving the second decoder and the second discriminator sent by the second communication device; wherein the first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding to the plurality of channel information one by one; the update condition comprises that x consecutive probability values in the first probability value are lower than a threshold value, x being a positive integer.

2. The method of claim 1, wherein, The method further comprises: receiving the second channel information sent by the second communication device.

3. The method of claim 1, wherein, The method further comprises: receiving a first encoding result sent by the second communication device, the first encoding result being obtained by the first encoder based on the second channel information; the first channel information is obtained by the first decoder based on the first encoding result.

4. The method of claim 2, wherein, The first communication device comprises a terminal device, and the second communication device comprises a network device; or, the first communication device comprises a network device, and the second communication device comprises a terminal device; the first wireless channel model comprises a first channel estimation model; the first channel information comprises a first channel estimation result; the second channel information comprises a reference signal.

5. The method according to claim 1 or 3, characterized in that, The first communication device comprises a network device, and the second communication device comprises a terminal device; the first wireless channel model comprises a first channel state information (CSI) self-encoding model, and the second wireless channel model comprises a second CSI self-encoding model; the first channel information comprises a first CSI recovery result.

6. The method of claim 1, wherein, The first communication device comprises a terminal device, and the second communication device comprises a network device; the first wireless channel model comprises a first CSI prediction self-encoding model, and the second wireless channel model comprises a second CSI prediction self-encoding model; the first channel information comprises a first CSI prediction result.

7. The method of claim 1, wherein, The first update indication is sent to the second communication device in a case where it is determined, based on the first probability value, that the first wireless channel model satisfies an update condition. The first update indication is sent to the second communication device in a case where it is determined, based on at least one probability value in the plurality of probability values, that the first wireless channel model satisfies the update condition.

8. A method of updating a wireless channel model, the method comprising: The method is performed by a second communication device, and the method comprises: receiving a first update indication sent by a first communication device, the first update indication being sent by the first communication device in a case where it is determined, based on a first probability value, that a first wireless channel model satisfies an update condition, the first update indication comprising a second discriminator; updating, based on a trigger of the first update indication, the first wireless channel model to obtain a second wireless channel model, the second wireless channel model comprising a second encoder and a second decoder; sending, to the first communication device, the second encoder and the second decoder; wherein the first probability value is obtained by a first discriminator evaluating first channel information, the first channel information being output by the first wireless channel model based on second channel information, the first wireless channel model being a generator in a first generative adversarial network, the first wireless channel model comprising a first encoder and a first decoder, the first decoder being deployed at a side of the first communication device, the first encoder being deployed at a side of the second communication device, the first discriminator being a discriminator in the first generative adversarial network, the first discriminator comprising a first temporal relationship discriminator and a first spatio-temporal relationship discriminator; the updating of the first wireless channel model is implemented based on a second generative adversarial network composed of the second discriminator and the first wireless channel model, the second discriminator being obtained by real-time training and updating the first discriminator with the second channel information as training samples; the first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding to the plurality of channel information one by one; the update condition comprises that x consecutive probability values in the first probability value are lower than a threshold value, x being a positive integer.

9. The method of claim 8, wherein, The method further comprises: sending, to the first communication device, the second channel information.

10. The method of claim 8, wherein, The method further comprises: obtaining, by the first encoder, a first encoding result based on the second channel information; sending, to the first communication device, the first encoding result, the first channel information being obtained by the first communication device based on the first encoding result by the first decoder.

11. The method of claim 9, wherein, The first communication device comprises a terminal device, and the second communication device comprises a network device; or, the first communication device comprises a network device, and the second communication device comprises a terminal device; The first wireless channel model comprises a first channel estimation model. The first channel information comprises a first channel estimation result. The second channel information comprises a reference signal.

12. The method of claim 8 or 10, wherein, The first communication device comprises a network device, and the second communication device comprises a terminal device; The first wireless channel model comprises a first channel state information (CSI) self-encoding model, and the second wireless channel model comprises a second CSI self-encoding model. The first channel information comprises a first CSI recovery result.

13. The method of claim 8, wherein, The first update indication is sent by the first communication device in a case where it is determined based on at least one of the plurality of probability values that the first wireless channel model satisfies the update condition.

14. A first communications device, characterized by: The first communication device comprises: a first model module, configured to obtain first channel information based on second channel information by a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network, the first wireless channel model comprising a first encoder and a first decoder, the first decoder being deployed at a side of the first communication device, and the first encoder being deployed at a side of a second communication device; a first evaluation submodule, configured to obtain a first probability value by evaluating the first channel information by a first discriminator, the first discriminator being a discriminator in the first generative adversarial network, the first discriminator comprising a first time sequence relationship discriminator and a first time-space relationship discriminator; a first update module, configured to update the first discriminator in real time to obtain a second discriminator by taking the second channel information as a training sample, the second discriminator being used to form a second generative adversarial network with the first wireless channel model to update the first wireless channel model; the first update module is further configured to send a first update indication to the second communication device in a case where it is determined based on the first probability value that the first wireless channel model satisfies an update condition, the first update indication being used to trigger the second communication device to update the first wireless channel model to obtain a second wireless channel model, the first update indication comprising the second discriminator, and the second wireless channel model comprising a second encoder and a second decoder; a first receiving module, configured to receive the second decoder and the second discriminator sent by the second communication device; wherein the first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding one-to-one to the plurality of channel information; the update condition comprises that x consecutive probability values in the first probability value are lower than a threshold value, x being a positive integer.

15. The apparatus of claim 14, wherein, The first receiving module is further configured to receive the second channel information sent by the second communication device.

16. The apparatus of claim 14, wherein, The first receiving module is further configured to receive a first encoding result sent by the second communication device, the first encoding result being obtained by the first encoder based on second channel information. The first model module is further configured to obtain the first channel information based on the first encoding result by the first decoder.

17. The apparatus of claim 14, wherein the first wireless channel model comprises a first channel estimation model; the first channel information comprises a first channel estimation result; the second channel information comprises a reference signal.

18. The apparatus of claim 14 or 16, wherein The first wireless channel model comprises a first channel state information (CSI) self-encoding model, and the second wireless channel model comprises a second CSI self-encoding model. The first channel information comprises a first CSI recovery result.

19. The apparatus of claim 14, wherein, The first wireless channel model comprises a first CSI prediction self-encoding model, and the second wireless channel model comprises a second CSI prediction self-encoding model. The first channel information comprises a first CSI prediction result.

20. The apparatus of claim 14, wherein, The first updating module is configured to, in a case where it is determined that the first wireless channel model satisfies the updating condition based on at least one of the plurality of probability values, send the first updating indication to the second communication apparatus.

21. A second communications device, characterized by The second communication apparatus comprises: The second receiving module is configured to receive the first updating indication sent by the first communication apparatus, the first updating indication being sent by the first communication apparatus in a case where it is determined that the first wireless channel model satisfies the updating condition based on the first probability value, and the first updating indication comprising a second discriminator. The second updating module is configured to, based on triggering of the first updating indication, update the first wireless channel model to obtain a second wireless channel model, the second wireless channel model comprising a second encoder and a second decoder. The second sending module is configured to send the second encoder and the second decoder to the first communication apparatus. The first probability value is obtained by a first discriminator evaluating first channel information, the first channel information being output by the first wireless channel model based on second channel information, the first wireless channel model being a generator in a first generative adversarial network, the first wireless channel model comprising a first encoder and a first decoder, the first decoder being deployed at the first communication apparatus side, the first encoder being deployed at the second communication apparatus side, the first discriminator being a discriminator in the first generative adversarial network, and the first discriminator comprising a first time sequence relationship discriminator and a first time-space relationship discriminator. Updating of the first wireless channel model is implemented based on a second generative adversarial network composed of the second discriminator and the first wireless channel model, the second discriminator being obtained by real-time training and updating of the first discriminator with the second channel information as training samples; The first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding to the plurality of channel information. The updating condition comprises that x consecutive probability values in the first probability value are lower than a threshold value, x being a positive integer.

22. The apparatus of claim 21, wherein, The second sending module is further configured to send the second channel information to the first communication apparatus.

23. The apparatus of claim 21, wherein, The second communication apparatus further comprises: The second model module is configured to obtain a first encoding result based on the second channel information by the first encoder. The second sending module is further configured to send the first encoding result to the first communication apparatus, and the first channel information is obtained by the first communication apparatus based on the first encoding result by the first decoder.

24. The apparatus of claim 21, wherein, The first wireless channel model comprises a first channel estimation model; The first channel information comprises a first channel estimation result; The second channel information comprises a reference signal.

25. The apparatus of claim 21 or 23, wherein The first wireless channel model comprises a first channel state information (CSI) autoencoder model, and the second wireless channel model comprises a second CSI autoencoder model; The first channel information comprises a first CSI recovery result.

26. The apparatus of claim 21, wherein, The first update indication is sent by the first communication device in a case that the first wireless channel model is determined to satisfy the update condition based on at least one of the plurality of probability values.

27. A first communication device, characterized by The first communication device comprises a processor, a transmitter and a receiver; wherein The processor is configured to obtain first channel information based on second channel information by a first wireless channel model, the first wireless channel model being a generator in a first generative adversarial network (GAN), the first wireless channel model comprising a first encoder and a first decoder, the first decoder being deployed at a side of the first communication device, and the first encoder being deployed at a side of a second communication device; The processor is configured to obtain a first probability value by evaluating the first channel information by a first discriminator, the first discriminator being a discriminator in the first GAN, the first discriminator comprising a first temporal relationship discriminator and a first spatio-temporal relationship discriminator; The processor is configured to update the first discriminator in real time to obtain a second discriminator by taking the second channel information as a training sample, the second discriminator being configured to form a second GAN with the first wireless channel model to update the first wireless channel model; The transmitter is configured to send a first update indication to the second communication device in a case that the first wireless channel model is determined to satisfy an update condition based on the first probability value, the first update indication being configured to trigger the second communication device to update the first wireless channel model to obtain a second wireless channel model, the first update indication comprising the second discriminator, and the second wireless channel model comprising a second encoder and a second decoder; The receiver is configured to receive the second decoder and the second discriminator sent by the second communication device; The first channel information comprises a plurality of channel information, and the first probability value comprises a plurality of probability values corresponding to the plurality of channel information. The update condition comprises that x consecutive probability values in the first probability value are lower than a threshold value, x being a positive integer.

28. A second communication device, characterized by The second communication device comprises a receiver, a processor and a transmitter; wherein The receiver is configured to receive a first update indication sent by the first communication device, the first update indication being sent by the first communication device in a case that a first wireless channel model is determined to satisfy an update condition based on a first probability value, and the first update indication comprising a second discriminator; The processor is configured to update the first wireless channel model to obtain a second wireless channel model based on triggering of the first update indication, the second wireless channel model comprising a second encoder and a second decoder; and The transmitter is configured to send the second encoder and the second discriminator to the first communication device. The transmitter is configured to send the second encoder and the second decoder to the first communication device. The first probability value is obtained by a first discriminator evaluating first channel information, the first channel information is output by a first wireless channel model based on second channel information, the first wireless channel model is a generator in a first generative adversarial network, the first wireless channel model includes a first encoder and a first decoder, the first decoder is deployed on the first communication device side, the first encoder is deployed on the second communication device side, the first discriminator is a discriminator in the first generative adversarial network, and the first discriminator includes a first time sequence relationship discriminator and a first space-time relationship discriminator. The updating of the first wireless channel model is realized based on a second generative adversarial network composed of a second discriminator and the first wireless channel model, and the second discriminator is obtained by real-time training and updating the first discriminator by taking the second channel information as a training sample. The first channel information includes a plurality of channel information, and the first probability value includes a plurality of probability values corresponding to the plurality of channel information. The updating condition includes that x consecutive probability values in the first probability value are lower than a threshold value, and x is a positive integer.

29. A first communication device, characterized by The first communication device includes: a processor; a transceiver connected to the processor; a memory for storing executable instructions of the processor; The processor is configured to load and execute the executable instructions to realize the wireless channel model updating method according to any one of claims 1 to 7.

30. A second communication device, characterized by The second communication device includes: a processor; a transceiver connected to the processor; a memory for storing executable instructions of the processor; The processor is configured to load and execute the executable instructions to realize the wireless channel model updating method according to any one of claims 8 to 13.

31. A computer readable storage medium, characterized in that, The readable storage medium stores executable instructions, and the executable instructions are loaded and executed by the processor to realize the wireless channel model updating method according to any one of claims 1 to 13.

32. A computer program product, characterised in that, The computer program product stores executable instructions, and the executable instructions are loaded and executed by the processor to realize the wireless channel model updating method according to any one of claims 1 to 13.

Images