Feedback model generation method, terminal and network equipment

Abstract

The invention relates to a feedback model generation method, a terminal and network equipment. The method comprises the following steps: first equipment firstly obtains a training data set, then trains an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder, and then keeps all parameters of the first encoder unchanged based on CSI associated with each application scene to obtain a second encoder and a second decoder; and carrying out fine tuning on the first decoder to obtain a second decoder associated with each application scene, or respectively carrying out fine tuning on the first decoder and the first encoder based on the CSI associated with each application scene by keeping part of parameters of the first encoder unchanged to obtain a second encoder and a second decoder associated with each application scene. Therefore, the model parameter scale of the CSI feedback model encoder is reduced to a certain extent, so that the storage overhead of the CSI feedback model at the terminal side is reduced.

Description

Feedback model generation method, terminal and network device Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular to a method for generating a feedback model, a terminal, and a network device. Background Art

[0002] In related technologies, a deep learning-based channel state information (CSI) feedback solution is used to reduce CSI feedback overhead. Using large-scale training data, a CSI feedback model is trained and generated. The encoder in the CSI feedback model is then deployed on the terminal side, and the decoder on the network device side. This allows for feature extraction and dimensionality reduction of the CSI data on the terminal side. The network device then decodes the received codewords and reconstructs them to output the original dimensional CSI data.

[0003] Summary of the Invention

[0004] The embodiments of the present disclosure provide a method for generating a feedback model, a terminal, and a network device, which to a certain extent solve the problem of how to reduce the model parameter scale of the CSI feedback model encoder, thereby reducing the storage overhead of the CSI feedback model in the user equipment (UE).

[0005] According to a first aspect of an embodiment of the present disclosure, a method for generating a channel state information (CSI) feedback model is proposed. The method is performed by a first device and includes:

[0006] Acquire a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0007] Training an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0008] Based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario; or, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters.

[0009] According to a second aspect of an embodiment of the present disclosure, a method for generating a channel state information (CSI) feedback model is proposed. The method is executed by a terminal, and the method includes:

[0010] Receiving a first encoder or multiple second encoders sent by the first device, wherein the multiple second encoders share some network parameters;

[0011] Deploy the first encoder or the plurality of second encoders.

[0012] According to a third aspect of an embodiment of the present disclosure, a method for generating a channel state information (CSI) feedback model is proposed. The method is performed by a network device and includes:

[0013] Acquire a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0014] Based on the CSI associated with each of the application scenarios, an initial encoder and an initial decoder in an initial CSI feedback model are trained to obtain a second decoder associated with each of the application scenarios and input data of the second decoder;

[0015] Sending a first data set associated with each application scenario to the terminal, wherein the first data set includes CSI and input data of the second decoder, and the first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0016] According to a fourth aspect of an embodiment of the present disclosure, a method for generating a channel state information (CSI) feedback model is proposed. The method is executed by a terminal, and the method includes:

[0017] receiving a first data set associated with each application scenario sent by a network device, wherein the data set includes channel state information CSI associated with each application scenario and input data of a second decoder;

[0018] A second encoder associated with each application scenario is obtained by training based on the first data set, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0019] According to a fifth aspect of an embodiment of the present disclosure, a method for generating a channel state information (CSI) feedback model is proposed, the method further comprising:

[0020] The first device obtains a training data set, wherein the training data set includes channel state information CSI in multiple application scenarios;

[0021] The first device trains an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0022] The first device, based on the CSI associated with each application scenario, keeps all parameters of the first encoder unchanged, and fine-tunes the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, keeps some parameters of the first encoder unchanged, and fine-tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or the second encoders associated with different application scenarios share some network parameters;

[0023] The first device sends a first encoder or multiple second encoders to the terminal, where the multiple second encoders share some network parameters;

[0024] The terminal receives a first encoder or multiple second encoders sent by the first device, wherein the multiple second encoders share some network parameters;

[0025] The terminal is equipped with the first encoder or the plurality of second encoders.

[0026] According to a sixth aspect of an embodiment of the present disclosure, a method for generating a channel state information (CSI) feedback model is proposed. The method is performed by a communication system, and the method includes:

[0027] The network device obtains a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0028] The network device trains an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtains a second decoder associated with each application scenario and input data of the second decoder;

[0029] The network device sends a first data set associated with each application scenario to the terminal, wherein the first data set includes CSI and input data of the second decoder;

[0030] The terminal obtains, based on the first data set, a second encoder associated with each application scenario through training, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0031] According to a seventh aspect of the embodiments of the present disclosure, a first device is provided, including:

[0032] A processing module, configured to obtain a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0033] The processing module is configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0034] The processing module is configured to, based on the CSI associated with each application scenario, keep all parameters of the first encoder unchanged, fine-tune the first decoder, and obtain a second decoder associated with each application scenario; or, based on the CSI associated with each application scenario, keep some parameters of the first encoder unchanged, fine-tune the first decoder and the first encoder respectively, and obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters.

[0035] According to an eighth aspect of an embodiment of the present disclosure, a terminal is provided, including:

[0036] a transceiver module, configured to receive a first encoder or multiple second encoders sent by a first device, wherein the multiple second encoders share some network parameters;

[0037] A processing module is used to deploy the first encoder or the multiple second encoders.

[0038] According to a ninth aspect of an embodiment of the present disclosure, a network device is provided, including:

[0039] A processing module, configured to obtain a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0040] The processing module is configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtain a second decoder associated with each application scenario and input data of the second decoder;

[0041] A transceiver module is configured to send a first data set associated with each application scenario to a terminal, wherein the first data set includes CSI and input data of the second decoder, and the first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0042] According to a tenth aspect of an embodiment of the present disclosure, a terminal is provided, including:

[0043] a transceiver module, configured to receive a first data set associated with each application scenario sent by a network device, wherein the data set includes channel state information CSI associated with each application scenario and input data of a second decoder;

[0044] A processing module is used to train a second encoder associated with each application scenario based on the first data set, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0045] According to an eleventh aspect of the present disclosure, a communication device is provided, including:

[0046] one or more processors;

[0047] The processor is used to call instructions to enable the communication device to execute the processing method described in any one of the first aspect, the second aspect, the third aspect, and the fourth aspect.

[0048] According to the twelfth aspect of an embodiment of the present disclosure, a communication system is proposed, characterized in that it includes a terminal and a network device, wherein the terminal is configured to implement the method for generating the channel state information CSI feedback model described in the second aspect and the fourth aspect, and the network device is configured to implement the method for generating the channel state information CSI feedback model described in the third aspect.

[0049] According to the thirteenth aspect of an embodiment of the present disclosure, a storage medium is proposed, which stores instructions, and is characterized in that when the instructions are executed on a communication device, the communication device executes the method for generating a channel state information CSI feedback model as described in any one of the first aspect, the second aspect, the third aspect, and the fourth aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following drawings required for describing the embodiments are introduced. The following drawings are merely some embodiments of the present disclosure and do not impose specific limitations on the protection scope of the present disclosure.

[0051] FIG1A is a schematic diagram illustrating an architecture of a communication system according to an embodiment of the present disclosure;

[0052] FIG1B is a schematic diagram of a CSI feedback model according to an embodiment of the present disclosure;

[0053] 2A-2D are interactive schematic diagrams illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure;

[0054] 3A-3H are schematic flow charts illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure;

[0055] 4A-4G are schematic flow diagrams illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure;

[0056] 5A-5C are schematic flow charts illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure;

[0057] 6A-6B are interactive schematic diagrams illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure;

[0058] FIG7A is a schematic structural diagram of a method for sharing an overall encoder model according to an embodiment of the present disclosure;

[0059] FIG7B is a structural diagram illustrating a method for sharing a partial encoder model according to an embodiment of the present disclosure;

[0060] FIG8A is a schematic structural diagram of a first device proposed in an embodiment of the present disclosure;

[0061] FIG8B is a schematic structural diagram of a terminal proposed in an embodiment of the present disclosure;

[0062] FIG8C is a schematic diagram of the structure of a network device proposed in an embodiment of the present disclosure;

[0063] FIG8D is a schematic diagram of the structure of a terminal proposed in an embodiment of the present disclosure;

[0064] FIG9A is a schematic structural diagram of a communication device proposed in an embodiment of the present disclosure;

[0065] FIG9B is a schematic diagram of the structure of the chip proposed in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0066] The embodiments of the present disclosure provide a method for generating a feedback model, a terminal, and a network device.

[0067] In a first aspect, an embodiment of the present disclosure provides a method for generating a channel state information (CSI) feedback model, the method being performed by a first device and including:

[0068] Acquire a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0069] Training an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0070] Based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario; or, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters.

[0071] In the above embodiment, the first device trains the encoder and decoder based on the CSI associated with each application scenario to obtain encoders and decoders suitable for different application scenarios, and then fine-tunes the obtained encoders and decoders for different application scenarios, so that the second encoders in different application scenarios share at least some network parameters, thereby reducing the storage overhead of the CSI feedback model in the terminal while ensuring the accuracy of the CSI feedback model.

[0072] In conjunction with some embodiments of the first aspect, in some embodiments, after obtaining the second encoder and the second decoder associated with each application scenario, the method further includes:

[0073] In the case where the different application scenarios are associated with the same first encoder, one first encoder is sent to a terminal, and multiple second decoders are sent to a network device; or

[0074] In the case where the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal, and multiple second decoders are sent to the network device.

[0075] In the above embodiment, when the first device is not a network device, based on whether the encoders associated with different application scenarios are the same or share some network parameters, the first device sends the corresponding encoder to the terminal and sends multiple decoders to the network device, thereby improving the accuracy of the CSI feedback model and ensuring the accuracy and reliability of the communication system.

[0076] In conjunction with some embodiments of the first aspect, in some embodiments, the first device is a network device, and after obtaining the second encoder and the second decoder associated with each application scenario, the method further includes:

[0077] In the case where the different application scenarios are associated with the same first encoder, sending one first encoder to the terminal; or,

[0078] In the case that the second encoders associated with the different application scenarios share part of the network parameters, the shared part of the network parameters and another part of the network parameters associated with each application scenario are sent to the terminal.

[0079] In the above embodiment, when the first device is a network device, the first device sends the corresponding encoder to the terminal based on whether the encoders associated with different application scenarios are the same or share some network parameters, thereby improving the accuracy of the CSI feedback model and reducing the storage overhead of the CSI feedback model in the terminal.

[0080] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes:

[0081] receiving first indication information sent by the terminal, wherein the first indication information is used to indicate a size of available storage space of the terminal;

[0082] Determining, according to the first indication information, a type of a second encoder corresponding to the terminal;

[0083] According to the type of the second encoder corresponding to the terminal, one second encoder applicable to all application scenarios is sent to the terminal, or multiple second encoders that share some network parameters are sent to the terminal.

[0084] In the above embodiment, the first device first confirms the size of the terminal's storage space by receiving the first indication information sent by the terminal, and then sends an encoder suitable for the terminal to the terminal, thereby improving the reliability of the CSI feedback model and reducing the storage overhead of the CSI feedback model in the terminal.

[0085] In a second aspect, an embodiment of the present disclosure proposes a method for generating a channel state information (CSI) feedback model, the method being executed by a terminal, and the method including:

[0086] Receiving a first encoder or multiple second encoders sent by the first device, wherein the multiple second encoders share some network parameters;

[0087] Deploy the first encoder or the plurality of second encoders.

[0088] In the above embodiment, the terminal receives and deploys one or more encoders sent by the first device, thereby reducing the storage overhead of the CSI feedback model in the terminal.

[0089] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes:

[0090] Sending first indication information to the first device, where the first indication information is used to indicate the performance of the terminal;

[0091] Receive a second encoder sent by the first device, or receive multiple second encoders sent by the first device, wherein the multiple second encoders share some network parameters, and the second encoder is used to encode channel state information CSI.

[0092] In the above embodiment, the terminal indicates its own performance by sending the first indication information to the first device, and then receives the encoder sent by the first device, thereby ensuring the accuracy and reliability of the generated CSI feedback model.

[0093] In combination with some embodiments of the second aspect, in some embodiments, the terminal performance includes the size of available storage space in the terminal.

[0094] In a third aspect, an embodiment of the present disclosure provides a method for generating a channel state information (CSI) feedback model, the method being executed by a network device, the method comprising:

[0095] Acquire a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0096] Based on the CSI associated with each of the application scenarios, an initial encoder and an initial decoder in an initial CSI feedback model are trained to obtain a second decoder associated with each of the application scenarios and input data of the second decoder;

[0097] Sending a first data set associated with each application scenario to the terminal, wherein the first data set includes CSI and input data of the second decoder, and the first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0098] In the above embodiment, the network device trains the encoder and decoder based on the CSI associated with each application scenario to obtain the decoder and decoder input data for each application scenario, and sends the CSI and decoder input data associated with each application scenario to the terminal, thereby providing conditions for reducing the storage overhead of the generated CSI feedback model in the terminal.

[0099] In a fourth aspect, an embodiment of the present disclosure proposes a method for generating a channel state information (CSI) feedback model, the method being executed by a terminal, and the method including:

[0100] receiving a first data set associated with each application scenario sent by a network device, wherein the data set includes channel state information CSI associated with each application scenario and input data of a second decoder;

[0101] A second encoder associated with each application scenario is obtained by training based on the first data set, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0102] In the above embodiment, the terminal trains the encoder associated with each application scenario by receiving the CSI and decoder input data associated with each application scenario sent by the network device, thereby improving the accuracy and reliability of the communication system.

[0103] In conjunction with some embodiments of the fourth aspect, in some embodiments, the training based on the first data set to obtain the second encoder associated with each application scenario includes:

[0104] determining a target training mode for a second encoder based on the performance of the terminal;

[0105] According to the target training mode, a second encoder associated with each application scenario is obtained by training based on the first data set.

[0106] In the above embodiment, the terminal first determines the target training mode of the encoder based on its own performance, and then trains the encoder based on the target training mode, as well as the CSI associated with each application scenario and the input data of the decoder, thereby reducing the storage overhead of the generated CSI feedback model in the terminal and improving the accuracy of the CSI feedback model.

[0107] In conjunction with some embodiments of the fourth aspect, in some embodiments, determining the target training mode of the second encoder according to the performance of the terminal includes:

[0108] When the size of the available storage space of the terminal is greater than a size threshold, determining the target training mode to be training the second encoder associated with each application scenario based on the CSI associated with each application scenario and input data of the second decoder, wherein the second encoders associated with different application scenarios share some network parameters;

[0109] When the size of the available storage space of the terminal is less than or equal to the size threshold, the target training mode is determined to be training a second encoder associated with all application scenarios based on the CSI associated with all application scenarios and the input data of the second decoder.

[0110] In a fifth aspect, an embodiment of the present disclosure proposes a method for generating a channel state information (CSI) feedback model, the method further comprising:

[0111] The first device obtains a training data set, wherein the training data set includes channel state information CSI in multiple application scenarios;

[0112] The first device trains an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0113] The first device, based on the CSI associated with each application scenario, keeps all parameters of the first encoder unchanged, and fine-tunes the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, keeps some parameters of the first encoder unchanged, and fine-tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or the second encoders associated with different application scenarios share some network parameters;

[0114] The first device sends a first encoder or multiple second encoders to the terminal, where the multiple second encoders share some network parameters;

[0115] The terminal receives a first encoder or multiple second encoders sent by the first device, wherein the multiple second encoders share some network parameters;

[0116] The terminal is equipped with the first encoder or the plurality of second encoders.

[0117] In a sixth aspect, an embodiment of the present disclosure provides a method for generating a channel state information (CSI) feedback model, the method being performed by a communication system, the method comprising:

[0118] The network device obtains a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0119] The network device trains an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtains a second decoder associated with each application scenario and input data of the second decoder;

[0120] The network device sends a first data set associated with each application scenario to the terminal, wherein the first data set includes CSI and input data of the second decoder;

[0121] The terminal obtains, based on the first data set, a second encoder associated with each application scenario through training, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0122] In a seventh aspect, an embodiment of the present disclosure provides a first device, the first device including:

[0123] A processing module, configured to obtain a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0124] The processing module is configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0125] The processing module is configured to, based on the CSI associated with each application scenario, keep all parameters of the first encoder unchanged, fine-tune the first decoder, and obtain a second decoder associated with each application scenario; or, based on the CSI associated with each application scenario, keep some parameters of the first encoder unchanged, fine-tune the first decoder and the first encoder respectively, and obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters.

[0126] In conjunction with some embodiments of the seventh aspect, in some embodiments, after obtaining the second encoder and the second decoder associated with each application scenario, the method further includes:

[0127] a transceiver module configured to, when the different application scenarios are associated with the same first encoder, send one first encoder to a terminal and send multiple second decoders to a network device; or

[0128] The transceiver module is used to send the shared network parameters and another part of the network parameters associated with each application scenario to the terminal when the second encoders associated with different application scenarios share some network parameters, and send multiple second decoders to the network device.

[0129] In conjunction with some embodiments of the seventh aspect, in some embodiments, the first device is a network device, and after obtaining the second encoder and the second decoder associated with each application scenario, the method further includes:

[0130] a transceiver module, configured to send one first encoder to a terminal when the different application scenarios are associated with the same first encoder; or

[0131] The transceiver module is configured to, when the second encoders associated with different application scenarios share some network parameters, send the shared network parameters and another part of network parameters associated with each application scenario to the terminal.

[0132] In conjunction with some embodiments of the seventh aspect, in some embodiments, the method further includes:

[0133] a transceiver module, configured to receive first indication information sent by the terminal, wherein the first indication information is used to indicate the size of the available storage space of the terminal;

[0134] a processing module, configured to determine a type of a second encoder corresponding to the terminal according to the first indication information;

[0135] The transceiver module is configured to send a second encoder applicable to all application scenarios to the terminal according to the type of the second encoder corresponding to the terminal, or send multiple second encoders that share some network parameters to the terminal.

[0136] In an eighth aspect, an embodiment of the present disclosure provides a terminal, comprising:

[0137] a transceiver module, configured to receive a first encoder or multiple second encoders sent by a first device, wherein the multiple second encoders share some network parameters;

[0138] A processing module is used to deploy the first encoder or the multiple second encoders.

[0139] In conjunction with some embodiments of the eighth aspect, in some embodiments, the transceiver module is further configured to:

[0140] Sending first indication information to the first device, where the first indication information is used to indicate the performance of the terminal;

[0141] Receive a second encoder sent by the first device, or receive multiple second encoders sent by the first device, wherein the multiple second encoders share some network parameters, and the second encoder is used to encode channel state information CSI.

[0142] In combination with some embodiments of the eighth aspect, in some embodiments, the terminal performance includes the size of the available storage space in the terminal.

[0143] In a ninth aspect, an embodiment of the present disclosure provides a network device, comprising:

[0144] A processing module, configured to obtain a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0145] The processing module is configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtain a second decoder associated with each application scenario and input data of the second decoder;

[0146] A transceiver module is configured to send a first data set associated with each application scenario to a terminal, wherein the first data set includes CSI and input data of the second decoder, and the first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0147] In a tenth aspect, an embodiment of the present disclosure provides a terminal, comprising:

[0148] a transceiver module, configured to receive a first data set associated with each application scenario sent by a network device, wherein the data set includes channel state information CSI associated with each application scenario and input data of a second decoder;

[0149] A processing module is used to train a second encoder associated with each application scenario based on the first data set, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0150] In conjunction with some embodiments of the tenth aspect, in some embodiments, the processing module is further configured to:

[0151] determining a target training mode for a second encoder based on the performance of the terminal;

[0152] According to the target training mode, a second encoder associated with each application scenario is obtained by training based on the first data set.

[0153] In conjunction with some embodiments of the tenth aspect, in some embodiments, the processing module is further configured to:

[0154] When the size of the available storage space of the terminal is greater than a size threshold, determining the target training mode to be training the second encoder associated with each application scenario based on the CSI associated with each application scenario and input data of the second decoder, wherein the second encoders associated with different application scenarios share some network parameters;

[0155] When the size of the available storage space of the terminal is less than or equal to the size threshold, the target training mode is determined to be training a second encoder associated with all application scenarios based on the CSI associated with all application scenarios and the input data of the second decoder.

[0156] In the eleventh aspect, an embodiment of the present disclosure proposes a communication device, which includes: one or more processors; wherein the processor is used to execute an optional implementation method of the method for generating a channel state information CSI feedback model proposed in the first aspect, the second aspect, the third aspect, and the fourth aspect.

[0157] In the twelfth aspect, an embodiment of the present disclosure proposes a communication system, which includes: a terminal and a network device; wherein the terminal is configured to execute the method described in the optional implementation of the second aspect and the fourth aspect, and the network device is configured to execute the method described in the optional implementation of the third aspect.

[0158] In the thirteenth aspect, an embodiment of the present disclosure proposes a storage medium, which stores instructions. When the instructions are executed on a communication device, the communication device executes the method described in the optional implementation of the first, second, third and fourth aspects.

[0159] In the fourteenth aspect, an embodiment of the present disclosure proposes a program product. When the program product is executed by a communication device, the communication device executes the method described in the optional implementation of the first aspect, the second aspect, the third aspect, and the fourth aspect.

[0160] In the fifteenth aspect, an embodiment of the present disclosure proposes a computer program, which, when executed on a computer, enables the computer to execute the method described in the optional implementation of the first, second, third, and fourth aspects.

[0161] In a sixteenth aspect, an embodiment of the present disclosure provides a chip or a chip system, which includes a processing circuit configured to execute the method described in the optional implementation of the first, second, third, and fourth aspects above.

[0162] It is understandable that the above-mentioned first device, terminal, network device, communication system, storage medium, program product, computer program, chip or chip system are all used to perform the method proposed in the embodiment of the present disclosure. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects of the corresponding method and will not be repeated here.

[0163] The present disclosure provides a method for generating a channel state information (CSI) feedback model. In some embodiments, the terms "channel state information (CSI) feedback model generation method," "measurement configuration method," "configuration method," and "communication method" are interchangeable. The terms "channel state information (CSI) feedback model generation device," "measurement configuration device," and "communication device" are interchangeable. The terms "channel state information (CSI) feedback model generation system," "measurement configuration system," and "communication system" are interchangeable.

[0164] The embodiments of the present disclosure are not exhaustive and are merely illustrative of some embodiments, and are not intended to be a specific limitation on the scope of protection of the present disclosure. In the absence of contradiction, each step in a certain embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a certain embodiment can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment can be arbitrarily exchanged. In addition, the optional implementation methods in a certain embodiment can be arbitrarily combined; in addition, the embodiments can be arbitrarily combined. For example, some or all steps of different embodiments can be arbitrarily combined, and a certain embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0165] In each embodiment of the present disclosure, unless otherwise specified or provided for by logic, the terms and / or descriptions between the embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form a new embodiment based on their inherent logical relationships.

[0166] The terms used in the embodiments of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0167] In the embodiments of the present disclosure, unless otherwise specified, elements expressed in the singular, such as "a", "an", "the", "above", "said", "the", "the", etc., may mean "one and only one", or "one or more", "at least one", etc. For example, when using articles such as "a", "an", "the" in English in translation, the noun following the article may be understood as a singular expression or a plural expression.

[0168] In the embodiments of the present disclosure, “plurality” refers to two or more.

[0169] In some embodiments, the terms "at least one," "one or more," "a plurality of," "multiple," etc. may be used interchangeably.

[0170] In some embodiments, descriptions such as "at least one of A and B," "A and / or B," "A in one case, B in another case," or "in response to one case A, in response to another case B" may include the following technical solutions depending on the situation: in some embodiments, A (A is executed independently of B); in some embodiments, B (B is executed independently of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); and in some embodiments, A and B (both A and B are executed). The above is also applicable when there are more branches such as A, B, and C.

[0171] In some embodiments, "A or B" and other descriptions may include the following technical solutions depending on the situation: in some embodiments, A (A is executed independently of B); in some embodiments, B (B is executed independently of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). The above is also applicable when there are more branches such as A, B, C, etc.

[0172] The prefixes such as "first" and "second" in the embodiments of the present disclosure are only used to distinguish different description objects and do not constitute any restriction on the position, order, priority, quantity or content of the description objects. For the statement of the description object, please refer to the description in the context of the claims or embodiments, and no unnecessary restriction should be constituted due to the use of prefixes. For example, if the description object is a "field", the ordinal number before the "field" in the "first field" and the "second field" does not limit the position or order between the "fields". "First" and "second" do not limit whether the "fields" they modify are in the same message, nor do they limit the order of the "first field" and the "second field". For another example, if the description object is a "level", the ordinal number before the "level" in the "first level" and the "second level" does not limit the priority between the "levels". For another example, the number of description objects is not limited by the ordinal number and can be one or more. Taking "first device" as an example, the number of "devices" can be one or more. In addition, the objects modified by different prefixes can be the same or different. For example, if the description object is "device", then the "first device" and the "second device" can be the same device or different devices, and their types can be the same or different; for another example, if the description object is "information", then the "first information" and the "second information" can be the same information or different information, and their contents can be the same or different.

[0173] In some embodiments, “including A,” “comprising A,” “used to indicate A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0174] In some embodiments, terms such as "in response to...", "in response to determining...", "in the case of...", "at the time of...", "when...", "if...", "if...", etc. can be used interchangeably.

[0175] In some embodiments, terms such as "greater than", "greater than or equal to", "not less than", "more than", "more than or equal to", "not less than", "higher than", "higher than or equal to", "not less than", and "above" can be replaced with each other, and terms such as "less than", "less than or equal to", "not greater than", "less than", "less than or equal to", "not more than", "lower than", "lower than or equal to", "not higher than", and "below" can be replaced with each other.

[0176] In some embodiments, devices and equipment can be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. In some cases, they can also be understood as "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "subject", etc.

[0177] In some embodiments, "network" can be interpreted as devices included in the network, such as access network equipment, core network equipment, etc.

[0178] In some embodiments, "access network device (AN device)" may also be referred to as "radio access network device (RAN device)", "base station (BS)", "radio base station", "fixed station", and in some embodiments may also be understood as "node", "access point", "transmission point (TP)", "reception point (RP)", "transmission and / or reception point (TRP)", "panel", "antenna panel", "antenna array", "cell", "macrocell", "smallcell", "femtocell", "picocell", "sector", "cell group", "serving cell", "carrier", "component carrier", "bandwidth part (BWP)", etc.

[0179] In some embodiments, "terminal" or "terminal device" may be referred to as "user equipment (UE)", "user terminal" "mobile station (MS)", "mobile terminal (MT)", subscriber station (subscriber station), mobile unit (mobile unit), subscriber unit (subscribe runit), wireless unit (wireless unit), remote unit (remote unit), mobile device (mobile device), wireless device (wireless device), wireless communication device (wireless communication device), remote device (remoted device), mobile subscriber station (mobile subscriber station), access terminal (access terminal), mobile terminal (mobile terminal), wireless terminal (wireless terminal), remote terminal (remote terminal), handset (handset), user agent (user agent), mobile client (mobile client), client (client), etc.

[0180] In some embodiments, obtaining data, information, etc. may comply with the laws and regulations of the country where the data is obtained.

[0181] In some embodiments, data, information, etc. may be obtained with the user's consent.

[0182] FIG1A is a schematic diagram showing the architecture of a communication system according to an embodiment of the present disclosure.

[0183] As shown in FIG1A , a communication system 1100 includes a first device 1101 , a terminal 1102 , and a network device 1103 .

[0184] In some embodiments, the first device 1101 may be a server, or may be the same device as the network device 1103 , which is not limited in this disclosure.

[0185] In some embodiments, the terminal 1102 includes, for example, a mobile phone, a wearable device, an Internet of Things device, a car with communication function, a smart car, a tablet computer, a computer with wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal device in industrial control, a wireless terminal device in self-driving, a wireless terminal device in remote medical surgery, a wireless terminal device in a smart grid, a wireless terminal device in transportation safety, a wireless terminal device in a smart city, and at least one of a wireless terminal device in a smart home, but is not limited thereto.

[0186] In some embodiments, the network device 1103 may include at least one of an access network device and a core network device.

[0187] In some embodiments, the access network device is, for example, a node or device that accesses a terminal to a wireless network. The access network device may include an evolved NodeB (eNB), a next generation evolved NodeB (ng-eNB), a next generation NodeB (gNB), a node B (NB), a home node B (HNB), a home evolved nodeB (HeNB), a wireless backhaul device, a radio network controller (RNC), a base station controller (BSC), a base transceiver station (BTS), a base band unit (BBU), a mobile switching center, a base station in a 6G communication system, an open base station (Open RAN), a cloud base station (Cloud RAN), a base station in other communication systems, and at least one of an access node in a Wi-Fi system, but is not limited thereto.

[0188] In some embodiments, the technical solution of the present disclosure can be applied to the Open RAN architecture. In this case, the interfaces between or within the access network devices involved in the embodiments of the present disclosure can be transformed into internal interfaces of the Open RAN, and the processes and information interactions between these internal interfaces can be implemented through software or programs.

[0189] In some embodiments, the access network device can be composed of a centralized unit (CU) and a distributed unit (DU), where the CU can also be called a control unit. The CU-DU structure can be used to split the protocol layer of the access network device, with the functions of some protocol layers centrally controlled by the CU, and the functions of the remaining part or all of the protocol layers distributed in the DU, which is centrally controlled by the CU, but is not limited to this.

[0190] In some embodiments, a core network device may be a device including one or more network elements, or may be multiple devices or device groups, each including all or part of the one or more network elements. The network element may be virtual or physical. The core network may include, for example, at least one of an Evolved Packet Core (EPC), a 5G Core Network (5GCN), and a Next Generation Core (NGC).

[0191] It can be understood that the communication system described in the embodiment of the present disclosure is for the purpose of more clearly illustrating the technical solution of the embodiment of the present disclosure, and does not constitute a limitation on the technical solution proposed in the embodiment of the present disclosure. Ordinary technicians in this field can know that with the evolution of the system architecture and the emergence of new business scenarios, the technical solution proposed in the embodiment of the present disclosure is also applicable to similar technical problems.

[0192] The following embodiments of the present disclosure may be applied to the communication system 1100 shown in FIG1A , or a portion thereof, but are not limited thereto. The entities shown in FIG1A are illustrative only. The communication system may include all or part of the entities shown in FIG1A , or may include other entities other than those shown in FIG1A . The number and form of the entities may be arbitrary. The entities may be physical or virtual. The connection relationships between the entities are illustrative only. The entities may be connected or disconnected, and the connection may be in any manner, including direct or indirect, wired or wireless.

[0193] The embodiments of the present disclosure can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G New Radio (NR), Future Radio Access (FRA), New Radio Access Technology (RAT), New Radio (NR), New Radio Access (NX), Future Generation Radio Access (FX), Global System for Mobile Communications (GSM (registered trademark)), CDMA2000, Ultra Mobile Broadband (Ultra Mobile Broadband), and other technologies. Broadband (UMB), IEEE802.11 (Wi-Fi (registered trademark)), IEEE802.16 (WiMAX (registered trademark)), IEEE802.20, Ultra-WideBand (UWB), Bluetooth (registered trademark), Public Land Mobile Network (PLMN) network, Device-to-Device (D2D) system, Machine-to-Machine (M2M) system, Internet of Things (IoT) system, Vehicle-to-Everything (V2X), systems using other communication methods, and next-generation systems based on them. In addition, multiple systems can also be combined (for example, a combination of LTE or LTE-A with 5G, etc.) for application.

[0194] The Channel State Information (CSI) feedback network structure is shown in Figure 1B, which is a schematic diagram of a CSI feedback model according to an embodiment of the present disclosure. The encoder model on the user equipment (UE) side completes feature extraction and dimensionality compression of the original CSI data, compressing the original CSI data into codewords with smaller data volume, and then transmits the codeword information to the base station (BS) via the uplink feedback link to fully reduce feedback overhead. The codeword received by the BS is reconstructed by the decoder model to output the CSI data of the original dimension.

[0195] Existing deep learning-based CSI feedback network technology solutions focus more on the two stages of model training and model inference. In the model training stage, a larger-scale training data set is used, and more training rounds are performed to obtain the feedback network model parameters. Due to different channel scenarios, different channel parameter configurations, or different base station coverage ranges (collectively referred to as different model application scenarios), the CSI data feature distribution may vary greatly. Therefore, existing technical solutions usually need to use data from various application scenarios to train and store different AI models for different application scenarios. In the model inference stage, that is, the application stage of the model in the actual system, encoder and decoder models for multiple different application scenarios will be deployed and stored in the UE and BS respectively to complete the compression and reconstruction of real-time CSI input data in different application scenarios, and then complete the CSI feedback work in large-scale multiple-input multiple-output (MIMO) systems.

[0196] To ensure CSI feedback accuracy, multiple AI models are used to complete CSI feedback tasks in different application scenarios. However, AI models typically have a large number of parameters, requiring a large amount of storage space to be allocated for multiple AI models in different application scenarios. Compared to base stations, user equipment (mobile terminals) have more limited storage resources, and the storage space available for AI model applications is even more limited.

[0197] Figure 2A is an interactive diagram illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in Figure 2A, the present disclosure embodiment relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101, a terminal 1102, and a network device 1103. The method includes:

[0198] Step S2101: The first device 1101 obtains a training data set.

[0199] In some embodiments, the first device 1101 may be a server.

[0200] In some embodiments, the training data set may include channel state information CSI in multiple application scenarios.

[0201] In some embodiments, terms such as "CSI", "Channel State Information", and "Channel State Information" can be used interchangeably.

[0202] In step S2102 , the first device 1101 trains an initial encoder and an initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0203] In some embodiments, the first device 1101 may first perform feature extraction and dimension compression on the CSI data for multiple application scenarios included in the training data set through an initial encoder to obtain corresponding codewords, and then perform CSI reconstruction through an initial decoder until a first encoder and a first decoder suitable for multiple application scenarios are obtained.

[0204] In step S2103, the first device 1101 keeps all parameters of the first encoder unchanged based on the CSI associated with each application scenario, and fine-tunes the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, keeps some parameters of the first encoder unchanged, and fine-tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0205] In some embodiments, different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters, etc., which is not limited in this disclosure.

[0206] In some embodiments, the first device 1101 may determine whether to keep all parameters of the first encoder unchanged or to keep some parameters of the first encoder unchanged based on performance requirements of a model in actual applications, etc., and this disclosure does not limit this.

[0207] In some embodiments, if it is determined that the first encoder can be used to perform feature extraction and dimensionality compression on CSI in all scenarios, that is, all parameters of the first encoder can remain unchanged. In this case, the CSI for each scenario can be input into the first encoder for feature extraction and dimensionality compression to obtain the corresponding codeword. The codeword output by the first encoder is then input into the first decoder. Based on the difference between the CSI output by the first decoder and the CSI input to the first encoder, only the parameters in the first decoder are fine-tuned, while all parameters of the first encoder remain unchanged, thereby obtaining a second decoder associated with each application scenario.

[0208] In some embodiments, the second encoder is an encoder that is more suitable for a certain application scenario obtained by updating some parameters in the first encoder, and can be used to perform feature extraction and dimensionality compression on the channel state information CSI in the scenario.

[0209] In some embodiments, when the CSI associated with different application scenarios is quite different, the model training and application method shared by the first encoder overall model may be difficult to meet the accuracy requirements of the CSI feedback in different application scenarios. At this time, it can be determined that some encoder parameters can be shared in different scenarios, that is, some parameters of the first encoder can be kept unchanged. At this time, the first device 1101 can first input the CSI associated with each application scenario into the first encoder for feature extraction and dimensionality compression to obtain the corresponding codeword, and then input the codeword output by the first encoder into the first decoder, and then fine-tune the first decoder and the first encoder based on the difference between the CSI output by the first decoder and the CSI input to the first encoder to obtain the second encoder and the second decoder associated with each application scenario.

[0210] It should be noted that when fine-tuning the first decoder and the first encoder, it is necessary to keep some network parameters in the first encoder unchanged, while only adjusting other parameters. For example, the parameters of the network layer used for feature extraction in the first encoder can be kept unchanged, while only the parameters of the network layer used for dimensionality compression can be fine-tuned; alternatively, the parameters of several network layers used for feature extraction in the first encoder can be kept unchanged, while only the parameters of the remaining network layers can be fine-tuned, etc. This disclosure is not limited to this.

[0211] In some embodiments, the part of the first encoder that keeps parameters unchanged can be called shared layers, which are applicable to all application scenarios, and the other part can be called adaptive layers, which are dedicated to different application scenarios. This disclosure does not limit this.

[0212] In some embodiments, while keeping some parameters of the first encoder unchanged, the first device 1101 can first separate the common layer in the first encoder and fix its parameters. Then, using the CSI associated with each application scenario, the first device 1101 can fine-tune the adaptive layer portion of the first encoder and the first decoder to obtain a second encoder and second decoder associated with each application scenario. The second encoder is composed of the common layer portion of the first encoder and the fine-tuned adaptive layer portion.

[0213] In step S2104 , when different application scenarios are associated with the same first encoder, the first device 1101 sends one first encoder to the terminal 1102 , and sends multiple second decoders to the network device 1103 .

[0214] In some embodiments, when the second encoders associated with different application scenarios are the same, it can be considered that each application scenario shares one encoder. In this case, the first device 1101 can send one first encoder to the terminal 1102.

[0215] In some embodiments, after receiving the second decoder sent by the first device 1101 , the network device 1103 may deploy the second decoder locally.

[0216] In step S2105, when the second encoders associated with different application scenarios share some network parameters, the first device 1101 sends the shared network parameters and another part of the network parameters associated with each application scenario to the terminal 1102, and sends multiple second decoders to the network device 1103.

[0217] In some embodiments, when the second encoders associated with different application scenarios share part of the network parameters, the other part of the second encoder is a part that does not share the network parameters.

[0218] In some embodiments, the multiple second decoders sent by the first device 1101 to the network device 1103 include a second decoder corresponding to each application scenario.

[0219] In some embodiments, when the second encoders associated with different application scenarios share some network parameters, it can be considered that another part of the network parameters in the second encoder is dedicated to different application scenarios. At this time, the first device 1101 can send the shared part of the network parameters and the other part of the network parameters associated with each application scenario to the terminal 1102.

[0220] In some embodiments, after receiving the second decoder sent by the first device 1101 , the network device 1103 may deploy the second decoder locally.

[0221] Step S2106: Terminal 1102 deploys a first encoder or multiple second encoders.

[0222] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S2101 to S2106. For example, steps S2101 and S2102 may be implemented as independent embodiments, and step S2103 may be implemented as an independent embodiment, etc., but the present invention is not limited thereto.

[0223] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0224] In this embodiment, the first device trains the encoder and decoder based on the CSI in different application scenarios in the training data set to obtain encoders and decoders suitable for different application scenarios. The obtained encoder and decoder are then fine-tuned for different application scenarios, so that the second encoders in different application scenarios share at least some network parameters. This reduces the storage overhead of the CSI feedback model in the terminal while ensuring the accuracy of the CSI feedback model.

[0225] FIG2B is an interactive diagram illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG2B , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and a terminal 1102, and includes:

[0226] Step S2201: The first device 1101 obtains a training data set.

[0227] In some embodiments, the first device 1101 may be a network device such as a base station, which is not limited in the present disclosure.

[0228] In step S2202 , the first device 1101 trains an initial encoder and an initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0229] In step S2203, the first device 1101 keeps all parameters of the first encoder unchanged based on the CSI associated with each application scenario, and fine-tunes the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, keeps some parameters of the first encoder unchanged, and fine-tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0230] Step S2204 : When different application scenarios are associated with the same first encoder, the first device 1101 sends a first encoder to the terminal 1102 .

[0231] In step S2205 , when the second encoders associated with different application scenarios share some network parameters, the first device 1101 sends the shared network parameters and another network parameter associated with each application scenario to the terminal 1102 .

[0232] For a detailed description of steps S2201 to S2205, reference may be made to steps S2101 to S2105 in the embodiment shown in FIG2A , which will not be repeated here.

[0233] In step S2206 , the first device 1101 deploys multiple second decoders locally.

[0234] In some embodiments, after obtaining multiple second decoders, the first device 1101 may deploy the multiple second decoders locally.

[0235] In step S2207 , the terminal 1102 deploys a first encoder or multiple second encoders.

[0236] For a detailed description of step S2207, please refer to step S2106 in the embodiment shown in FIG2A , which will not be repeated here.

[0237] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S2201 to S2207. For example, step S2201 may be implemented as an independent embodiment, step S2202 may be implemented as an independent embodiment, and steps S2201+S2202 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0238] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0239] In this embodiment, when the first device is a network device, the first device trains an encoder and a decoder based on the CSI in multiple application scenarios in the training data set to obtain encoders and decoders suitable for multiple application scenarios. The first device then fine-tunes the obtained encoder and decoder for different application scenarios. Finally, the first device deploys the fine-tuned decoder locally, thereby reducing the storage overhead of the CSI feedback model in the terminal and ensuring the accuracy of the CSI feedback model.

[0240] FIG2C is an interactive diagram illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG2C , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and a terminal 1102. The method includes:

[0241] Step S2301: The first device 1101 obtains a training data set.

[0242] In some embodiments, the first device 1101 may be a server, or a base station, etc., which is not limited in the present disclosure.

[0243] In step S2302 , the first device 1101 trains an initial encoder and an initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0244] In step S2303, the first device 1101 keeps all parameters of the first encoder unchanged based on the CSI associated with each application scenario, and fine-tunes the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, keeps some parameters of the first encoder unchanged, and fine-tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0245] In some embodiments, when the first device 1101 is a network device, after obtaining the second decoder associated with each application scenario, multiple decoders can be deployed locally.

[0246] For a detailed introduction to steps S2301-S2303, please refer to steps S2101-S2103 and steps S2201-S2203 in the embodiments shown in FIG. 2A and FIG. 2B, which will not be repeated here.

[0247] Step S2304: The terminal 1102 sends first indication information to the first device 1101.

[0248] In some embodiments, the first indication information may be used to indicate the performance of the terminal.

[0249] In some embodiments, the performance of the terminal may be any performance such as the size of the available storage space of the terminal, the data processing capability, etc., and this disclosure does not limit this.

[0250] In some embodiments, terminal capabilities may include the amount of storage space available in the terminal.

[0251] In some embodiments, after receiving the first indication information sent by the terminal 1102 , the first device 1101 may obtain the size of the available storage space of the terminal 1102 from the first indication information.

[0252] Step S2305: The first device 1101 determines the type of the second encoder corresponding to the terminal 1102 according to the first indication information.

[0253] In some embodiments, the type of the second encoder can be an encoder suitable for the application scenario used, or it can also be an encoder that shares some network parameters and uses another part of network parameters specifically for different application scenarios, etc. This disclosure does not limit this.

[0254] In step S2306 , the first device 1101 sends a second encoder applicable to all application scenarios to the terminal 1102 according to the type of the second encoder corresponding to the terminal 1102 , or sends multiple second encoders that share some network parameters to the terminal 1102 .

[0255] In some embodiments, if the type of the second encoder corresponding to the terminal 1102 is applicable to all application scenarios, the first device 1101 may send a second encoder applicable to all application scenarios to the terminal 1102 .

[0256] In some embodiments, if the type of the second encoder corresponding to the terminal 1102 is to share part of the network parameters, and the other part of the network parameters is dedicated to different application scenarios, then the first device 1101 can send multiple second encoders that share part of the network parameters to the terminal 1102.

[0257] In some embodiments, the terminal 1102 receives a second encoder sent by the first device 1101, or receives multiple second encoders sent by the first device 1101, wherein the multiple second encoders share some network parameters.

[0258] In step S2307 , the terminal 1102 deploys a second encoder or multiple second encoders.

[0259] In some embodiments, when a second encoder applicable to all application scenarios is obtained, the terminal may deploy the second encoder.

[0260] In some embodiments, when obtaining the second encoders respectively associated with each application scenario, the terminal may deploy the multiple second encoders.

[0261] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S2301 to S2307. For example, steps S2301 and S2302 may be implemented as independent embodiments, and step S2303 may be implemented as an independent embodiment, etc., but the present invention is not limited thereto.

[0262] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0263] In this embodiment, the first device trains the encoder and decoder based on the CSI in multiple application scenarios in the training data set to obtain encoders and decoders suitable for multiple application scenarios. The obtained encoder and decoder are then fine-tuned based on the CSI associated with different application scenarios. The terminal then indicates its own performance by sending indication information to the first device. Finally, the first device sends the corresponding encoder to the terminal, thereby reducing the storage overhead of the CSI feedback model in the terminal and improving the accuracy of the CSI feedback model.

[0264] FIG2D is an interactive diagram illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG2D , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used for terminal 1102 and network device 1103. The method includes:

[0265] In step S2401 , the network device 1103 obtains a training data set.

[0266] In some embodiments, the training data set may include channel state information CSI in multiple application scenarios.

[0267] For a detailed description of step S2401, please refer to step S2201 in the embodiment shown in FIG2B , which will not be repeated here.

[0268] In step S2402 , the network device 1103 trains an initial encoder and an initial decoder in the initial CSI feedback model based on the CSI associated with each application scenario, and obtains a second decoder and input data of the second decoder associated with each application scenario.

[0269] In some embodiments, the input data of the second decoder may be the CSI in each application scenario, which is subjected to feature extraction and dimension compression by a corresponding encoder to obtain a corresponding codeword.

[0270] In some embodiments, when the network device 1103 trains the initial encoder and the initial decoder in the initial CSI feedback model based on the CSI associated with each application scenario, the CSI associated with each application scenario is first feature extracted and dimensionally compressed through initial encoding to obtain the corresponding codeword, and then the codeword is input into the initial decoder as input data. Thereafter, based on the difference between the CSI output by the initial decoder and the CSI input into the initial encoder, the initial encoder and the initial decoder are trained respectively to obtain the second decoder and the input data of the second decoder associated with each application scenario.

[0271] In step S2403 , the network device 1103 locally deploys the second decoder associated with each application scenario.

[0272] For a detailed description of step S2403, please refer to step S2206 in the embodiment shown in FIG2B , which will not be repeated here.

[0273] In step S2404 , the network device 1103 sends the first data set associated with each application scenario to the terminal.

[0274] In some embodiments, the first data set may include CSI associated with each application scenario and input data of the second decoder.

[0275] In some embodiments, the first data set can be used to assist the terminal in training to obtain a second encoder associated with each application scenario.

[0276] In some embodiments, the second encoders associated with different application scenarios have the same or share some network parameters, which is not limited in this disclosure.

[0277] In some embodiments, after obtaining the input data of the second decoder associated with each application scenario, the network device 1103 may send the CSI associated with each application scenario and the input data of the second decoder to the terminal 1102 .

[0278] In some embodiments, the terminal 1102 receives a first data set associated with each application scenario sent by the network device 1103 .

[0279] Step S2405: Terminal 1102 determines a target training mode for the second encoder based on the performance of the terminal.

[0280] In some embodiments, the target training mode of the second encoder may be to train an encoder applicable to all application scenarios, or may be to train multiple encoders that share some network parameters, etc., which is not limited in this disclosure.

[0281] In some embodiments, when the size of the available storage space of the terminal is greater than a size threshold, the target training mode is determined to be based on the CSI associated with each application scenario and the input data of the second decoder, and the second encoder associated with each application scenario is trained, wherein the second encoders associated with different scenarios share some network parameters.

[0282] That is to say, if there is enough storage space on the terminal side to store the network parameters of multiple encoders, then a second encoder sharing some parameters can be trained for each application scenario.

[0283] In some embodiments, the scale threshold is a critical value of the scale of the terminal available storage space for determining the target training mode of the second encoder, which can be pre-set or determined according to model performance requirements, etc., and this disclosure does not limit this.

[0284] In some embodiments, when the size of the available storage space of the terminal is less than or equal to a size threshold, the target training mode is determined to be training a second encoder associated with all application scenarios based on the CSI associated with all application scenarios and the input data of the second decoder.

[0285] That is to say, the storage space on the terminal side is limited. If multiple encoders are stored at this time, the terminal performance may be reduced. At this time, a second encoder can be trained for multiple application scenarios.

[0286] In step S2406 , the terminal 1102 trains the first data set according to the target training mode to obtain a second encoder associated with each application scenario.

[0287] In some embodiments, when the target training mode of the second encoder determined by terminal 1102 is to train an encoder applicable to all application scenarios, terminal 1102 can train the encoder based on the received first training data set to obtain a second encoder applicable to all application scenarios.

[0288] In some embodiments, when the target training mode of the second encoder determined by the terminal 1102 is to train multiple encoders that share some network parameters, the terminal 1102 can first separate the common layer and the adaptive layer of the shared network parameters in the encoder, and then keep the parameters of the common layer unchanged, and use the first training data set it received to fine-tune the parameters of the adaptive layer of the encoder to obtain adaptive layers dedicated to different application scenarios, thereby obtaining a second encoder associated with each application scenario, wherein the second encoder includes a common layer and an adaptive layer dedicated to each application scenario.

[0289] In some embodiments, when training the encoder, the terminal 1102 first inputs the CSI into the encoder, and then after obtaining the output data of the encoder, compares the output data of the encoder with the input data of the decoder in the corresponding application scenario, and then can correct the encoder based on the comparison result.

[0290] In step S2407 , the terminal 1102 deploys a second encoder or multiple second encoders.

[0291] For a detailed description of step S2407, please refer to step S2307 in the embodiment shown in FIG2C , which will not be repeated here.

[0292] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S2401 to S2407. For example, steps S2401 and S2402 may be implemented as independent embodiments, and step S2402 may be implemented as an independent embodiment, etc., but the present invention is not limited thereto.

[0293] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0294] In this embodiment, the network device first trains the initial encoder and the initial decoder based on the CSI associated with each application scenario, obtains the second decoder associated with each application scenario and the input data of the second decoder, and sends the CSI associated with each application scenario and the input data of the second decoder to the terminal. Then, the terminal trains the encoder based on the CSI associated with each application scenario and the input data of the second decoder according to the target training mode of the encoder to obtain the second encoder associated with each application scenario, thereby improving the accuracy of the CSI feedback model and reducing the storage overhead of the CSI feedback model in the terminal.

[0295] FIG3A is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3A , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0296] Step S3101, obtain a training data set.

[0297] In some embodiments, the first device 1101 may be a server.

[0298] In some embodiments, the training data set may include channel state information CSI in multiple application scenarios.

[0299] Step S3102 : Training an initial encoder and an initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0300] In step S3103, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0301] Step S3104: When different application scenarios are associated with the same first encoder, one first encoder is sent to the terminal, and multiple second decoders are sent to the network device.

[0302] In some embodiments, the terminal 1102 receives a first encoder sent by the first device 1101.

[0303] In some embodiments, the network device 1103 receives multiple second decoders sent by the first device 1101 .

[0304] Step S3105 , when the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal 1102 , and multiple second decoders are sent to the network device 1103 .

[0305] In some embodiments, the terminal 1102 receives a portion of shared network parameters sent by the first device 1101 and another portion of network parameters associated with each application scenario.

[0306] In some embodiments, the network device 1103 receives multiple second decoders sent by the first device 1101 .

[0307] In some embodiments, after receiving the second decoder sent by the first device 1101 , the network device 1103 may deploy the second decoder locally.

[0308] For a detailed description of steps S3101 to S3105, please refer to steps S2101 to S2105 in the embodiment shown in FIG2A , which will not be repeated here.

[0309] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3101 to S3105. For example, step S3101 may be implemented as an independent embodiment, step S3102 may be implemented as an independent embodiment, and steps S3101+S3102 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0310] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0311] In this embodiment, the first device trains an initial encoder and an initial decoder based on a training data set to obtain an encoder and a decoder suitable for all application scenarios. Then, based on the CSI associated with different application scenarios, the parameters of the obtained decoder are fine-tuned so that the second encoders in different application scenarios share at least some network parameters. This reduces the storage overhead of the CSI feedback model in the terminal while ensuring the accuracy of the CSI feedback model.

[0312] FIG3B is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3B , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0313] Step S3201: Obtain a training data set.

[0314] In some embodiments, the first device 1101 may be a network device such as a base station.

[0315] Step S3202: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0316] In step S3203, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0317] Step S3204 : When different application scenarios are associated with the same first encoder, send the first encoder to the terminal 1102 .

[0318] In some embodiments, the terminal 1102 receives a first encoder sent by the first device 1101.

[0319] Step S3205 : When the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal 1102 .

[0320] In some embodiments, the terminal 1102 receives a portion of shared network parameters sent by the first device 1101 and another portion of network parameters associated with each application scenario.

[0321] Step S3206: deploy multiple second decoders locally.

[0322] For a detailed description of steps S3201 to S3206, reference may be made to steps S2201 to S2206 in the embodiment shown in FIG2B , which will not be repeated here.

[0323] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3201 to S3206. For example, step S3201 may be implemented as an independent embodiment, step S3202 may be implemented as an independent embodiment, and steps S3201+S3202 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0324] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0325] In this embodiment, when the first device is a network device, an initial encoder and an initial decoder are trained based on a training data set to obtain encoders and decoders suitable for all application scenarios. Then, based on the CSI associated with different application scenarios, the parameters of the obtained decoder are fine-tuned to obtain decoders dedicated to different application scenarios. Thereafter, the first device sends the encoder to the terminal and deploys the decoder locally, thereby reducing the storage overhead of the CSI feedback model in the terminal and ensuring the accuracy of the CSI feedback model.

[0326] FIG3C is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3C , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0327] Step S3301, obtain a training data set.

[0328] In some embodiments, the first device 1101 may be a server, or may also be a base station, etc.

[0329] Step S3302: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0330] In step S3303, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0331] Step S3304: receiving first indication information sent by the terminal.

[0332] In some embodiments, the first indication information is used to indicate the performance of the terminal.

[0333] In some embodiments, the first device 1101 receives the first indication information sent by the terminal 1102

[0334] Step S3305: Determine the type of the second encoder corresponding to the terminal according to the first indication information.

[0335] Step S3306 : According to the type of the second encoder corresponding to the terminal 1102 , a second encoder applicable to all application scenarios is sent to the terminal 1102 , or multiple second encoders that share some network parameters are sent to the terminal 1102 .

[0336] For a detailed description of steps S3301 to S3306, please refer to steps S2301 to S2306 in the embodiment shown in FIG2C , which will not be repeated here.

[0337] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3301 to S3306. For example, step S3301 may be implemented as an independent embodiment, step S3302 may be implemented as an independent embodiment, and steps S3301+S3302 may be implemented as independent embodiments, but the present disclosure is not limited thereto.

[0338] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0339] In this embodiment, the first device first trains the encoder and decoder based on the training data set to obtain the encoder and decoder associated with each application scenario, and then sends the corresponding encoder to the terminal based on the indication information sent by the terminal, thereby improving the accuracy of the CSI feedback model.

[0340] FIG3D is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3D , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0341] Step S3401: Obtain a training data set.

[0342] In some embodiments, the first device 1101 may be a server.

[0343] In some embodiments, the training data set may include channel state information CSI in multiple application scenarios.

[0344] Step S3402: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0345] In step S3403, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0346] Step S3404 , when different application scenarios are associated with the same first encoder, one first encoder is sent to the terminal 1102 , and multiple second decoders are sent to the network device 1103 .

[0347] In some embodiments, the terminal 1102 receives a first encoder sent by the first device 1101.

[0348] In some embodiments, the network device 1103 receives multiple second decoders sent by the first device 1101 .

[0349] For a detailed description of steps S3401 - S3404 , please refer to steps S3101 - S3104 in the embodiment shown in FIG3A , which will not be repeated here.

[0350] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3401 to S3404. For example, step S3401 may be implemented as an independent embodiment, step S3402 may be implemented as an independent embodiment, and steps S3401+S3402 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0351] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0352] In this embodiment, the first device trains the encoder and decoder based on the training data set to obtain the encoder and decoder associated with each application scenario. When the second encoders associated with different application scenarios are the same, the first device sends one encoder to the terminal and sends multiple decoders to the network device, thereby reducing the storage overhead of the CSI feedback model at the terminal.

[0353] FIG3E is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3E , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0354] Step S3501: Obtain a training data set.

[0355] In some embodiments, the first device 1101 may be a server.

[0356] In some embodiments, the training data set may include channel state information CSI in multiple application scenarios.

[0357] Step S3502: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0358] In step S3503, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0359] For a detailed description of steps S3501-S3503, please refer to steps S3101-S3103 in the embodiment shown in FIG3A, which will not be repeated here.

[0360] Step S3504: When the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal 1102, and multiple second decoders are sent to the network device 1103.

[0361] In some embodiments, the terminal 1102 receives a portion of shared network parameters sent by the first device 1101 and another portion of network parameters associated with each application scenario.

[0362] In some embodiments, the network device 1103 receives multiple second decoders sent by the first device 1101 .

[0363] In some embodiments, after receiving the second decoder sent by the first device 1101 , the network device 1103 may deploy the second decoder locally.

[0364] For a detailed description of step S3504, please refer to step S3105 in the embodiment shown in FIG3A , which will not be repeated here.

[0365] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3501 to S3504. For example, step S3501 may be implemented as an independent embodiment, step S3502 may be implemented as an independent embodiment, and steps S3501+S3502 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0366] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0367] In this embodiment, the first device trains the encoder and decoder based on the training data set to obtain the encoder and decoder associated with each application scenario. When the second encoders associated with different application scenarios share some network parameters, the first device sends the shared network parameters and another part of the network parameters associated with each application scenario to the terminal, and sends multiple second decoders to the network device, thereby reducing the storage overhead of the CSI feedback model in the terminal.

[0368] FIG3F is a flow chart illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3F , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0369] Step S3601: Obtain a training data set.

[0370] In some embodiments, the first device 1101 may be a network device such as a base station.

[0371] Step S3602: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0372] In step S3603, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0373] Step S3604 : When different application scenarios are associated with the same first encoder, send the first encoder to the terminal 1102 .

[0374] In some embodiments, the terminal 1102 receives a first encoder sent by the first device 1101.

[0375] For a detailed description of steps S3601 to S3604, please refer to steps S3201 to S3204 in the embodiment shown in FIG3B , which will not be repeated here.

[0376] Step S3605: deploy multiple second decoders locally.

[0377] For a detailed description of step S3605, please refer to step S3206 in the embodiment shown in FIG3B , which will not be repeated here.

[0378] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3601 to S3605. For example, step S3601 may be implemented as an independent embodiment, step S3602 may be implemented as an independent embodiment, and steps S3601+S3602 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0379] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0380] In this embodiment, when the first device is a network device, the encoder and decoder are trained based on a training data set to obtain an encoder and decoder associated with each application scenario. When the second encoders associated with different application scenarios are the same, the first device sends an encoder to the terminal, and then the first device locally deploys the decoder associated with each application scenario, thereby improving the working efficiency of the communication system and ensuring the accuracy of the CSI feedback model.

[0381] FIG3G is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3G , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0382] Step S3701: Obtain a training data set.

[0383] In some embodiments, the first device 1101 may be a network device such as a base station.

[0384] Step S3702: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0385] In step S3703, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0386] For a detailed description of steps S3701-S3703, please refer to steps S3201-S3203 in the embodiment shown in FIG3B , which will not be repeated here.

[0387] Step S3704 : When the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal 1102 .

[0388] In some embodiments, the terminal 1102 receives a portion of shared network parameters sent by the first device 1101 and another portion of network parameters associated with each application scenario.

[0389] Step S3705: deploy multiple second decoders locally.

[0390] For a detailed description of steps S3704-S3705, please refer to steps S3205-S3206 in the embodiment shown in FIG3B , which will not be repeated here.

[0391] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3701 to S3705. For example, step S3701 may be implemented as an independent embodiment, step S3702 may be implemented as an independent embodiment, and steps S3701+S3702 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0392] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0393] In this embodiment, when the first device is a network device, the encoder and decoder are trained based on a training data set to obtain an encoder and decoder associated with each application scenario. When the second encoders associated with different application scenarios share some network parameters, the first device sends the shared network parameters and another part of the network parameters associated with each application scenario to the terminal, thereby improving the working efficiency of the communication system and ensuring the accuracy of the CSI feedback model.

[0394] FIG3H is a flow chart illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG3H , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and includes:

[0395] Step S3801: Obtain a training data set.

[0396] In some embodiments, the first device 1101 may be a server, a base station, or the like.

[0397] In some embodiments, the training data set may include channel state information CSI in multiple application scenarios.

[0398] Step S3802: Train the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0399] In step S3803, based on the CSI associated with each application scenario, all parameters of the first encoder are kept unchanged, and the first decoder is fine-tuned to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, some parameters of the first encoder are kept unchanged, and the first decoder and the first encoder are fine-tuned respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0400] In some embodiments, after obtaining the second encoder and the second decoder associated with each application scenario, the method further includes:

[0401] In the case where the same first encoder is associated with different application scenarios, one first encoder is sent to the terminal, and multiple second decoders are sent to the network device; or

[0402] When the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal, and multiple second decoders are sent to the network device.

[0403] In some embodiments, the first device is a network device, and after obtaining the second encoder and the second decoder associated with each application scenario, further includes:

[0404] In the case where different application scenarios are associated with the same first encoder, one first encoder is sent to the terminal; or

[0405] In the case that the second encoders associated with different application scenarios share some network parameters, the shared network parameters and another part of network parameters associated with each application scenario are sent to the terminal.

[0406] In some embodiments, it further includes:

[0407] Receiving first indication information sent by a terminal, wherein the first indication information is used to indicate the size of available storage space of the terminal;

[0408] Determining, according to the first indication information, a type of a second encoder corresponding to the terminal;

[0409] According to the type of the second encoder corresponding to the terminal, one second encoder applicable to all application scenarios is sent to the terminal, or multiple second encoders sharing some network parameters are sent to the terminal.

[0410] For a detailed description of steps S3801-S3803, please refer to the above embodiment description.

[0411] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S3801 to S3803. For example, step S3801 may be implemented as an independent embodiment, step S3802 may be implemented as an independent embodiment, and steps S3801+S3802 may be implemented as independent embodiments, but the present disclosure is not limited thereto.

[0412] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0413] In this embodiment, the first device trains the encoder and decoder based on a training data set to obtain encoders and decoders suitable for different application scenarios. The obtained encoder and decoder are then fine-tuned for different application scenarios, so that the second encoders in different application scenarios share at least some network parameters. This reduces the storage overhead of the CSI feedback model in the terminal while ensuring the accuracy of the CSI feedback model.

[0414] FIG4A is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4A , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in terminal 1102 and includes:

[0415] Step S4101: When different application scenarios are associated with the same first encoder, a first encoder sent by the first device 1101 is received.

[0416] Step S4102 , when the second encoders associated with different application scenarios share some network parameters, receives the shared part of network parameters sent by the first device 1101 and another part of network parameters associated with each application scenario.

[0417] Step S4103: deploy a first encoder or multiple second encoders.

[0418] For a detailed description of steps S4101-S4103, reference may be made to steps S2104-S2106 and steps S2204-S2206 in the embodiments shown in FIG. 2A and FIG. 2B, which will not be repeated here.

[0419] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4101 to S4103. For example, step S4101 may be implemented as an independent embodiment, step S4102 may be implemented as an independent embodiment, and steps S4101+S4102 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0420] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0421] In this embodiment, the terminal improves the accuracy of the CSI feedback model by receiving the corresponding encoder sent by the first device.

[0422] FIG4B is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4B , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model for terminal 1102, the method comprising:

[0423] Step S4201: Send first indication information to the first device 1101.

[0424] In some embodiments, the first indication information is used to indicate the performance of the terminal.

[0425] For a detailed description of step S4201, please refer to step S2304 in the embodiment shown in FIG2C , which will not be repeated here.

[0426] Step S4202: Receive a second encoder applicable to all application scenarios, or multiple second encoders that share some network parameters, sent by the first device 1101.

[0427] Step S4203: deploy a second encoder or multiple second encoders.

[0428] For a detailed description of steps S4202 to S4203, please refer to steps S2306 to S2307 in the embodiment shown in FIG2C , which will not be repeated here.

[0429] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4201 to S4203. For example, step S4201 may be implemented as an independent embodiment, step S4202 may be implemented as an independent embodiment, and steps S4201+S4202 may be implemented as independent embodiments, but are not limited thereto.

[0430] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0431] In this embodiment, the terminal indicates the terminal performance by sending indication information to the first device, and then receives the encoder sent by the first device, thereby reducing the storage overhead of the CSI feedback model in the terminal and improving the accuracy of the determined encoder.

[0432] FIG4C is a flow chart illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4C , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in terminal 1102 and includes:

[0433] Step S4301: Receive a first data set associated with each application scenario sent by the network device 1103.

[0434] In some embodiments, the first data set may include CSI associated with each application scenario and input data of the second decoder.

[0435] Step S4302 : Determine a target training mode for the second encoder based on the performance of the terminal 1102 .

[0436] Step S4303 : According to the target training mode, a second encoder associated with each application scenario is obtained by training based on the first data set.

[0437] Step S4304: deploy a second encoder or multiple second encoders.

[0438] For a detailed description of steps S4301-S4304, please refer to steps S2404-S2407 in the embodiment shown in FIG2D, which will not be repeated here.

[0439] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4301 to S4304. For example, step S4301 may be implemented as an independent embodiment, step S4302 may be implemented as an independent embodiment, and steps S4301+S4302 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0440] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0441] In this embodiment, after receiving the first data set associated with each application scenario sent by the network device, the terminal can determine the target training mode of the encoder based on its own performance, and then train the encoder associated with each application scenario based on the first data set according to the target training mode, thereby reducing the storage overhead of the encoder in the CSI feedback model at the terminal.

[0442] FIG4D is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4D , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model for terminal 1102, the method comprising:

[0443] Step S4401: When different application scenarios are associated with the same first encoder, a first encoder sent by the first device 1101 is received.

[0444] For a detailed description of step S4401, please refer to step S4101 in the embodiment shown in FIG4A , which will not be repeated here.

[0445] Step S4402: deploy a first encoder.

[0446] For a detailed description of step S4402, please refer to step S4103 in the embodiment shown in FIG4A , which will not be repeated here.

[0447] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4401 and S4402. For example, step S4401 may be implemented as an independent embodiment, step S4402 may be implemented as an independent embodiment, and steps S4401+S4402 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0448] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0449] In this embodiment, when the second encoders associated with different application scenarios are the same, the terminal improves the accuracy of the CSI feedback model by receiving the encoders applicable to all application scenarios sent by the first device.

[0450] FIG4E is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4E , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in terminal 1102 and includes:

[0451] Step S4501: When the second encoders associated with different application scenarios share some network parameters, receive the shared part of network parameters sent by the first device 1101 and another part of network parameters associated with each application scenario.

[0452] Step S4502: deploy multiple second encoders.

[0453] For a detailed description of steps S4501 to S4502S, please refer to steps S4102 to S4103 in the embodiment shown in FIG4A , which will not be repeated here.

[0454] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4501 and S4502. For example, step S4501 may be implemented as an independent embodiment, step S4502 may be implemented as an independent embodiment, and steps S4501+S4502 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0455] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0456] In this embodiment, when the second encoders associated with different application scenarios share some network parameters, the terminal improves the accuracy of the CSI feedback model by receiving the shared network parameters sent by the first device and another part of the network parameters associated with each application scenario.

[0457] FIG4F is a flow chart illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4F , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model for terminal 1102, the method comprising:

[0458] Step S4601: Receive a first encoder or multiple second encoders sent by a first device, where the multiple second encoders share some network parameters.

[0459] In some embodiments, the method further comprises:

[0460] Sending first indication information to the first device, where the first indication information is used to indicate performance of the terminal;

[0461] A second encoder is received from the first device, or multiple second encoders are received from the first device, where the multiple second encoders share some network parameters, and the second encoders are used to encode channel state information CSI.

[0462] In some embodiments, the terminal capabilities include the amount of storage space available in the terminal.

[0463] Step S4602: deploy a first encoder or multiple second encoders.

[0464] For a detailed description of steps S4601-S4602, please refer to the above embodiment description.

[0465] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4601 and S4602. For example, step S4601 may be implemented as an independent embodiment, step S4602 may be implemented as an independent embodiment, and steps S4601+S4602 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0466] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0467] In this embodiment, the terminal receives and deploys the encoder sent by the first device, thereby reducing the storage overhead of the CSI feedback model in the terminal and improving the accuracy of the determined encoder.

[0468] FIG4G is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG4G , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model for terminal 1102, the method comprising:

[0469] Step S4701 : receiving a first data set associated with each application scenario sent by the network device 1103 .

[0470] In some embodiments, the data set includes channel state information CSI associated with each application scenario and input data of the second decoder.

[0471] Step S4702: obtaining a second encoder associated with each application scenario through training based on the first data set.

[0472] In some embodiments, the second encoders associated with different application scenarios have the same or share some network parameters.

[0473] In some embodiments, training a second encoder associated with each application scenario based on the first data set includes:

[0474] determining a target training mode for the second encoder based on the performance of the terminal;

[0475] According to the target training mode, a second encoder associated with each application scenario is obtained by training based on the first data set.

[0476] In some embodiments, determining a target training mode for the second encoder based on the performance of the terminal includes:

[0477] When the size of the available storage space of the terminal is greater than a size threshold, determining a target training mode to train the second encoder associated with each application scenario based on the CSI associated with each application scenario and input data of the second decoder, wherein the second encoders associated with different application scenarios share some network parameters;

[0478] When the size of the available storage space of the terminal is less than or equal to the size threshold, the target training mode is determined to be training a second encoder associated with all application scenarios based on the CSI associated with all application scenarios and the input data of the second decoder.

[0479] For a detailed description of steps S4701-S4702, please refer to the above embodiment description.

[0480] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S4701 and S4702. For example, step S4701 may be implemented as an independent embodiment, step S4702 may be implemented as an independent embodiment, and steps S4701+S4702 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0481] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0482] In this embodiment, after receiving the first data set associated with each application scenario sent by the network device, the terminal obtains an encoder associated with each application scenario based on training of the first data set, thereby reducing the storage overhead of the encoder in the CSI feedback model at the terminal.

[0483] FIG5A is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG5A , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a network device 1103. The method includes:

[0484] Step S5101 , receiving multiple second decoders sent by the first device 1101 .

[0485] For a detailed description of step S5101, please refer to steps S2104 and S2105 in the embodiment shown in FIG2A , which will not be repeated here.

[0486] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0487] In this embodiment, the network device improves the accuracy and reliability of the CSI feedback model by receiving multiple second decoders sent by the first device.

[0488] FIG5B is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG5B , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model for use in network device 1103. The method includes:

[0489] Step S5201: Obtain a training data set.

[0490] In some embodiments, the training data set includes channel state information CSI in multiple application scenarios.

[0491] Step S5202 : Based on the CSI associated with each application scenario, the initial encoder and the initial decoder in the initial CSI feedback model are trained to obtain the second decoder and input data of the second decoder associated with each application scenario.

[0492] Step S5203: deploy the second decoder associated with each application scenario locally.

[0493] Step S5204: Send the first data set associated with each application scenario to the terminal.

[0494] In some embodiments, the first data set may include CSI associated with each application scenario and input data of the second decoder.

[0495] For a detailed description of steps S5201 to S5204, please refer to steps S2401 to S2404 in the embodiment shown in FIG2D , which will not be repeated here.

[0496] The method for generating a channel state information (CSI) feedback model according to the embodiments of the present disclosure may include at least one of steps S5201 to S5204. For example, step S5201 may be implemented as an independent embodiment, step S5202 may be implemented as an independent embodiment, and steps S5201+S5202 may be implemented as independent embodiments, but the present invention is not limited thereto.

[0497] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0498] In this embodiment, after obtaining a training data set, the network device trains the encoder and decoder based on the training data set to obtain an encoder and decoder associated with each application scenario, and then deploys the obtained decoder locally. Thereafter, the CSI associated with each application scenario and the obtained decoder input data are sent to the terminal, thereby providing conditions for reducing the storage overhead of the CSI feedback model in the terminal.

[0499] FIG5C is a flow chart of a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG5C , the embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model for network device 1103, the method comprising:

[0500] Step S5301: Obtain a training data set.

[0501] In some embodiments, the training data set includes channel state information CSI in multiple application scenarios.

[0502] Step S5302: Based on the CSI associated with each application scenario, the initial encoder and the initial decoder in the initial CSI feedback model are trained to obtain the second decoder and input data of the second decoder associated with each application scenario.

[0503] Step S5303: Send the first data set associated with each application scenario to the terminal.

[0504] In some embodiments, the first data set includes CSI and input data of the second decoder.

[0505] In some embodiments, the first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario.

[0506] In some embodiments, the second encoders associated with different application scenarios have the same or share some network parameters.

[0507] For a detailed description of steps S5301-S5303, please refer to the above embodiment description.

[0508] The method for generating a channel state information (CSI) feedback model involved in the embodiments of the present disclosure may include at least one of steps S5301 to S5303. For example, step S5301 may be implemented as an independent embodiment, step S5302 may be implemented as an independent embodiment, and steps S5301+S5302 may be implemented as independent embodiments, but are not limited thereto.

[0509] In the embodiments of the present disclosure, some or all of the steps and their optional implementations may be arbitrarily combined with some or all of the steps in other embodiments, or may be arbitrarily combined with the optional implementations of other embodiments.

[0510] In this embodiment, after obtaining a training data set, the network device trains the encoder and decoder based on the training data set to obtain an encoder and decoder associated with each application scenario. The CSI associated with each application scenario and the obtained decoder input data are then sent to the terminal, thereby providing conditions for reducing the storage overhead of the CSI feedback model at the terminal.

[0511] FIG6A is an interactive diagram illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG6A , an embodiment of the present disclosure relates to a method for generating a channel state information (CSI) feedback model, which is used in a first device 1101 and a terminal 1102. The method includes:

[0512] Step S6101: The first device 1101 obtains a training data set.

[0513] In some embodiments, the training data set includes channel state information CSI in multiple application scenarios.

[0514] In step S6102 , the first device 1101 trains an initial encoder and an initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder.

[0515] In step S6103, the first device 1101 keeps all parameters of the first encoder unchanged based on the CSI associated with each application scenario, and fine-tunes the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, keeps some parameters of the first encoder unchanged, and fine-tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario.

[0516] In some embodiments, different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters.

[0517] Step S6104 : The first device 1101 sends a first encoder or multiple second encoders to the terminal 1102 .

[0518] In some embodiments, the plurality of second encoders share some network parameters.

[0519] In some embodiments, the terminal 1102 receives a first encoder or multiple second encoders sent by the first device.

[0520] In some embodiments, the plurality of second encoders share some network parameters.

[0521] Step S6105: Terminal 1102 deploys a first encoder or multiple second encoders.

[0522] For a detailed description of steps S6101 - S6105 , please refer to the above embodiment description.

[0523] In the disclosed embodiment, a first device trains an encoder and a decoder based on the CSI in different application scenarios in a training data set to obtain encoders and decoders suitable for different application scenarios. The obtained encoders and decoders are then fine-tuned for different application scenarios, so that the second encoders in different application scenarios share at least some network parameters. This reduces the storage overhead of the CSI feedback model in the terminal while ensuring the accuracy of the CSI feedback model.

[0524] FIG6B is an interactive diagram illustrating a method for generating a channel state information (CSI) feedback model according to an embodiment of the present disclosure. As shown in FIG6B , the present disclosure embodiment relates to a method for generating a channel state information (CSI) feedback model for a communication system including a terminal 1102 and a network device 1103. The method includes:

[0525] In step S6201 , the network device 1103 obtains a training data set.

[0526] In some embodiments, the training data set includes channel state information CSI in multiple application scenarios.

[0527] In step S6202 , the network device 1103 trains an initial encoder and an initial decoder in the initial CSI feedback model based on the CSI associated with each application scenario, and obtains a second decoder and input data of the second decoder associated with each application scenario.

[0528] In step S6203 , the network device 1103 sends the first data set associated with each application scenario to the terminal 1101 .

[0529] In some embodiments, the first data set includes CSI and input data of the second decoder.

[0530] Step S6204: Terminal 1101 obtains a second encoder associated with each application scenario through training based on the first data set.

[0531] In some embodiments, the second encoders associated with different application scenarios have the same or share some network parameters.

[0532] For a detailed description of steps S6201-S6204, please refer to the above embodiment description.

[0533] In an embodiment of the present disclosure, after training the encoder and decoder based on the training data set, the network device obtains the decoder and decoder input data associated with each application scenario. The network device then sends the CSI associated with each application scenario and the decoder input data to the terminal. The terminal then trains based on the CSI associated with each application scenario and the decoder input data to obtain the decoder associated with each application scenario, thereby reducing the storage overhead of the CSI feedback model in the terminal.

[0534] The following is an exemplary introduction to the above method.

[0535] The present disclosure is used to solve the problem of how to reduce the memory overhead of user equipment (UE) when generating a CSI feedback model. The optional implementation solutions are as follows:

[0536] The present disclosure relates to a method for generating a channel state information (CSI) feedback model. Taking a base station as an example, the method includes:

[0537] (1) Method 1: Model training and application method shared by the entire encoder model

[0538] In order to reduce the amount of encoder model parameters that need to be stored in user equipment in multiple application scenarios, the present disclosure proposes a method for sharing the overall encoder model, that is, only one encoder model is used to complete CSI data compression in different application scenarios, and then the compressed codeword is transmitted to the base station (BS) via the uplink, and the CSI data is reconstructed using dedicated decoder models in different application scenarios.

[0539] The model training process of method 1 is as follows:

[0540] a. In the case that there are N application scenarios in the actual system application, obtain the model training data set Data for each application scenario i ,i=1,2,…N;

[0541] b. By using the dataset Data under N application scenarios i , i = 1, 2, ... N, to obtain the encoder model encoder and decoder model decoder, where encoder is a common encoder model for different application scenarios and is directly deployed on the UE;

[0542] c. Fix the parameters of the encoder model encoder, use the model training data under N application scenarios to fine-tune the parameters of the decoder model, and obtain the dedicated decoder model decoder under the i-th application scenario i ,i=1,2,…,N, and deployed at BS.

[0543] The model application structure of method 1 is shown in Figure 7A, which is a structural diagram of the encoder overall model sharing method according to an embodiment of the present disclosure. The specific model application process is: the original CSI data is directly compressed into codewords by the shared encoder model encoder deployed on the UE, and then fed back to the BS via the uplink, and the received codewords are input into the decoder model decoder in its corresponding application scenario. i , i=1,2,…,N performs data reconstruction to obtain reconstructed CSI data.

[0544] (2) Method 2: Model training and application method shared by some encoder models

[0545] When CSI data distribution varies significantly across different application scenarios, a model training and application method that uses a shared encoder model may not be able to meet the high feedback accuracy requirements across multiple application scenarios. Therefore, this disclosure proposes a model training and application method that uses a shared encoder model. This method enables the sharing of some layers in the encoder network model across different application scenarios. Specifically, some layers in the encoder are shared layers, while others are adaptive layers. The adaptive layers and the decoder are each dedicated to different application scenarios.

[0546] The model training process of method 2 is as follows:

[0547] a. In the case that there are N application scenarios in the actual system application, obtain the model training data set Data for each application scenario i ,i=1,2,…N;

[0548] b. Obtain the encoder model encoder and decoder model decoder by training with a mixed dataset consisting of datasets from N application scenarios;

[0549] c. Separate the shared layers in the encoder and fix their model parameters, and use the model training data of N application scenarios respectively i , i=1,2,…N, fine-tune the model parameters of the encoder adaptive layers and decoder decoder to obtain adaptive layers in N application scenarios i ,i=1,2,…,N,shared layers, adaptivelayers in N application scenarios i,i=1,2,…,N models are deployed on UE, decoder models in N application scenarios i ,i=1,2,…,N are deployed at BS.

[0550] When applying a CSI feedback solution system based on deep learning, different specific model training methods may be considered. In some cases, the method proposed in this disclosure may have different specific implementation processes. For example, for the following methods: the BS first uses training data sets under different application scenarios for training, obtains the decoder model and deploys it locally, and then passes the training data sets that match the trained decoder model under each application scenario to the UE. The UE uses the received data sets to complete the encoder model training under different application scenarios. At this time, the model training process of method 2 is specifically as follows:

[0551] a. In the case that there are N application scenarios in the actual system application, obtain the model training data set Data for each application scenario i ,i=1,2,…N;

[0552] b.BS trains decoder models for N application scenarios by using a mixed dataset composed of datasets for N application scenarios i ,i=1,2,…,N, and deployed locally;

[0553] c. The BS sends the datasets matching the trained decoder model in N application scenarios to the UE;

[0554] d. The UE obtains an encoder model encoder by training using the received mixed data sets under N application scenarios;

[0555] e.UE separates the shared layers and adaptive layers in the encoder, fixes the model parameters of the shared layers, and uses the model training data of the N application scenarios it receives to fine-tune the model parameters of the encoder adaptive layers, obtaining the adaptive layers under the N application scenarios. i ,i=1,2,…,N, the shared layers and adaptive layers are deployed locally.

[0556] The model application structure of method 2 is shown in Figure 7B, which is a structural diagram of the encoder partial model sharing method according to an embodiment of the present disclosure. The specific model application process is as follows: the original CSI data is compressed into codewords in turn through the shared layers of the encoder deployed on the UE and the adaptive layers in the application scenario corresponding to the CSI data. After being fed back to the BS via the uplink, the BS uses the decoder model decoder in the corresponding application scenario to receive the received codewords. i , i=1,2,…,N performs data reconstruction to obtain reconstructed CSI data.

[0557] The embodiments of the present disclosure further provide an apparatus for implementing any of the above methods. For example, an apparatus is provided, comprising units or modules for implementing each step performed by a terminal in any of the above methods. For another example, another apparatus is provided, comprising units or modules for implementing each step performed by a network device (e.g., a RAN) in any of the above methods.

[0558] It should be understood that the division of the various units or modules in the above device is merely a division of logical functions. In actual implementation, they may be fully or partially integrated into a physical entity, or they may be physically separated. In addition, the units or modules in the device may be implemented in the form of a processor calling software: for example, the device includes a processor, the processor is connected to a memory, and the memory stores instructions. The processor calls the instructions stored in the memory to implement any of the above methods or implement the functions of the various units or modules of the above device, wherein the processor is, for example, a general-purpose processor, such as a central processing unit (CPU) or a microprocessor, and the memory is a memory within the device or a memory outside the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits, and the functions of some or all of the units or modules can be realized by designing the hardware circuits. The above-mentioned hardware circuits can be understood as one or more processors; for example, in one implementation, the above-mentioned hardware circuit is an application-specific integrated circuit (ASIC), which realizes the functions of some or all of the above units or modules by designing the logical relationship of the components in the circuit; for example, in another implementation, the above-mentioned hardware circuit can be realized by a programmable logic device (PLD). Taking a field programmable gate array (FPGA) as an example, it can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured by configuring the configuration file, thereby realizing the functions of some or all of the above units or modules. All units or modules of the above devices can be realized in the form of software called by the processor, or in the form of hardware circuits, or in part by the form of software called by the processor, and the rest by hardware circuits.

[0559] In the embodiments of the present disclosure, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction reading and execution capabilities, such as a central processing unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationship of the hardware circuit. The logical relationship of the above-mentioned hardware circuit is fixed or reconfigurable. For example, the processor is a hardware circuit implemented by an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and implementing the hardware circuit configuration can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a neural network processing unit (NPU), a tensor processing unit (TPU), a deep learning processing unit (DPU), etc.

[0560] FIG8A is a schematic diagram of the structure of the first device proposed in an embodiment of the present disclosure. As shown in FIG8A , the first device 8100 may include: at least one of a transceiver module 8101 and a processing module 8102. The first device 8100 may include:

[0561] The processing module 8102 is configured to obtain a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0562] The processing module 8102 is configured to train an initial encoder and an initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder;

[0563] The processing module 8102 is configured to, based on the CSI associated with each application scenario, maintain all parameters of the first encoder unchanged and fine-tune the first decoder to obtain a second decoder associated with each application scenario; or, based on the CSI associated with each application scenario, maintain some parameters of the first encoder unchanged and fine-tune the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario, wherein different application scenarios are associated with the same first encoder, or second encoders associated with different application scenarios share some network parameters.

[0564] Optionally, after obtaining the second encoder and the second decoder associated with each application scenario, the method further includes:

[0565] The transceiver module 8101 is configured to send one first encoder to the terminal and multiple second decoders to the network device when different application scenarios are associated with the same first encoder; or

[0566] The transceiver module 8101 is used to send the shared network parameters and another part of the network parameters associated with each application scenario to the terminal when the second encoders associated with different application scenarios share some network parameters, and send multiple second decoders to the network device.

[0567] Optionally, the first device is a network device, and after obtaining the second encoder and the second decoder associated with each application scenario, further includes:

[0568] The transceiver module 8101 is configured to send a first encoder to the terminal when different application scenarios are associated with the same first encoder; or

[0569] The transceiver module 8101 is configured to, when second encoders associated with different application scenarios share some network parameters, send the shared network parameters and another part of network parameters associated with each application scenario to the terminal.

[0570] Optionally, it also includes:

[0571] The transceiver module 8101 is configured to receive first indication information sent by a terminal, where the first indication information is used to indicate the size of available storage space of the terminal;

[0572] The processing module 8102 is configured to determine the type of the second encoder corresponding to the terminal according to the first indication information;

[0573] The transceiver module 8101 is configured to send a second encoder applicable to all application scenarios to the terminal according to the type of the second encoder corresponding to the terminal, or send multiple second encoders that share some network parameters to the terminal.

[0574] FIG8B is a schematic diagram of the structure of a terminal proposed in an embodiment of the present disclosure. As shown in FIG8B , the terminal 8200 may include at least one of a transceiver module 8201 and a processing module 8202. The terminal 8200 may include:

[0575] The transceiver module 8201 is configured to receive a first encoder or multiple second encoders sent by a first device, where the multiple second encoders share some network parameters;

[0576] The processing module 8202 is used to deploy a first encoder or multiple second encoders.

[0577] Optionally, the transceiver module 8201 is configured to: send first indication information to the first device, wherein the first indication information is used to indicate the performance of the terminal;

[0578] A second encoder is received from the first device, or multiple second encoders are received from the first device, where the multiple second encoders share some network parameters, and the second encoders are used to encode channel state information CSI.

[0579] Optionally, the terminal performance includes the size of available storage space in the terminal.

[0580] FIG8C is a schematic diagram of the structure of a network device proposed in an embodiment of the present disclosure. As shown in FIG8C , the network device 8300 may include: at least one of a transceiver module 8301 and a processing module 8302. The network device 8300 may include:

[0581] The processing module 8302 is configured to obtain a training data set, wherein the training data set includes channel state information (CSI) under multiple application scenarios;

[0582] The processing module 8302 is configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtain a second decoder and input data of the second decoder associated with each application scenario;

[0583] The transceiver module 8301 is used to send a first data set associated with each application scenario to the terminal, wherein the first data set includes CSI and input data of the second decoder. The first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0584] FIG8D is a schematic diagram of the structure of a terminal proposed in an embodiment of the present disclosure. As shown in FIG8D , the terminal 8400 may include at least one of a transceiver module 8401 and a processing module 8402. The terminal 8400 may include:

[0585] The transceiver module 8401 is configured to receive a first data set associated with each application scenario sent by a network device, wherein the data set includes channel state information CSI associated with each application scenario and input data of a second decoder;

[0586] The processing module 8402 is configured to obtain a second encoder associated with each application scenario through training based on the first data set, wherein the second encoders associated with different application scenarios have the same or share some network parameters.

[0587] Optionally, the processing module 8402 is further configured to:

[0588] determining a target training mode for the second encoder based on the performance of the terminal;

[0589] According to the target training mode, a second encoder associated with each application scenario is obtained by training based on the first data set.

[0590] Optionally, the processing module 8402 is further configured to:

[0591] When the size of the available storage space of the terminal is greater than a size threshold, determining a target training mode to train the second encoder associated with each application scenario based on the CSI associated with each application scenario and input data of the second decoder, wherein the second encoders associated with different application scenarios share some network parameters;

[0592] When the size of the available storage space of the terminal is less than or equal to the size threshold, the target training mode is determined to be training a second encoder associated with all application scenarios based on the CSI associated with all application scenarios and the input data of the second decoder.

[0593] In some embodiments, the transceiver module may include a transmitting module and / or a receiving module, and the transmitting module and the receiving module may be separate or integrated. Optionally, the transceiver module may be interchangeable with the transceiver.

[0594] In some embodiments, the processing module can be a single module or can include multiple submodules. Optionally, the multiple submodules respectively execute all or part of the steps required to be executed by the processing module. Optionally, the processing module can be interchangeable with the processor.

[0595] Figure 9A is a schematic diagram of the structure of a communication device 9100 proposed in an embodiment of the present disclosure. Communication device 9100 can be a terminal, a network device, a chip, a chip system, or a processor that supports a terminal implementing any of the above methods, or a chip, a chip system, or a processor that supports a network device implementing any of the above methods. Communication device 9100 can be used to implement the methods described in the above method embodiments. For details, please refer to the description of the above method embodiments.

[0596] As shown in Figure 9A, the communication device 9100 includes one or more processors 9101. The processor 9101 can be a general-purpose processor or a dedicated processor, for example, a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication device (such as a base station, a baseband chip, a terminal device, a terminal device chip, a DU or a CU, etc.), execute programs, and process program data. The communication device 9100 is used to perform any of the above methods.

[0597] In some embodiments, the communication device 9100 further includes one or more memories 9102 for storing instructions. Optionally, all or part of the memories 9102 may be located outside the communication device 9100.

[0598] In some embodiments, the communication device 9100 further includes one or more transceivers 9103. When the communication device 9100 includes one or more transceivers 9103, the transceiver 9103 performs at least one of the communication steps such as sending and / or receiving in the above method (for example, step S2104, step S2105, step S2204, step S2205, step S2304, step S2306, step S2404, but not limited thereto), and the processor 9101 performs other steps (for example, step S2101, step S2102, step S2103, step S2104, step S2105, step S2204, step S2205, step S2304, step S2306, step S2404, but not limited thereto). S2102, step S2103, step S2106, step S2201, step S2202, step S2203, step S2206, step S2207, step S2301, step S2302, step S2303, step S2305, step S2307, step S2401, step S2402, step S2403, step S2405, step S2406, step S2407).

[0599] In some embodiments, a transceiver may include a receiver and / or a transmitter. The receiver and transmitter may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, and transceiver circuit may be used interchangeably; the terms transmitter, transmitting unit, transmitter, and transmitting circuit may be used interchangeably; and the terms receiver, receiving unit, receiver, and receiving circuit may be used interchangeably.

[0600] In some embodiments, the communication device 9100 may include one or more interface circuits 9104. Optionally, the interface circuit 9104 is connected to the memory 9102. The interface circuit 9104 may be configured to receive signals from the memory 9102 or other devices, and may be configured to send signals to the memory 9102 or other devices. For example, the interface circuit 9104 may read instructions stored in the memory 9102 and send the instructions to the processor 9101.

[0601] The communication device 9100 described in the above embodiments may be a terminal, a network device, or a third entity, but the scope of the communication device 9100 described in the present disclosure is not limited thereto, and the structure of the communication device 9100 may not be limited by FIG. 9A. The communication device may be an independent device or may be part of a larger device. For example, the communication device may be: 1) an independent integrated circuit IC, or a chip, or a chip system or subsystem; (2) a collection of one or more ICs, optionally, the above IC collection may also include a storage component for storing data or programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, a terminal device, an intelligent terminal device, a cellular phone, a wireless device, a handheld device, a mobile unit, an in-vehicle device, a network device, a cloud device, an artificial intelligence device, etc.; (6) others, etc.

[0602] 9B is a schematic diagram of the structure of a chip 9200 according to an embodiment of the present disclosure. If the communication device 9100 can be a chip or a chip system, please refer to the schematic diagram of the structure of the chip 9200 shown in FIG9B , but the present disclosure is not limited thereto.

[0603] The chip 9200 includes one or more processors 9201 , and the chip 9200 is configured to execute any of the above methods.

[0604] In some embodiments, the chip 9200 further includes one or more interface circuits 9202. Optionally, the interface circuit 9202 is connected to the memory 9203. The interface circuit 9202 can be used to receive signals from the memory 9203 or other devices, and can be used to send signals to the memory 9203 or other devices. For example, the interface circuit 9202 can read instructions stored in the memory 9203 and send the instructions to the processor 9201.

[0605] In some embodiments, the interface circuit 9202 performs at least one of the communication steps such as sending and / or receiving in the above method (for example, step S2104, step S2105, step S2204, step S2205, step S2304, step S2306, step S2404, but not limited to this), and the processor 9201 performs other steps (for example, step S2101, step S2102, step S2103, step S2106, step S2201, step S2202, step S2203, step S2206, step S2207, step S2301, step S2302, step S2303, step S2305, step S2307, step S2401, step S2402, step S2403, step S2405, step S2406, step S2407).

[0606] In some embodiments, terms such as interface circuit, interface, transceiver pin, and transceiver may be used interchangeably.

[0607] In some embodiments, the chip 9200 further includes one or more memories 9203 for storing instructions. Alternatively, all or part of the memories 9203 may be located outside the chip 9200.

[0608] The present disclosure also proposes a storage medium having instructions stored thereon, which, when executed on the communication device 9100, causes the communication device 9100 to execute any of the above methods. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but is not limited thereto and may also be a storage medium readable by other devices. Optionally, the storage medium may be a non-transitory storage medium, but is not limited thereto and may also be a temporary storage medium.

[0609] The present disclosure also provides a program product, which, when executed by the communication device 9100, enables the communication device 9100 to perform any of the above methods. Optionally, the program product is a computer program product.

[0610] The present disclosure also proposes a computer program, which, when executed on a computer, causes the computer to perform any one of the above methods.

Claims

1. A method for generating a channel state information (CSI) feedback model, characterized in that, The method is executed by a first device, and the method includes: Obtain a training data set, where the training data set includes channel state information (CSI) in multiple application scenarios; Train an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder; Based on the CSI associated with each application scenario, while keeping all the parameters of the first encoder unchanged, fine-tune the first decoder to obtain a second decoder associated with each application scenario, or, based on the CSI associated with each application scenario, while keeping some of the parameters of the first encoder unchanged, fine-tune the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario, where the same first encoder is associated with different application scenarios, or the second encoders associated with different application scenarios share some network parameters.

2. The method according to claim 1, characterized in that, After obtaining the second encoder and the second decoder associated with each application scenario, it further includes: In the case where the same first encoder is associated with different application scenarios, send one of the first encoders to a terminal, and send multiple second decoders to a network device; or, In the case where the second encoders associated with different application scenarios share some network parameters, send the shared part of the network parameters and the other part of the network parameters associated with each application scenario to the terminal, and send multiple second decoders to the network device.

3. The method according to claim 1, characterized in that, When the first device is a network device, after obtaining the second encoder and the second decoder associated with each application scenario, it further includes: In the case where the same first encoder is associated with different application scenarios, send one of the first encoders to the terminal; or, In the case where the second encoders associated with different application scenarios share some network parameters, send the shared part of the network parameters and the other part of the network parameters associated with each application scenario to the terminal.

4. The method according to any one of claims 1 - 3, characterized in that, It further includes: Receive first indication information sent by the terminal, where the first indication information is used to indicate the scale of the available storage space of the terminal; Determine the type of the second encoder corresponding to the terminal according to the first indication information; According to the type of the second encoder corresponding to the terminal, send one second encoder applicable to all application scenarios to the terminal, or send multiple second encoders sharing some network parameters to the terminal.

5. A method for generating a channel state information (CSI) feedback model, characterized in that, The method is executed by a terminal, and the method includes: Receive one first encoder or multiple second encoders sent by a first device, where the multiple second encoders share some network parameters; Deploy the first encoder or the multiple second encoders.

6. The method according to claim 5, characterized in that, The method further includes: Send first indication information to the first device, where the first indication information is used to indicate the performance of the terminal. Receive a second encoder sent by the first device, or receive multiple second encoders sent by the first device, where the multiple second encoders share some network parameters, and the second encoder is used to encode channel state information (CSI).

7. The method according to claim 6, characterized in that, The terminal performance includes the scale of the available storage space in the terminal.

8. A method for generating a channel state information (CSI) feedback model, characterized in that, The method is executed by a network device, and the method includes: Obtain a training data set, where the training data set includes channel state information (CSI) in multiple application scenarios; Based on the CSI associated with each application scenario, train the initial encoder and the initial decoder in the initial CSI feedback model; Obtain the second decoder associated with each application scenario and the input data of the second decoder; Send the first data set associated with each application scenario to the terminal, where the first data set includes CSI and the input data of the second decoder, and the first data set is used to assist the terminal in training to obtain the second encoder associated with each application scenario, where the second encoders associated with different application scenarios are the same or share some network parameters.

9. A method for generating a channel state information (CSI) feedback model, characterized in that, The method is executed by a terminal, and the method includes: Receive the first data set associated with each application scenario sent by the network device, where the data set includes the channel state information (CSI) associated with each application scenario and the input data of the second decoder; Based on the first data set, train the second encoder associated with each application scenario, where the second encoders associated with different application scenarios are the same or share some network parameters.

10. The method according to claim 9, characterized in that,The training the second encoder associated with each application scenario based on the first data set includes: Determine the target training mode of the second encoder according to the performance of the terminal; Based on the target training mode, train the second encoder associated with each application scenario based on the first data set.

11. The method according to claim 10, wherein, The determining the target training mode of the second encoder according to the performance of the terminal includes: When the scale of the available storage space in the terminal is greater than the scale threshold, determine that the target training mode is to train the second encoder associated with each application scenario based on the CSI and the input data of the second decoder associated with each application scenario, where the second encoders associated with different application scenarios share some network parameters; When the scale of the available storage space in the terminal is less than or equal to the scale threshold, determine that the target training mode is to train one second encoder associated with all application scenarios based on the CSI and the input data of the second decoder associated with all application scenarios.

12. A method for generating a channel state information (CSI) feedback model, wherein, The method includes: The first device obtains a training data set, where the training data set includes channel state information (CSI) in multiple application scenarios; The first device trains the initial encoder and the initial decoder in the initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder; The first device keeps all parameters of the first encoder unchanged based on the CSI associated with each application scenario, and finely tunes the first decoder to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, the first device keeps some parameters of the first encoder unchanged, and finely tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario. Here, the same first encoder is associated with different application scenarios, or the second encoders associated with different application scenarios share some network parameters; The first device sends a first encoder or multiple second encoders to the terminal, where the multiple second encoders share some network parameters; The terminal receives a first encoder or multiple second encoders sent by the first device, where the multiple second encoders share some network parameters; The terminal deploys the first encoder or the multiple second encoders.

13. A method for generating a channel state information (CSI) feedback model, wherein, The method is executed by a communication system, and the method includes: The network device obtains a training data set, where the training data set includes channel state information (CSI) in multiple application scenarios; The network device trains an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtains a second decoder associated with each application scenario and the input data of the second decoder; The network device sends a first data set associated with each application scenario to the terminal, where the first data set includes CSI and the input data of the second decoder; The terminal trains a second encoder associated with each application scenario based on the first data set, where the second encoders associated with different application scenarios are the same or share some network parameters.

14. A first device, wherein, The first device includes: A processing module, configured to obtain a training data set, where the training data set includes channel state information CSI in multiple application scenarios; The processing module is configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the training data set to obtain a first encoder and a first decoder; The processing module is configured to keep all parameters of the first encoder unchanged based on the CSI associated with each application scenario, and finely tune the first decoder to obtain a second decoder associated with each application scenario. Alternatively, based on the CSI associated with each application scenario, the processing module keeps some parameters of the first encoder unchanged, and finely tunes the first decoder and the first encoder respectively to obtain a second encoder and a second decoder associated with each application scenario. Here, the same first encoder is associated with different application scenarios, or the second encoders associated with different application scenarios share some network parameters.

15. A terminal, wherein, The terminal includes: A transceiver module, configured to receive a first encoder or multiple second encoders sent by the first device, where the multiple second encoders share some network parameters; A processing module, configured to deploy the first encoder or the multiple second encoders.

16. A network device, wherein, The network device includes: A processing module, configured to obtain a training data set, where the training data set includes channel state information CSI in multiple application scenarios; The processing module, configured to train an initial encoder and an initial decoder in an initial CSI feedback model based on the CSI associated with each application scenario, and obtain a second decoder and input data of the second decoder associated with each application scenario; A transceiver module, configured to send a first data set associated with each application scenario to a terminal, where the first data set includes CSI and input data of the second decoder, and the first data set is used to assist the terminal in training to obtain a second encoder associated with each application scenario, and the second encoders associated with different application scenarios are the same or share some network parameters.

17. A terminal, wherein, The terminal includes: A transceiver module, configured to receive the first data set associated with each application scenario sent by the network device, where the data set includes CSI and input data of the second decoder associated with each application scenario; A processing module, configured to train a second encoder associated with each application scenario based on the first data set, and the second encoders associated with different application scenarios are the same or share some network parameters.

18. A communication device, characterized in that, including: One or more processors; Wherein, the terminal is configured to execute the method for generating a channel state information CSI feedback model according to any one of claims 1-11.

19. A communication system, characterized in that, including a terminal and a network device, wherein the terminal is configured to implement the method for generating a channel state information CSI feedback model according to any one of claims 5-7 and 9-11, and the network device is configured to implement the method for generating a channel state information CSI feedback model according to any one of claims 1-5 and 8.

20. A storage medium, the storage medium stores instructions, characterized in that, When the instruction runs on the communication device, the communication device is caused to execute the method for generating a channel state information CSI feedback model according to any one of claims 1-11.

Images