CSI feedback method and training method and device of machine learning model

Abstract

The embodiment of the invention provides a CSI (Channel State Information) feedback method and a training method and device of a machine learning model, and the method comprises the steps that a terminal inputs CSI to be fed back and shared environment information of an area to which the terminal belongs into a CSI coding model, and compressed CSI is obtained; and sending the compressed CSI to a network device.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a CSI feedback method, a training method and a device for a machine learning model. Background Art

[0002] In current wireless communication systems, each terminal (or user equipment, User Equipment, UE) independently feeds back its own channel state information (CSI), which results in high feedback overhead or low feedback accuracy. Summary of the invention

[0003] In response to the problems existing in the prior art, the embodiments of the present application provide a CSI feedback method, a training method and a device for a machine learning model.

[0004] In a first aspect, an embodiment of the present application provides a channel state information CSI feedback method, which is applied to a terminal, including:

[0005] Inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI;

[0006] The compressed CSI is sent to the network device.

[0007] In some embodiments, the CSI coding model includes 1 CSI encoder and 1 joint encoder;

[0008] The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes:

[0009] Inputting the CSI to be fed back into the CSI encoder for encoding;

[0010] The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

[0011] In some embodiments, the CSI coding model is part of a two-layer autoencoder;

[0012] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0013] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0014] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0015] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0016] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0017] In some embodiments, the method further comprises:

[0018] Receive the CSI coding model and the shared environment information sent by the network device.

[0019] In some embodiments, the method further comprises:

[0020] According to the mobile state of the terminal, requesting the network device for the shared environment information of the new area to which the terminal belongs; or,

[0021] According to the moving state of the terminal, the shared environment information of the new area to which the terminal belongs is called locally.

[0022] In a second aspect, an embodiment of the present application further provides a channel state information CSI feedback method, which is applied to a network device, including:

[0023] receiving the compressed CSI sent by the terminal;

[0024] The compressed CSI is input into a CSI decoding model to obtain the CSI fed back by the terminal.

[0025] In some embodiments, the CSI decoding model includes 1 joint decoder and 1 CSI decoder;

[0026] The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes:

[0027] Inputting the compressed CSI into the joint decoder for joint decoding;

[0028] The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

[0029] In some embodiments, the CSI decoding model is part of a two-layer autoencoder;

[0030] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0031] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0032] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0033] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0034] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0035] In some embodiments, the method further comprises:

[0036] The two-layer autoencoder is trained based on the CSI data set of each sub-area within the coverage area of ​​the network device.

[0037] In some embodiments, the training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0038] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0039] In some embodiments, the method further comprises:

[0040] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0041] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0042] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0043] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0044] In some embodiments, the method further comprises:

[0045] Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

[0046] In some embodiments, the method further comprises:

[0047] Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder;

[0048] The CSI coding model is sent to terminals in each sub-area.

[0049] In some embodiments, the method further comprises:

[0050] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0051] In some embodiments, the method further comprises:

[0052] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0053] In some embodiments, the method further comprises:

[0054] Sending a CSI reporting notification to each terminal in the sub-area;

[0055] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0056] In a third aspect, the present application also provides a method for training a machine learning model, including:

[0057] Train the machine learning model based on the channel state information (CSI) dataset of each sub-area within the coverage area of ​​the network device;

[0058] The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

[0059] In some embodiments, the dual-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0060] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0061] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0062] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0063] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0064] In some embodiments, the training of the machine learning model based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0065] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0066] In some embodiments, the method further comprises:

[0067] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0068] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0069] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0070] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0071] In some embodiments, the method further comprises:

[0072] Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

[0073] In some embodiments, the method further comprises:

[0074] Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder;

[0075] The CSI coding model is sent to terminals in each sub-area.

[0076] In some embodiments, the method further comprises:

[0077] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0078] In some embodiments, the method further comprises:

[0079] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0080] In some embodiments, the method further comprises:

[0081] Sending a CSI reporting notification to each terminal in the sub-area;

[0082] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0083] In a fourth aspect, an embodiment of the present application further provides a terminal, including a memory, a transceiver, and a processor;

[0084] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0085] Inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI;

[0086] The compressed CSI is sent to the network device.

[0087] In some embodiments, the CSI coding model includes 1 CSI encoder and 1 joint encoder;

[0088] The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes:

[0089] Inputting the CSI to be fed back into the CSI encoder for encoding;

[0090] The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

[0091] In some embodiments, the CSI coding model is part of a two-layer autoencoder;

[0092] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0093] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0094] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0095] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0096] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0097] In some embodiments, the operations further include:

[0098] Receive the CSI coding model and the shared environment information sent by the network device.

[0099] In some embodiments, the operations further include:

[0100] According to the mobile state of the terminal, requesting the network device for the shared environment information of the new area to which the terminal belongs; or,

[0101] According to the moving state of the terminal, the shared environment information of the new area to which the terminal belongs is called locally.

[0102] In a fifth aspect, an embodiment of the present application further provides a network device, including a memory, a transceiver, and a processor;

[0103] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0104] receiving the compressed CSI sent by the terminal;

[0105] The compressed CSI is input into a CSI decoding model to obtain the CSI fed back by the terminal.

[0106] In some embodiments, the CSI decoding model includes 1 joint decoder and 1 CSI decoder;

[0107] The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes:

[0108] Inputting the compressed CSI into the joint decoder for joint decoding;

[0109] The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

[0110] In some embodiments, the CSI decoding model is part of a two-layer autoencoder;

[0111] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0112] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0113] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0114] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0115] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0116] In some embodiments, the operations further include:

[0117] The two-layer autoencoder is trained based on the CSI data set of each sub-area within the coverage area of ​​the network device.

[0118] In some embodiments, the training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0119] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0120] In some embodiments, the operations further include:

[0121] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0122] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0123] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0124] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0125] In some embodiments, the operations further include:

[0126] Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

[0127] In some embodiments, the operations further include:

[0128] Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder;

[0129] The CSI coding model is sent to terminals in each sub-area.

[0130] In some embodiments, the operations further include:

[0131] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0132] In some embodiments, the operations further include:

[0133] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0134] In some embodiments, the operations further include:

[0135] Sending a CSI reporting notification to each terminal in the sub-area;

[0136] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0137] In a sixth aspect, an embodiment of the present application further provides a network device, including a memory, a transceiver, and a processor;

[0138] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0139] Training a machine learning model based on a channel state information (CSI) data set of each sub-area within the coverage area of ​​the network device;

[0140] The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

[0141] In some embodiments, the dual-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0142] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0143] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0144] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0145] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0146] In some embodiments, the training of the machine learning model based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0147] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0148] In some embodiments, the operations further include:

[0149] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0150] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0151] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0152] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0153] In some embodiments, the operations further include:

[0154] Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

[0155] In some embodiments, the operations further include:

[0156] Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder;

[0157] The CSI coding model is sent to terminals in each sub-area.

[0158] In some embodiments, the operations further include:

[0159] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0160] In some embodiments, the operations further include:

[0161] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0162] In some embodiments, the operations further include:

[0163] Sending a CSI reporting notification to each terminal in the sub-area;

[0164] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0165] In a seventh aspect, an embodiment of the present application further provides a channel state information CSI feedback device, including:

[0166] A coding unit, used for inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI;

[0167] The first sending unit is configured to send the compressed CSI to a network device.

[0168] In some embodiments, the CSI coding model includes 1 CSI encoder and 1 joint encoder;

[0169] The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes:

[0170] Inputting the CSI to be fed back into the CSI encoder for encoding;

[0171] The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

[0172] In some embodiments, the CSI coding model is part of a two-layer autoencoder;

[0173] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0174] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0175] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0176] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0177] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0178] In some embodiments, the apparatus further comprises:

[0179] The first receiving unit is used to receive the CSI coding model and the shared environment information sent by the network device.

[0180] In some embodiments, the apparatus further comprises:

[0181] a requesting unit, configured to request, from the network device, shared environment information of a new area to which the terminal belongs according to a mobile state of the terminal; or

[0182] The calling unit is used to call the shared environment information of the new area to which the terminal belongs from the local according to the moving state of the terminal.

[0183] In an eighth aspect, an embodiment of the present application further provides a channel state information CSI feedback device, including:

[0184] A second receiving unit, configured to receive the compressed CSI sent by the terminal;

[0185] The decoding unit is used to input the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal.

[0186] In some embodiments, the CSI decoding model includes 1 joint decoder and 1 CSI decoder;

[0187] The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes:

[0188] Inputting the compressed CSI into the joint decoder for joint decoding;

[0189] The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

[0190] In some embodiments, the CSI decoding model is part of a two-layer autoencoder;

[0191] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0192] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0193] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0194] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0195] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0196] In some embodiments, the apparatus further comprises:

[0197] The first training unit is used to train the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device.

[0198] In some embodiments, the training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0199] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0200] In some embodiments, the apparatus further comprises:

[0201] The first extraction unit is used to extract the shared environment information of the sub-region based on the CSI dataset of the sub-region and the trained double-layer autoencoder after training the double-layer autoencoder based on the CSI dataset of the sub-region each time.

[0202] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0203] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0204] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0205] In some embodiments, the apparatus further comprises a first deployment unit configured to:

[0206] Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

[0207] In some embodiments, the apparatus further comprises a second deployment unit, configured to:

[0208] Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder;

[0209] The CSI coding model is sent to terminals in each sub-area.

[0210] In some embodiments, the apparatus further comprises:

[0211] The second sending unit is configured to send the shared environment information of the corresponding sub-area to the terminals in each sub-area.

[0212] In some embodiments, the apparatus further comprises:

[0213] The third sending unit is used to send the shared environment information of the new area to which the terminal belongs to the terminal according to the request information sent by the terminal.

[0214] In some embodiments, the device further comprises a first collecting unit, configured to:

[0215] Sending a CSI reporting notification to each terminal in the sub-area;

[0216] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0217] In a ninth aspect, an embodiment of the present application further provides a training device for a machine learning model, comprising:

[0218] A second training unit is used to train a machine learning model based on a channel state information CSI data set of each sub-area within the coverage area of ​​the network device;

[0219] The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

[0220] In some embodiments, the dual-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0221] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0222] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0223] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0224] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0225] In some embodiments, the training of the machine learning model based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0226] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0227] In some embodiments, the apparatus further comprises:

[0228] The second extraction unit is used to extract the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained double-layer autoencoder after training the double-layer autoencoder based on the CSI data set of the sub-area each time.

[0229] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0230] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0231] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0232] In some embodiments, the apparatus further comprises a third deployment unit, configured to:

[0233] Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

[0234] In some embodiments, the apparatus further comprises a fourth deployment unit, configured to:

[0235] Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder;

[0236] The CSI coding model is sent to terminals in each sub-area.

[0237] In some embodiments, the apparatus further comprises:

[0238] The fourth sending unit is configured to send the shared environment information of the corresponding sub-area to the terminals in each sub-area.

[0239] In some embodiments, the apparatus further comprises:

[0240] The fifth sending unit is used to send the shared environment information of the new area to which the terminal belongs to the terminal according to the request information sent by the terminal.

[0241] In some embodiments, the device further comprises a second collecting unit, configured to:

[0242] Sending a CSI reporting notification to each terminal in the sub-area;

[0243] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0244] In the tenth aspect, an embodiment of the present application also provides a non-transitory readable storage medium, wherein the non-transitory readable storage medium stores a computer program, and the computer program is used to enable a processor to execute the CSI feedback method described in the first aspect as described above, or execute the CSI feedback method described in the second aspect as described above, or execute the training method of the machine learning model described in the third aspect as described above.

[0245] In the eleventh aspect, an embodiment of the present application also provides a communication device, in which a computer program is stored, and the computer program is used to enable the communication device to execute the CSI feedback method described in the first aspect as described above, or execute the CSI feedback method described in the second aspect as described above, or execute the training method of the machine learning model described in the third aspect as described above.

[0246] In the twelfth aspect, an embodiment of the present application also provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute the CSI feedback method described in the first aspect as described above, or execute the CSI feedback method described in the second aspect as described above, or execute the training method of the machine learning model described in the third aspect as described above.

[0247] In the thirteenth aspect, an embodiment of the present application further provides a chip product, in which a computer program is stored, and the computer program is used to enable the chip product to execute the CSI feedback method described in the first aspect as described above, or execute the CSI feedback method described in the second aspect as described above, or execute the training method of the machine learning model described in the third aspect as described above.

[0248] The CSI feedback method, machine learning model training method and device provided in the embodiments of the present application use AI technology to learn the propagation environment characteristics shared by multiple terminals in the same communication environment, and input the CSI to be fed back and the shared environment information of the area to which the terminal belongs into the trained CSI coding model during CSI coding compression, so that the CSI coding model can perform concise compression on the CSI to be fed back based on the shared environment information of the area to which the terminal belongs, remove redundant information, thereby reducing the CSI feedback overhead and improving the CSI feedback accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0249] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the following is a brief introduction to the drawings required for use in the embodiments or the related technical descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0250] Figure 1 One of the flow charts of the CSI feedback method provided in the embodiment of the present application;

[0251] Figure 2 A schematic diagram of the structure of a double-layer autoencoder provided in an embodiment of the present application;

[0252] Figure 3 One of the schematic diagrams of the deployment of the double-layer autoencoder provided in the embodiment of the present application;

[0253] Figure 4 The second schematic diagram of the deployment of a double-layer autoencoder provided in an embodiment of the present application;

[0254] Figure 5 A second flowchart of the CSI feedback method provided in an embodiment of the present application;

[0255] Figure 6 A flowchart of a method for training a machine learning model provided in an embodiment of the present application;

[0256] Figure 7 A schematic diagram of the structure of a terminal provided in an embodiment of the present application;

[0257] Figure 8One of the structural diagrams of the network device provided in the embodiment of the present application;

[0258] Fig. 9 A second structural diagram of a network device provided in an embodiment of the present application;

[0259] Fig.10 One of the structural schematic diagrams of the CSI feedback device provided in an embodiment of the present application;

[0260] Fig.11 The second structural diagram of the CSI feedback device provided in the embodiment of the present application;

[0261] Fig.12 A schematic diagram of the structure of a training device for a machine learning model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0262] In the embodiments of the present application, the term "and / or" describes the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0263] In the embodiments of the present application, the term "plurality" refers to two or more than two, and other quantifiers are similar.

[0264] In the embodiments of the present application, the terms "first", "second", etc. are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first" and "second" are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more.

[0265] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0266] In order to facilitate a clearer understanding of the technical solutions of the embodiments of the present application, some technical contents related to the embodiments of the present application are first introduced.

[0267] Currently, there are three main ways to provide CSI feedback:

[0268] 1. Implicit feedback: The terminal feeds back CSI in the form of recommended transmission parameters (such as Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), Rank Indicator (RI), CSI-RS Resource Indicator (CRI), etc.), and the base station can directly use the parameters recommended by the terminal for transmission. The focus of implicit feedback is on the design of the codebook, striving to accurately reflect the channel state information. Specifically, 5G New Radio (NR) supports two types of CSI feedback: one is ordinary precision Type I CSI feedback, which has ordinary precision spatial resolution and PMI feedback based on Type I codebook; the other is high-precision Type II CSI feedback, which has higher spatial resolution and aims to improve the performance of multi-user multiple input multiple output (MU-MIMO), and PMI feedback is based on Type II codebook.

[0269] 2. Explicit feedback: Directly feed back channel parameters, such as the channel correlation matrix, channel matrix, or channel eigenvector information. Explicit feedback has a large overhead and requires further dimensionality reduction to reduce the amount of feedback.

[0270] 3. Feedback based on channel reciprocity: It can be divided into feedback based on complete channel reciprocity and feedback based on partial channel reciprocity.

[0271] The CSI feedback of 5G NR is mainly implicit feedback, supplemented by channel reciprocity feedback of the time division duplex (TDD) system.

[0272] In current wireless communication systems, no matter which CSI feedback mode is used, each user independently feeds back its own CSI. However, in fact, multiple users in a wireless communication system share the same propagation environment, that is, there is a correlation between the CSIs of multiple users. Therefore, this application proposes a multi-user CSI fusion technology based on artificial intelligence (AI) to learn the propagation environment characteristics shared by multiple users in the system, fuse multi-user CSI, reduce the overhead of CSI feedback, and improve the accuracy of CSI feedback.

[0273] Figure 1 One of the flow charts of the CSI feedback method provided in the embodiment of the present application, the method can be applied to the terminal, such as Figure 1 As shown, the method comprises the following steps:

[0274] Step 100: Input the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI.

[0275] Step 101: Send the compressed CSI to a network device.

[0276] Specifically, the CSI coding model is a machine learning model for CSI coding (or compression). The CSI coding model is a machine learning model of the coding part obtained after training the complete coding and decoding machine learning model. After training, the complete coding and decoding machine learning model can learn the propagation environment characteristics shared by multiple terminals in the same communication environment, so that when encoding the CSI to be fed back by the terminal, the propagation environment characteristics shared by multiple terminals in the same communication environment can be effectively used to reduce and compress the CSI to be fed back, remove redundant information, and finally achieve the purpose of reducing CSI feedback overhead and improving CSI feedback accuracy.

[0277] In some embodiments, the complete codec machine learning model can be trained using the CSI of multiple groups of terminals as training samples, where each group of terminals is in the same communication environment (the large-scale channel information in the same communication environment is very similar).

[0278] In some embodiments, after the complete encoding and decoding machine learning model is trained on the network device (e.g., base station) side, the CSI encoding model can be deployed to the terminal side, and the corresponding CSI decoding model can be deployed on the network device side to decode and obtain the CSI fed back by the terminal.

[0279] When the terminal performs CSI compression, it first needs to determine the CSI coding model used and the shared environment information of the area to which the terminal belongs. In some embodiments, the coverage of the network device can be divided into several sub-areas according to the characteristics of the propagation environment. The terminal can learn about the division of the sub-area and determine the sub-area it is in (i.e., the area to which the terminal belongs) based on its own location information; or, the network device can obtain the location information of the terminal, determine the sub-area to which the terminal belongs, and notify the terminal of the sub-area to which the terminal belongs.

[0280] In some embodiments, if there is more than one CSI coding model stored locally, the terminal can select one of them for CSI compression as needed, or use the latest CSI coding model for CSI compression, or select a CSI coding model for CSI compression according to the location (area) of the terminal, etc., without specific limitations here.

[0281] In some embodiments, if there is more than one shared environment information stored locally, for example, there are multiple regions of shared environment information stored locally, or there are versions of the shared environment information of the region to which the terminal belongs that are updated at different time points, the terminal can use the latest shared environment information of the region to which it belongs. Specifically, the terminal can first determine its own location information, and then determine which region it is in based on its own location information, and then use the shared environment information of the corresponding region to perform CSI compression; or the network device can obtain the location information of the terminal, determine the sub-region to which the terminal belongs, notify the terminal of the sub-region to which the terminal belongs, and the terminal uses the shared environment information of the corresponding region to perform CSI compression.

[0282] In this application, shared environment information refers to information that can characterize the propagation environment characteristics shared by multiple terminals in the same communication environment, and its specific form is not limited in this application. For example, it can be similar large-scale channel information extracted from CSI fed back by multiple terminals in the same communication environment; or it can be the typical value (such as average, median, mode, etc.) of the encoded information after encoding the CSI fed back by multiple terminals in the same communication environment as shared environment information; or it can be other forms of information.

[0283] After determining the CSI coding model to be used and the shared environment information of the area to which the terminal belongs, the terminal can input the CSI to be fed back and the shared environment information of the area to which the terminal belongs into the CSI coding model to obtain the compressed CSI (or encoded CSI) output by the CSI coding model.

[0284] Then, the terminal sends the compressed CSI to the network device. After receiving the compressed CSI, the network device can use the CSI decoding model corresponding to the CSI encoding model on the terminal side to decode it, obtain the CSI fed back by the terminal, and realize CSI feedback.

[0285] The CSI feedback method provided in the embodiment of the present application utilizes AI technology to learn the propagation environment characteristics shared by multiple terminals in the same communication environment, and inputs the CSI to be fed back and the shared environment information of the area to which the terminal belongs into the trained CSI coding model during CSI coding compression, so that the CSI coding model can perform precise compression on the CSI to be fed back based on the shared environment information of the area to which the terminal belongs, remove redundant information, thereby reducing the CSI feedback overhead and improving the CSI feedback accuracy.

[0286] In some embodiments, the CSI coding model includes 1 CSI encoder and 1 joint encoder;

[0287] The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes:

[0288] Inputting the CSI to be fed back into the CSI encoder for encoding;

[0289] The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

[0290] Specifically, the CSI coding model can be a two-layer structure, including a CSI encoder on the first layer and a joint encoder on the second layer. When performing CSI compression coding, the CSI encoder on the first layer can first encode the input CSI to be fed back and output the coding information. The coding information is then input together with the shared environment information into the joint encoder on the second layer for joint coding. The information output by the joint encoder is sent to the network device as compressed CSI.

[0291] In some embodiments, the CSI coding model is part of a two-layer autoencoder;

[0292] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0293] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0294] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0295] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0296] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0297] Specifically, in an embodiment of the present application, the complete encoding and decoding machine learning model can adopt a two-layer auto-encoder (auto-encoder) structure, that is, it includes a two-layer encoder (Encoder) and a two-layer decoder (Decoder), and extracts the propagation environment characteristics shared by multiple terminals through the second-layer joint encoder (Co-encoder).

[0298] Figure 2 A schematic diagram of the structure of a double-layer autoencoder provided in an embodiment of the present application is shown in FIG. Figure 2As shown, the architecture of the two-layer autoencoder includes N first-layer Encoders on the encoding side, 1 Co-encoder on the second layer on the encoding side, 1 Co-decoder on the first layer on the decoding side, and N second-layer Decoders on the decoding side.

[0299] Among them, the N Encoders in the first layer of the encoding side are responsible for compressing and encoding the CSI of each terminal respectively; the 1 Co-encoder in the second layer of the encoding side is responsible for jointly compressing and encoding all the input N terminals to extract the environmental information shared between the N terminals for CSI compression; the 1 Co-decoder in the first layer of the decoding side is responsible for jointly decoding the jointly encoded N terminal CSI to restore the environmental information shared between the N terminals; the N Decoders in the second layer of the decoding side are responsible for decoding the CSI unique information of each terminal to obtain the CSI of different terminals.

[0300] Figure 3 One of the schematic diagrams of the deployment of the double-layer autoencoder provided in the embodiment of the present application is as follows: Figure 3 As shown, based on Figure 2 The double-layer autoencoder structure shown in the figure, after the double-layer autoencoder is trained, part of the encoding side model can be deployed on the terminal, and part of the decoding side model can be deployed on the base station. Among them, the CSI encoding model deployed on the terminal side includes an encoder on the first layer of the encoding side and a co-encoder on the second layer of the encoding side. The input of the co-encoder includes the encoding information output by the encoder and the shared environment information of the area to which the terminal belongs; the CSI decoding model deployed on the base station side includes a co-decoder on the first layer of the decoding side and a decoder on the second layer of the decoding side. The decoder is used to decode the terminal-specific information to obtain the CSI of different terminals.

[0301] Figure 4 The second schematic diagram of the deployment of the double-layer autoencoder provided in the embodiment of the present application is as follows: Figure 4 As shown, the model is deployed in the same way as Figure 3 The CSI coding model deployed on the terminal side is different from Figure 3 The difference is the model deployed on the base station side. Figure 4 In the present invention, the CSI decoding model deployed on the base station side includes one Co-decoder on the first layer of the decoding side and N decoders on the second layer of the decoding side.

[0302] In some embodiments, the method further comprises:

[0303] Receive the CSI coding model and the shared environment information sent by the network device.

[0304] Specifically, the complete encoding and decoding machine learning model can be trained on the network device side. After the training is completed, the network device can send the CSI coding model that needs to be deployed on the terminal side to the terminal, and send the shared environment information of the area to which the terminal belongs to the terminal.

[0305] Optionally, the network device may send the shared environment information of all sub-areas within the coverage of the network device to the terminal; or, the network device may determine the area to which the terminal belongs according to the location information of the terminal, and then send the shared environment information of the area to which the terminal belongs to the terminal.

[0306] In some embodiments, the network device can periodically update the machine learning model (a complete encoding and decoding machine learning model, a CSI encoding model deployed on the terminal side, or a CSI decoding model deployed on the network device side), as well as the shared environment information of each sub-area, and deploy the updated machine learning model and shared environment information to the corresponding devices.

[0307] In some embodiments, the method further comprises:

[0308] According to the mobile state of the terminal, requesting the network device for the shared environment information of the new area to which the terminal belongs; or,

[0309] According to the moving state of the terminal, the shared environment information of the new area to which the terminal belongs is called locally.

[0310] Specifically, the area to which the terminal belongs may change during the movement process, so the terminal can request updated shared environment information from the network device in a timely manner according to its movement status to ensure the accuracy of CSI feedback. For example, if the terminal determines that its position at a certain moment exceeds the scope of the original area and enters another area, the terminal can send a request message to the network device, and the request message can include information about the other area, so that the network device can send new shared environment information to the terminal, or the request message does not carry information about the other area. After receiving the request message, the network device can obtain the location information of the terminal to determine the area to which the terminal belongs, and then send the shared environment information of the corresponding area to the terminal.

[0311] In some embodiments, the terminal stores the shared environment information of each area locally, and the terminal can call the shared environment information of the new area to which the terminal belongs in a timely manner from the local according to its mobile state to ensure the accuracy of CSI feedback. For example, if the terminal determines that its position at a certain moment exceeds the scope of the original area and enters another area, the terminal can call the shared environment information of the other area from the local in a timely manner, and use the new shared environment information for CSI compression encoding when performing CSI feedback.

[0312] Figure 5 The second flow chart of the CSI feedback method provided in the embodiment of the present application is as follows. The method can be applied to a network device (eg, a base station). Figure 5 As shown, the method comprises the following steps:

[0313] Step 500: Receive compressed CSI sent by a terminal.

[0314] Step 501: Input the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal.

[0315] Specifically, when the terminal performs CSI feedback, it first determines the CSI coding model used and the shared environment information of the area to which the terminal belongs, and then inputs the CSI to be fed back and the shared environment information of the area to which the terminal belongs into the CSI coding model to obtain compressed CSI and send it to the network device.

[0316] After receiving the compressed CSI, the network device can use the CSI decoding model corresponding to the CSI encoding model on the terminal side to decode it, obtain the CSI fed back by the terminal, and realize CSI feedback.

[0317] The CSI feedback method provided in the embodiment of the present application utilizes AI technology to learn the propagation environment characteristics shared by multiple terminals in the same communication environment, and inputs the CSI to be fed back and the shared environment information of the area to which the terminal belongs into the trained CSI coding model during CSI coding compression, so that the CSI coding model can perform a compact compression on the CSI to be fed back based on the shared environment information of the area to which the terminal belongs, remove redundant information, and then use the corresponding CSI decoding model on the network device side to decode the CSI fed back by the terminal, thereby reducing the CSI feedback overhead and improving the CSI feedback accuracy.

[0318] In some embodiments, the CSI decoding model includes 1 joint decoder and 1 CSI decoder;

[0319] The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes:

[0320] Inputting the compressed CSI into the joint decoder for joint decoding;

[0321] The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

[0322] Specifically, the CSI decoding model can be a two-layer structure, including a joint decoder on the first layer and a CSI decoder on the second layer. When performing CSI decoding (or decoding), the joint decoder on the first layer can first jointly decode the received compressed CSI, output joint decoding information, and then input the joint decoding information into the CSI decoder on the second layer for decoding to obtain the CSI fed back by the terminal.

[0323] In some embodiments, the CSI decoding model is part of a two-layer autoencoder;

[0324] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0325] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0326] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0327] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0328] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0329] Specifically, in an embodiment of the present application, a complete encoding and decoding machine learning model can adopt a two-layer auto-encoder structure, that is, it includes a two-layer encoder (Encoder) and a two-layer decoder (Decoder), and extracts the propagation environment characteristics shared by multiple terminals through the second-layer joint encoder (Co-encoder).

[0330] like Figure 2 As shown, the architecture of the two-layer autoencoder includes N first-layer Encoders on the encoding side, 1 Co-encoder on the second layer on the encoding side, 1 Co-decoder on the first layer on the decoding side, and N second-layer Decoders on the decoding side.

[0331] Among them, the N Encoders in the first layer of the encoding side are responsible for compressing and encoding the CSI of each terminal respectively; the 1 Co-encoder in the second layer of the encoding side is responsible for jointly compressing and encoding all the input N terminals to extract the environmental information shared between the N terminals for CSI compression; the 1 Co-decoder in the first layer of the decoding side is responsible for jointly decoding the jointly encoded N terminal CSI to restore the environmental information shared between the N terminals; the N Decoders in the second layer of the decoding side are responsible for decoding the CSI unique information of each terminal to obtain the CSI of different terminals.

[0332] In some embodiments, the method further comprises:

[0333] The two-layer autoencoder is trained based on the CSI data set of each sub-area within the coverage area of ​​the network device.

[0334] Specifically, the two-layer autoencoder can be trained on the network device side, and the data used for training includes CSI data sets of each sub-area within the coverage range of the network device. The CSI data set of each sub-area contains the CSI reported by the terminal in the corresponding sub-area.

[0335] In some embodiments, the method further comprises:

[0336] Sending a CSI reporting notification to each terminal in the sub-area;

[0337] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0338] Specifically, the network device may notify the terminals within the coverage area to report CSI, and form a CSI data set for each sub-area according to the collected CSI reported by the terminals in each sub-area.

[0339] The following is an exemplary description of the process of collecting CSI data sets, which mainly includes:

[0340] (1) The base station divides its coverage area into multiple sub-areas so that the propagation environment from the base station to the terminals in each sub-area is similar, that is, the terminals in each sub-area share the same propagation environment.

[0341] (2) The base station notifies the terminals in each sub-area to report the specified CSI type within the specified idle time.

[0342] (3) The base station stores the CSI fed back by all terminals received in all sub-areas.

[0343] (4) The base station generates a CSI data set for each sub-area according to the sub-areas, wherein the CSI data set for each sub-area stores a specified type of CSI according to the terminal.

[0344] In some embodiments, the training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0345] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0346] Specifically, the arrangement order of the sub-regions can be arbitrary, as long as the two-layer autoencoder is trained in sequence according to the order after arrangement. This application does not limit the specific arrangement.

[0347] For example, the CSI data set of the first sub-area can be used to train a double-layer autoencoder. After the training reaches the convergence condition, the obtained machine learning model parameters are stored, and then the CSI data set of the second sub-area is used to further train the double-layer autoencoder trained with the CSI data set of the first sub-area. Similarly, after the training reaches the convergence condition, the obtained machine learning model parameters are stored, and then the CSI data set of the third sub-area is used to further train the double-layer autoencoder trained with the CSI data set of the second sub-area, and so on. The double-layer autoencoder is trained based on the CSI data set of each sub-area in turn until all sub-areas are traversed to obtain the final trained machine learning model. The machine learning model can learn the propagation environment characteristics shared by terminals in all sub-areas, so that it can be used for CSI feedback to effectively reduce feedback overhead and improve feedback accuracy.

[0348] Since the autoencoder is a self-supervised learning model, that is, the input and output of the autoencoder are the same, ideally, the output of the decoder should be equal to the input of the encoder, but in reality there will always be some errors. As long as the error is less than the threshold value, the output of the autoencoder can be considered correct, that is, the convergence condition is met.

[0349] In some embodiments, the method further comprises:

[0350] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0351] For example, after the network device completes training of a double-layer autoencoder using the CSI data set of the first sub-area, it can extract the shared environment information of the first sub-area based on the CSI data set of the first sub-area and the currently trained double-layer autoencoder.

[0352] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0353] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0354] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0355] Taking the extraction of shared environment information of the first sub-area as an example, assuming that the two-layer autoencoder includes N CSI encoders, the CSI of any N terminals in the CSI data set of the first sub-area can be input into the N CSI encoders respectively to obtain N encoding information, and then based on the N encoding information, the shared environment information of the first sub-area is obtained.

[0356] In some embodiments, the process of "inputting the CSI of any N terminals in the CSI data set of the first sub-area into N CSI encoders respectively to obtain N coding information" can be repeated multiple times to obtain N×m coding information (where m is the number of repetitions), and then based on these N×m coding information, the shared environment information of the first sub-area is obtained.

[0357] In some implementations, a typical value of the above-mentioned multiple coded information (N or N×m coded information) can be used as the shared environment information of the first sub-region. The typical value can be, for example, an average value, a median, a mode, or other value that can reflect the typical characteristics of the multiple coded information.

[0358] The following is an exemplary description of the training process of the machine learning model, which mainly includes:

[0359] (1) The base station performs machine learning model training on a two-layer autoencoder for the first divided sub-area according to the collected CSI data set until convergence, wherein each encoder or decoder can adopt a transformer structure.

[0360] (2) After obtaining the trained machine learning model in the first sub-area, the obtained machine learning model parameters are stored, and the corresponding shared environment information in the first sub-area is also stored.

[0361] (3) Based on the CSI data set of the second sub-area and the machine learning model obtained from the first sub-area, continue to train the machine learning model obtained based on the CSI data set of the first sub-area until convergence.

[0362] (4) The machine learning model parameter set updated based on the CSI data set of the second sub-area is stored, and the shared environment information of the second sub-area is also stored.

[0363] (5) Repeat steps (3) and (4) above, that is, based on the CSI data set of the n+1th sub-area, train and update the machine learning model obtained based on the CSI data sets of the 1st to nth sub-areas, and store the updated machine learning model parameter set and the shared environment information in the n+1th sub-area until all sub-areas within the coverage area of ​​the base station are traversed.

[0364] In some embodiments, the value of N may be selected from the following options:

[0365] (1) In each sub-area, the same number of N terminals is selected to train the machine learning model to reduce the cost of training and storage.

[0366] (2) In each sub-area, there are multiple different levels of numbers N. A set of machine learning models is trained for each value of N at different levels. In actual use, it is selected according to needs. However, it is worth noting that the number of terminals N in the shared environment information of the machine learning model needs to match the N in the machine learning model used. This method can obtain a more accurate machine learning model based on different levels of N, but the training and storage costs are higher and can be selected according to needs.

[0367] In some embodiments, the method further comprises:

[0368] Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

[0369] Specifically, after the two-layer autoencoder is trained on the network device side, the network device can deploy the CSI decoding model on the network device side for decoding CSI feedback information, such as Figure 3 As shown, the CSI decoding model deployed on the base station side includes a Co-decoder on the first layer of the decoding side and a Decoder on the second layer of the decoding side. The Decoder is used to decode the terminal-specific information to obtain the CSI of different terminals.

[0370] In some embodiments, the method further comprises:

[0371] Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder;

[0372] The CSI coding model is sent to terminals in each sub-area.

[0373] Specifically, after the two-layer autoencoder is trained on the network device side, the network device can first determine the CSI coding model that needs to be deployed on the terminal side, and then send the determined deployed CSI coding model to the terminals in each sub-area.

[0374] For example, network devices are determined to be deployed on the terminal side as follows Figure 3 The CSI coding model shown includes an encoder on the first layer of the coding side and a co-encoder on the second layer of the coding side. The input of the co-encoder includes the coding information output by the encoder and the shared environment information of the area to which the terminal belongs. The network device can send the CSI coding model to each terminal.

[0375] In some embodiments, the method further comprises:

[0376] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0377] Specifically, after the two-layer autoencoder is trained on the network device side, the network device can send the CSI coding model that needs to be deployed on the terminal side to the terminal, and send the shared environment information of the area to which the terminal belongs to the terminal.

[0378] Optionally, the network device may send the shared environment information of all sub-areas within the coverage of the network device to the terminal; or, the network device may determine the area to which the terminal belongs according to the location information of the terminal, and then send the shared environment information of the area to which the terminal belongs to the terminal.

[0379] In some embodiments, the network device can periodically update the machine learning model (a two-layer autoencoder, a CSI encoding model deployed on the terminal side, or a CSI decoding model deployed on the network device side) and the shared environment information of each sub-area, and deploy the updated machine learning model and shared environment information to the corresponding device.

[0380] In some embodiments, the method further comprises:

[0381] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0382] Specifically, the area to which the terminal belongs may change during the movement process, so the terminal can request updated shared environment information from the network device in a timely manner according to its movement status to ensure the accuracy of CSI feedback. For example, if the terminal determines that its position at a certain moment exceeds the scope of the original area and enters another area, the terminal can send a request message to the network device, and the request message can include information about the other area, so that the network device can send new shared environment information to the terminal, or the request message does not carry information about the other area. After receiving the request message, the network device can obtain the location information of the terminal to determine the area to which the terminal belongs, and then send the shared environment information of the corresponding area to the terminal.

[0383] Figure 6 A flow chart of a training method for a machine learning model provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the method comprises the following steps:

[0384] Step 600: Train a machine learning model based on a channel state information CSI data set of each sub-area within the coverage area of ​​the network device; the machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding the CSI of multiple terminals in the same sub-area.

[0385] Specifically, the execution subject of the method can be a network device (for example, a base station), or can be any other device used for model training, which is not limited here.

[0386] In an embodiment of the present application, the trained machine learning model is a two-layer auto-encoder, which can be used to jointly encode and jointly decode the CSI of multiple terminals in the same sub-area, that is, the input of the machine learning model is the CSI of multiple terminals in the same sub-area, and the output of the machine learning model is the CSI corresponding to each terminal after decoding.

[0387] In an embodiment of the present application, the coverage range of the network device can be divided into several sub-areas according to the characteristics of the propagation environment. When performing model training, the CSI data sets of each sub-area within the coverage range of the network device can be used to train the machine learning model, so that the machine learning model obtained after training can learn the propagation environment characteristics shared by the terminals in each sub-area within the coverage range of the network device.

[0388] In some embodiments, the method further comprises:

[0389] Sending a CSI reporting notification to each terminal in the sub-area;

[0390] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0391] Specifically, the device used for model training can notify the terminals within the coverage of the network device to report CSI, and form a CSI data set for each sub-area based on the collected CSI reported by the terminals in each sub-area. The CSI data set for each sub-area contains the CSI reported by the terminals in the corresponding sub-area.

[0392] The training method of the machine learning model provided in the embodiment of the present application trains the machine learning model by utilizing the CSI data sets of each sub-area within the coverage range of the network device, so that the machine learning model obtained after training can learn the propagation environment characteristics shared by the terminals in each sub-area within the coverage range of the network device. When the machine learning model is subsequently applied to CSI feedback, the propagation environment characteristics shared by multiple terminals in the same communication environment can be effectively utilized to reduce and compress the CSI to be fed back, remove redundant information, and ultimately achieve the purpose of reducing the CSI feedback overhead and improving the CSI feedback accuracy.

[0393] In some embodiments, the dual-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0394] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0395] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0396] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0397] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0398] Specifically, the two-layer autoencoder may include a two-layer encoder (Encoder) and a two-layer decoder (Decoder), and extract the propagation environment characteristics shared by multiple terminals through a joint encoder (Co-encoder) of the second layer.

[0399] like Figure 2 As shown, the architecture of the two-layer autoencoder includes N first-layer Encoders on the encoding side, 1 Co-encoder on the second layer on the encoding side, 1 Co-decoder on the first layer on the decoding side, and N second-layer Decoders on the decoding side.

[0400] Among them, the N Encoders in the first layer of the encoding side are responsible for compressing and encoding the CSI of each terminal respectively; the 1 Co-encoder in the second layer of the encoding side is responsible for jointly compressing and encoding all the input N terminals to extract the environmental information shared between the N terminals for CSI compression; the 1 Co-decoder in the first layer of the decoding side is responsible for jointly decoding the jointly encoded N terminal CSI to restore the environmental information shared between the N terminals; the N Decoders in the second layer of the decoding side are responsible for decoding the CSI unique information of each terminal to obtain the CSI of different terminals.

[0401] In some embodiments, the training of the machine learning model based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0402] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0403] Specifically, the arrangement order of the sub-regions can be arbitrary, as long as the two-layer autoencoder is trained in sequence according to the order after arrangement. This application does not limit the specific arrangement.

[0404] For example, the CSI data set of the first sub-area can be used to train a double-layer autoencoder. After the training reaches the convergence condition, the obtained machine learning model parameters are stored, and then the CSI data set of the second sub-area is used to further train the double-layer autoencoder trained with the CSI data set of the first sub-area. Similarly, after the training reaches the convergence condition, the obtained machine learning model parameters are stored, and then the CSI data set of the third sub-area is used to further train the double-layer autoencoder trained with the CSI data set of the second sub-area, and so on. The double-layer autoencoder is trained based on the CSI data set of each sub-area in turn until all sub-areas are traversed to obtain the final trained machine learning model. The machine learning model can learn the propagation environment characteristics shared by terminals in all sub-areas, so that it can be used for CSI feedback to effectively reduce feedback overhead and improve feedback accuracy.

[0405] In some embodiments, the method further comprises:

[0406] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0407] For example, after the double-layer autoencoder is trained using the CSI data set of the first sub-region, the shared environment information of the first sub-region can be extracted based on the CSI data set of the first sub-region and the currently trained double-layer autoencoder.

[0408] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0409] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0410] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0411] Taking the extraction of shared environment information of the first sub-area as an example, assuming that the two-layer autoencoder includes N CSI encoders, the CSI of any N terminals in the CSI data set of the first sub-area can be input into the N CSI encoders respectively to obtain N encoding information, and then based on the N encoding information, the shared environment information of the first sub-area is obtained.

[0412] In some embodiments, the process of "inputting the CSI of any N terminals in the CSI data set of the first sub-area into N CSI encoders respectively to obtain N coding information" can be repeated multiple times to obtain N×m coding information (where m is the number of repetitions), and then based on these N×m coding information, the shared environment information of the first sub-area is obtained.

[0413] In some implementations, a typical value of the above-mentioned multiple coded information (N or N×m coded information) can be used as the shared environment information of the first sub-region. The typical value can be, for example, an average value, a median, a mode, or other value that can reflect the typical characteristics of the multiple coded information.

[0414] In some embodiments, the method further comprises:

[0415] Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

[0416] Specifically, after the two-layer autoencoder is trained, the CSI decoding model can be deployed on the network device side for decoding CSI feedback information, such as Figure 3As shown, the CSI decoding model deployed on the base station side includes a Co-decoder on the first layer of the decoding side and a Decoder on the second layer of the decoding side. The Decoder is used to decode the terminal-specific information to obtain the CSI of different terminals.

[0417] In some embodiments, the method further comprises:

[0418] Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder;

[0419] The CSI coding model is sent to terminals in each sub-area.

[0420] Specifically, after the training of the two-layer autoencoder is completed, the CSI coding model that needs to be deployed on the terminal side can be determined first, and then the determined deployed CSI coding model is sent to the terminals in each sub-area.

[0421] For example, the following are deployed on the terminal side: Figure 3 The CSI coding model shown includes an Encoder of the first layer on the coding side and a Co-encoder of the second layer on the coding side. The input of the Co-encoder includes the coding information output by the Encoder and the shared environment information of the area to which the terminal belongs.

[0422] In some embodiments, the method further comprises:

[0423] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0424] Specifically, after the training of the two-layer autoencoder is completed, the CSI coding model to be deployed on the terminal side can be sent to the terminal, and the shared environment information of the area to which the terminal belongs can be sent to the terminal.

[0425] Optionally, the shared environment information of all sub-areas within the coverage of the network device may be sent to the terminal; or, the area to which the terminal belongs may be determined according to the location information of the terminal, and then the shared environment information of the area to which the terminal belongs may be sent to the terminal.

[0426] In some embodiments, the machine learning model (a two-layer autoencoder, a CSI encoding model deployed on the terminal side, or a CSI decoding model deployed on the network device side) and the shared environment information of each sub-area can be updated regularly, and the updated machine learning model and shared environment information can be deployed to the corresponding devices.

[0427] In some embodiments, the method further comprises:

[0428] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0429] Specifically, the area to which the terminal belongs may change during the movement process, so the terminal can request updated shared environment information from the device used for model training in a timely manner according to its mobile state to ensure the accuracy of CSI feedback. For example, if the terminal determines that its position at a certain moment exceeds the scope of the original area and enters another area, the terminal can send a request message to the device used for model training, and the request message can include information about the other area, so that the device used for model training can send new shared environment information to the terminal, or the request message does not carry information about the other area. After receiving the request message, the device used for model training can obtain the location information of the terminal to determine the area to which the terminal belongs, and then send the shared environment information of the corresponding area to the terminal.

[0430] The methods provided in the embodiments of the present application are based on the same application concept, so the implementation of each method can refer to each other, and the repeated parts will not be repeated.

[0431] The following describes the methods provided in the above embodiments of the present application by using examples of specific application scenarios.

[0432] Example 1:

[0433] 1. The base station divides the coverage area into multiple sub-areas according to the propagation environment.

[0434] 2. The base station collects channel feature vectors of 10 terminals in each sub-area during a specified time period.

[0435] 3. Based on Figure 2 The machine learning model architecture shown and the collected data set are used to train the machine learning model. Each encoder and decoder uses the transformer structure until the model converges.

[0436] 4. According to Figure 3 The trained machine learning model is deployed on the terminal side and the base station side in a manner, that is, only a first-layer CSI encoder, shared information of each sub-area, and a second-layer joint encoder are deployed on the terminal side, and only a first-layer joint decoder and a CSI decoder corresponding to the terminal side are deployed on the base station side.

[0437] 5. The terminal determines which sub-area divided by the base station the terminal is in based on its own location information.

[0438] 6. Based on the machine learning model deployed on the terminal side and the shared information of the sub-area to which the terminal belongs, the terminal inputs the CSI that the terminal needs to feedback into the deployed two-layer autoencoder machine learning model to obtain the compressed CSI after the first layer encoding and the second layer joint encoding.

[0439] 7. The terminal sends the compressed CSI to the base station.

[0440] 8. The base station receives the compressed CSI.

[0441] 9. The base station inputs the received CSI into the machine learning model of the two-layer autoencoder deployed on the base station side. After joint decoding of the first layer and independent decoding of the second layer, the decoded CSI fed back by the terminal is obtained.

[0442] The methods and devices provided in the various embodiments of the present application are based on the same application concept. Since the methods and devices solve problems based on similar principles, the implementation of the devices and methods can refer to each other, and the repeated parts will not be repeated.

[0443] Figure 7 A schematic diagram of the structure of a terminal provided in an embodiment of the present application, such as Figure 7 As shown, the terminal includes a memory 720, a transceiver 710 and a processor 700; wherein the processor 700 and the memory 720 may also be arranged physically separately.

[0444] The memory 720 is used to store computer programs; the transceiver 710 is used to send and receive data under the control of the processor 700.

[0445] Specifically, the transceiver 710 is used to receive and send data under the control of the processor 700 .

[0446] Among them, Figure 7 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 700 and various circuits of memory represented by memory 720 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described in this application. The bus interface provides an interface. The transceiver 710 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which may include transmission media such as wireless channels, wired channels, and optical cables. For different user devices, the user interface 730 may also be an interface capable of externally and internally connecting required devices, and the connected devices include but are not limited to a keypad, a display, a speaker, a microphone, a joystick, and the like.

[0447] The processor 700 is responsible for managing the bus architecture and general processing, and the memory 720 can store data used by the processor 700 when performing operations.

[0448] The processor 700 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0449] The processor 700 calls the computer program stored in the memory 720 to execute any of the methods provided in the embodiments of the present application according to the obtained executable instructions, for example: inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into the CSI coding model to obtain compressed CSI; and sending the compressed CSI to the network device.

[0450] In some embodiments, the CSI coding model includes 1 CSI encoder and 1 joint encoder;

[0451] The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes:

[0452] Inputting the CSI to be fed back into the CSI encoder for encoding;

[0453] The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

[0454] In some embodiments, the CSI coding model is part of a two-layer autoencoder;

[0455] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0456] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0457] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0458] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0459] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0460] In some embodiments, the method further comprises:

[0461] Receive the CSI coding model and the shared environment information sent by the network device.

[0462] In some embodiments, the method further comprises:

[0463] According to the mobile state of the terminal, requesting the network device for the shared environment information of the new area to which the terminal belongs; or,

[0464] According to the moving state of the terminal, the shared environment information of the new area to which the terminal belongs is called locally.

[0465] Figure 8 One of the structural diagrams of the network device provided in the embodiment of the present application is as follows: Figure 8 As shown, the network device includes a memory 820, a transceiver 810 and a processor 800; wherein the processor 800 and the memory 820 may also be arranged physically separately.

[0466] The memory 820 is used to store computer programs; the transceiver 810 is used to send and receive data under the control of the processor 800.

[0467] Specifically, the transceiver 810 is used to receive and send data under the control of the processor 800 .

[0468] Among them, Figure 8 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 800 and memory represented by memory 820. The bus architecture may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described in this application. The bus interface provides an interface. The transceiver 810 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which may include a wireless channel, a wired channel, an optical cable, and other transmission media.

[0469] The processor 800 is responsible for managing the bus architecture and general processing, and the memory 820 can store data used by the processor 800 when performing operations.

[0470] The processor 800 may be a CPU, an ASIC, an FPGA or a CPLD, and the processor may also adopt a multi-core architecture.

[0471] The processor 800 calls the computer program stored in the memory 820 to execute any of the methods provided in the embodiments of the present application according to the obtained executable instructions, for example: receiving the compressed CSI sent by the terminal; inputting the compressed CSI into the CSI decoding model to obtain the CSI fed back by the terminal.

[0472] In some embodiments, the CSI decoding model includes 1 joint decoder and 1 CSI decoder;

[0473] The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes:

[0474] Inputting the compressed CSI into the joint decoder for joint decoding;

[0475] The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

[0476] In some embodiments, the CSI decoding model is part of a two-layer autoencoder;

[0477] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0478] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0479] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0480] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0481] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0482] In some embodiments, the method further comprises:

[0483] The two-layer autoencoder is trained based on the CSI data set of each sub-area within the coverage area of ​​the network device.

[0484] In some embodiments, the training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0485] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0486] In some embodiments, the method further comprises:

[0487] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0488] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0489] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0490] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0491] In some embodiments, the method further comprises:

[0492] Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

[0493] In some embodiments, the method further comprises:

[0494] Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder;

[0495] The CSI coding model is sent to terminals in each sub-area.

[0496] In some embodiments, the method further comprises:

[0497] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0498] In some embodiments, the method further comprises:

[0499] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0500] In some embodiments, the method further comprises:

[0501] Sending a CSI reporting notification to each terminal in the sub-area;

[0502] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0503] Fig. 9 The second structural diagram of the network device provided in the embodiment of the present application is as follows: Fig. 9 As shown, the network device includes a memory 920, a transceiver 910 and a processor 900; wherein the processor 900 and the memory 920 may also be arranged physically separately.

[0504] The memory 920 is used to store computer programs; the transceiver 910 is used to send and receive data under the control of the processor 900.

[0505] Specifically, the transceiver 910 is used to receive and send data under the control of the processor 900 .

[0506] Among them, Fig. 9 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 900 and memory represented by memory 920. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described in this application. The bus interface provides an interface. The transceiver 910 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which may include a wireless channel, a wired channel, an optical cable, and other transmission media.

[0507] The processor 900 is responsible for managing the bus architecture and general processing, and the memory 920 can store data used by the processor 900 when performing operations.

[0508] The processor 900 may be a CPU, an ASIC, an FPGA or a CPLD, and the processor may also adopt a multi-core architecture.

[0509] The processor 900 calls the computer program stored in the memory 920 to execute any of the methods provided in the embodiments of the present application according to the obtained executable instructions, for example: training a machine learning model based on a channel state information CSI data set of each sub-area within the coverage range of the network device; the machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding the CSI of multiple terminals in the same sub-area.

[0510] In some embodiments, the dual-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0511] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0512] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0513] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0514] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0515] In some embodiments, the training of the machine learning model based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0516] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0517] In some embodiments, the method further comprises:

[0518] After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

[0519] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0520] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0521] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0522] In some embodiments, the method further comprises:

[0523] Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

[0524] In some embodiments, the method further comprises:

[0525] Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder;

[0526] The CSI coding model is sent to terminals in each sub-area.

[0527] In some embodiments, the method further comprises:

[0528] The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

[0529] In some embodiments, the method further comprises:

[0530] According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

[0531] In some embodiments, the method further comprises:

[0532] Sending a CSI reporting notification to each terminal in the sub-area;

[0533] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0534] It should be noted here that the above-mentioned terminal and network device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

[0535] Fig.10 One of the structural diagrams of the CSI feedback device provided in the embodiment of the present application is as follows Fig.10 As shown, the device comprises:

[0536] The coding unit 1000 is used to input the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI;

[0537] The first sending unit 1010 is configured to send the compressed CSI to a network device.

[0538] In some embodiments, the CSI coding model includes 1 CSI encoder and 1 joint encoder;

[0539] The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes:

[0540] Inputting the CSI to be fed back into the CSI encoder for encoding;

[0541] The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

[0542] In some embodiments, the CSI coding model is part of a two-layer autoencoder;

[0543] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0544] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0545] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0546] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0547] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0548] In some embodiments, the apparatus further comprises:

[0549] The first receiving unit is used to receive the CSI coding model and the shared environment information sent by the network device.

[0550] In some embodiments, the apparatus further comprises:

[0551] a requesting unit, configured to request, from the network device, shared environment information of a new area to which the terminal belongs according to a mobile state of the terminal; or

[0552] The calling unit is used to call the shared environment information of the new area to which the terminal belongs from the local according to the moving state of the terminal.

[0553] Fig.11 The second structural diagram of the CSI feedback device provided in the embodiment of the present application is as follows: Fig.11 As shown, the device comprises:

[0554] The second receiving unit 1100 is configured to receive the compressed CSI sent by the terminal;

[0555] The decoding unit 1110 is used to input the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal.

[0556] In some embodiments, the CSI decoding model includes 1 joint decoder and 1 CSI decoder;

[0557] The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes:

[0558] Inputting the compressed CSI into the joint decoder for joint decoding;

[0559] The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

[0560] In some embodiments, the CSI decoding model is part of a two-layer autoencoder;

[0561] The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0562] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0563] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0564] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0565] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0566] In some embodiments, the apparatus further comprises:

[0567] The first training unit is used to train the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device.

[0568] In some embodiments, the training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0569] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0570] In some embodiments, the apparatus further comprises:

[0571] The first extraction unit is used to extract the shared environment information of the sub-region based on the CSI dataset of the sub-region and the trained double-layer autoencoder after training the double-layer autoencoder based on the CSI dataset of the sub-region each time.

[0572] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0573] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0574] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0575] In some embodiments, the apparatus further comprises a first deployment unit configured to:

[0576] Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

[0577] In some embodiments, the apparatus further comprises a second deployment unit, configured to:

[0578] Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder;

[0579] The CSI coding model is sent to terminals in each sub-area.

[0580] In some embodiments, the apparatus further comprises:

[0581] The second sending unit is configured to send the shared environment information of the corresponding sub-area to the terminals in each sub-area.

[0582] In some embodiments, the apparatus further comprises:

[0583] The third sending unit is used to send the shared environment information of the new area to which the terminal belongs to the terminal according to the request information sent by the terminal.

[0584] In some embodiments, the device further comprises a first collecting unit, configured to:

[0585] Sending a CSI reporting notification to each terminal in the sub-area;

[0586] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0587] Fig.12A structural diagram of a training device for a machine learning model provided in an embodiment of the present application, such as Fig.12 As shown, the device comprises:

[0588] The second training unit 1200 is used to train a machine learning model based on a channel state information CSI data set of each sub-area within the coverage area of ​​the network device;

[0589] The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

[0590] In some embodiments, the dual-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1;

[0591] The N CSI encoders are used to encode the CSIs of the N terminals respectively;

[0592] The joint encoder is used to jointly encode the coded information output by the N CSI encoders;

[0593] The joint decoder is used to jointly decode the joint coded information output by the joint encoder;

[0594] The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

[0595] In some embodiments, the training of the machine learning model based on the CSI data set of each sub-area within the coverage area of ​​the network device includes:

[0596] According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

[0597] In some embodiments, the apparatus further comprises:

[0598] The second extraction unit is used to extract the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained double-layer autoencoder after training the double-layer autoencoder based on the CSI data set of the sub-area each time.

[0599] In some embodiments, extracting the shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes:

[0600] Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information;

[0601] Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

[0602] In some embodiments, the apparatus further comprises a third deployment unit, configured to:

[0603] Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

[0604] In some embodiments, the apparatus further comprises a fourth deployment unit, configured to:

[0605] Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder;

[0606] The CSI coding model is sent to terminals in each sub-area.

[0607] In some embodiments, the apparatus further comprises:

[0608] The fourth sending unit is configured to send the shared environment information of the corresponding sub-area to the terminals in each sub-area.

[0609] In some embodiments, the apparatus further comprises:

[0610] The fifth sending unit is used to send the shared environment information of the new area to which the terminal belongs to the terminal according to the request information sent by the terminal.

[0611] In some embodiments, the device further comprises a second collecting unit, configured to:

[0612] Sending a CSI reporting notification to each terminal in the sub-area;

[0613] Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

[0614] It should be noted that the division of units in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0615] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0616] It should be noted here that the above-mentioned device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0617] On the other hand, an embodiment of the present application also provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute the CSI feedback method or the machine learning model training method provided in the above embodiments.

[0618] It should be noted here that the non-transitory readable storage medium provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0619] The non-transitory readable storage medium can be any available medium or data storage device that can be accessed by a computer, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.

[0620] The technical solution provided in the embodiment of the present application can be applicable to a variety of systems, especially 5G systems or future 6G systems. For example, the applicable system can be a global system of mobile communication (GSM) system, a code division multiple access (CDMA) system, a wideband code division multiple access (WCDMA) general packet radio service (GPRS) system, a long term evolution (LTE) system, a LTE frequency division duplex (FDD) system, a LTE time division duplex (TDD) system, an advanced long term evolution (LTE-A) system, a universal mobile telecommunication system (UMTS), a world-wide interoperability for microwave access (WiMAX) system, a 5G new radio (NR) system, a 6G system, etc. These various systems include terminal equipment and network equipment. The system may also include core network parts, such as the Evolved Packet System (EPS), 5G System (5GS), etc.

[0621] The terminal involved in the embodiment of the present application may be a device that provides voice and / or data connectivity to a user, a handheld device with a wireless connection function, or other processing devices connected to a wireless modem. In different systems, the name of the terminal may also be different. For example, in a 5G system, the terminal may be called a user equipment (UE). A wireless terminal device can communicate with one or more core networks (CN) via a radio access network (RAN). The wireless terminal device may be a mobile terminal device, such as a mobile phone (or a "cellular" phone) and a computer with a mobile terminal device. For example, it may be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device that exchanges language and / or data with a wireless access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs) and other devices. The wireless terminal device may also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, an access point, a remote terminal device, an access terminal device, a user terminal device, a user agent, and a user device, but is not limited in the embodiments of the present application.

[0622] The network device involved in the embodiment of the present application may be a base station, which may include multiple cells providing services for the terminal. Depending on the specific application scenario, the base station may also be called an access point, or may be a device in the access network that communicates with the wireless terminal device through one or more sectors on the air interface, or other names. The network device may be used to interchange received air frames with Internet Protocol (IP) packets, and serve as a router between the wireless terminal device and the rest of the access network, wherein the rest of the access network may include an Internet Protocol (IP) communication network. The network device may also coordinate the attribute management of the air interface. For example, the network device involved in the embodiments of the present application may be a network device (Base Transceiver Station, BTS) in the Global System for Mobile communications (Global System for Mobile communications, GSM) or Code Division Multiple Access (Code Division Multiple Access, CDMA), or a network device (NodeB) in Wide-band Code Division Multiple Access (WCDMA), or an evolutionary network device (evolutional Node B, eNB or e-NodeB) in the long term evolution (long term evolution, LTE) system, a 5G base station (gNB) in the 5G network architecture (next generation system), or a home evolved Node B (Home evolved Node B, HeNB), a relay node, a home base station (femto), a pico base station (pico), etc., which is not limited in the embodiments of the present application. In some network structures, the network device may include a centralized unit (CU) node and a distributed unit (DU) node, and the centralized unit and the distributed unit may also be arranged geographically separately.

[0623] Network devices and terminals can each use one or more antennas for multiple-input multiple-output (MIMO) transmission. MIMO transmission can be single-user MIMO (SU-MIMO) or multi-user MIMO (MU-MIMO). Depending on the form and number of antenna combinations, MIMO transmission can be 2D-MIMO, 3D-MIMO, FD-MIMO or massive-MIMO, or it can be diversity transmission, precoded transmission or beamforming transmission, etc.

[0624] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0625] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer executable instructions. These computer executable instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0626] These processor executable instructions may also be stored in a processor readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the processor readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0627] These processor-executable instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0628] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A channel state information (CSI) feedback method, characterized in that: Applied to terminals, including: Inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI; The compressed CSI is sent to the network device.

2. The CSI feedback method according to claim 1, characterized in that: The CSI coding model includes 1 CSI encoder and 1 joint encoder; The step of inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes: Inputting the CSI to be fed back into the CSI encoder for encoding; The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

3. The CSI feedback method according to claim 2, characterized in that: The CSI coding model is part of a two-layer autoencoder; The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1; The N CSI encoders are used to encode the CSIs of the N terminals respectively; The joint encoder is used to jointly encode the coded information output by the N CSI encoders; The joint decoder is used to jointly decode the joint coded information output by the joint encoder; The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

4. The CSI feedback method according to any one of claims 1 to 3, characterized in that: The method further comprises: Receive the CSI coding model and the shared environment information sent by the network device.

5. The CSI feedback method according to any one of claims 1 to 3, characterized in that: The method further comprises: According to the mobile state of the terminal, requesting the network device for the shared environment information of the new area to which the terminal belongs; or, According to the moving state of the terminal, the shared environment information of the new area to which the terminal belongs is called locally.

6. A channel state information (CSI) feedback method, characterized in that: Applied to network equipment, including: receiving the compressed CSI sent by the terminal; The compressed CSI is input into a CSI decoding model to obtain the CSI fed back by the terminal.

7. The CSI feedback method according to claim 6, characterized in that: The CSI decoding model includes a joint decoder and a CSI decoder; The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes: Inputting the compressed CSI into the joint decoder for joint decoding; The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

8. The CSI feedback method according to claim 7, characterized in that: The CSI decoding model is part of a two-layer autoencoder; The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1; The N CSI encoders are used to encode the CSIs of the N terminals respectively; The joint encoder is used to jointly encode the coded information output by the N CSI encoders; The joint decoder is used to jointly decode the joint coded information output by the joint encoder; The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

9. The CSI feedback method according to claim 8, characterized in that: The method further comprises: The two-layer autoencoder is trained based on the CSI data set of each sub-area within the coverage area of ​​the network device.

10. The CSI feedback method according to claim 9, characterized in that: The training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes: According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

11. The CSI feedback method according to claim 10, characterized in that: The method further comprises: After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

12. The CSI feedback method according to claim 11, characterized in that: The extracting shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes: Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information; Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

13. The CSI feedback method according to claim 9 or 10, characterized in that: The method further comprises: Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

14. The CSI feedback method according to any one of claims 9 to 12, characterized in that: The method further comprises: Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder; The CSI coding model is sent to terminals in each sub-area.

15. The CSI feedback method according to claim 14, characterized in that: The method further comprises: The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

16. The CSI feedback method according to claim 11 or 12, characterized in that: The method further comprises: According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

17. The CSI feedback method according to any one of claims 9 to 12, characterized in that: The method further comprises: Sending a CSI reporting notification to each terminal in the sub-area; Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

18. A method for training a machine learning model, characterized in that: include: Train the machine learning model based on the channel state information (CSI) dataset of each sub-area within the coverage area of ​​the network device; The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

19. The method for training a machine learning model according to claim 18, wherein: The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1; The N CSI encoders are used to encode the CSIs of the N terminals respectively; The joint encoder is used to jointly encode the coded information output by the N CSI encoders; The joint decoder is used to jointly decode the joint coded information output by the joint encoder; The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

20. The method for training a machine learning model according to claim 18 or 19, characterized in that: The CSI data set of each sub-area within the coverage area of ​​the network device is used to train the machine learning model, including: According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

21. The method for training a machine learning model according to claim 20, characterized in that: The method further comprises: After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

22. The method for training a machine learning model according to claim 21, characterized in that: The extracting shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes: Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information; Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

23. The method for training a machine learning model according to claim 19, wherein: The method further comprises: Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

24. The method for training a machine learning model according to claim 19, wherein: The method further comprises: Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder; The CSI coding model is sent to terminals in each sub-area.

25. The method for training a machine learning model according to claim 24, characterized in that: The method further comprises: The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

26. The method for training a machine learning model according to claim 21 or 22, characterized in that: The method further comprises: According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

27. The method for training a machine learning model according to claim 18 or 19, characterized in that: The method further comprises: Sending a CSI reporting notification to each terminal in the sub-area; Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

28. A terminal, characterized in that: Including memory, transceiver, processor; A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations: Inputting the channel state information CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI; The compressed CSI is sent to the network device.

29. The terminal according to claim 28, characterized in that The CSI coding model includes 1 CSI encoder and 1 joint encoder; The step of inputting the channel state information CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI includes: Inputting the CSI to be fed back into the CSI encoder for encoding; The coding information output by the CSI encoder and the shared environment information are input into the joint encoder for joint coding to obtain the compressed CSI.

30. The terminal according to claim 29, characterized in that The CSI coding model is part of a two-layer autoencoder; The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1; The N CSI encoders are used to encode the CSIs of the N terminals respectively; The joint encoder is used to jointly encode the coded information output by the N CSI encoders; The joint decoder is used to jointly decode the joint coded information output by the joint encoder; The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

31. The terminal according to any one of claims 28 to 30, characterized in that: The operations also include: Receive the CSI coding model and the shared environment information sent by the network device.

32. The terminal according to any one of claims 28 to 30, characterized in that: The operations also include: According to the mobile state of the terminal, requesting the network device for the shared environment information of the new area to which the terminal belongs; or, According to the moving state of the terminal, the shared environment information of the new area to which the terminal belongs is called locally.

33. A network device, characterized in that: Including memory, transceiver, processor; a memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; A processor is configured to read the computer program in the memory and perform the following operations: receiving the compressed CSI sent by the terminal; The compressed CSI is input into a CSI decoding model to obtain the CSI fed back by the terminal.

34. The network device according to claim 33, characterized in that: The CSI decoding model includes a joint decoder and a CSI decoder; The step of inputting the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal includes: Inputting the compressed CSI into the joint decoder for joint decoding; The joint decoding information output by the joint decoder is input into the CSI decoder for decoding, and the CSI fed back by the terminal is output.

35. The network device according to claim 34, characterized in that: The CSI decoding model is part of a two-layer autoencoder; The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1; The N CSI encoders are used to encode the CSIs of the N terminals respectively; The joint encoder is used to jointly encode the coded information output by the N CSI encoders; The joint decoder is used to jointly decode the joint coded information output by the joint encoder; The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

36. The network device according to claim 35, characterized in that The operations also include: The two-layer autoencoder is trained based on the CSI data set of each sub-area within the coverage area of ​​the network device.

37. The network device according to claim 36, characterized in that: The training of the two-layer autoencoder based on the CSI data set of each sub-area within the coverage area of ​​the network device includes: According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

38. The network device according to claim 37, characterized in that: The operations also include: After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

39. The network device according to claim 38, characterized in that: The extracting shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes: Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information; Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

40. The network device according to claim 36 or 37, characterized in that: The operations also include: Based on the trained two-layer autoencoder, the CSI decoding model is deployed on the network device side.

41. The network device according to any one of claims 36 to 39, characterized in that: The operations also include: Determine a CSI coding model deployed on the terminal side based on the trained two-layer autoencoder; The CSI coding model is sent to terminals in each sub-area.

42. The network device according to claim 41, characterized in that The operations also include: The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

43. The network device according to claim 38 or 39, characterized in that: The operations also include: According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

44. The network device according to any one of claims 36 to 39, characterized in that: The operations also include: Sending a CSI reporting notification to each terminal in the sub-area; Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

45. A network device, characterized in that: Including memory, transceiver, processor; A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations: Training a machine learning model based on a channel state information (CSI) data set of each sub-area within the coverage area of ​​the network device; The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

46. ​​The network device according to claim 45, characterized in that The double-layer autoencoder includes N CSI encoders, 1 joint encoder, 1 joint decoder and N CSI decoders, where N is an integer greater than 1; The N CSI encoders are used to encode the CSIs of the N terminals respectively; The joint encoder is used to jointly encode the coded information output by the N CSI encoders; The joint decoder is used to jointly decode the joint coded information output by the joint encoder; The N CSI decoders are used to decode the joint decoding information output by the joint decoder, and output the CSI of the N terminals respectively.

47. The network device according to claim 45 or 46, characterized in that: The CSI data set of each sub-area within the coverage area of ​​the network device is used to train the machine learning model, including: According to the arrangement order of the sub-regions, the two-layer autoencoder is trained based on the CSI data set of each sub-region in turn until all the sub-regions are traversed.

48. The network device according to claim 47, characterized in that The operations also include: After the two-layer autoencoder is trained based on the CSI data set of one sub-region each time, the shared environment information of the sub-region is extracted based on the CSI data set of the sub-region and the trained two-layer autoencoder.

49. The network device according to claim 48, characterized in that The extracting shared environment information of the sub-area based on the CSI data set of the sub-area and the trained two-layer autoencoder includes: Inputting the CSI in the CSI data set of the sub-region into the trained CSI encoder of the double-layer autoencoder respectively to obtain multiple encoding information; Based on the multiple pieces of encoding information, shared environment information of the sub-areas is obtained.

50. The network device according to claim 46, characterized in that The operations also include: Based on the trained two-layer autoencoder, a CSI decoding model is deployed on the network device side, and the CSI decoding model includes one joint decoder and one CSI decoder.

51. The network device according to claim 46, characterized in that The operations also include: Based on the trained two-layer autoencoder, determine a CSI coding model deployed on the terminal side, where the CSI coding model includes one CSI encoder and one joint encoder; The CSI coding model is sent to terminals in each sub-area.

52. The network device according to claim 51, characterized in that The operations also include: The shared environment information of the corresponding sub-area is sent to the terminals in each sub-area.

53. The network device according to claim 48 or 49, characterized in that: The operations also include: According to the request information sent by the terminal, the shared environment information of the new area to which the terminal belongs is sent to the terminal.

54. The network device according to claim 45 or 46, characterized in that: The operations also include: Sending a CSI reporting notification to each terminal in the sub-area; Based on the CSI reported by the terminals in each of the sub-areas, a CSI data set of the corresponding sub-area is formed.

55. A channel state information (CSI) feedback device, characterized in that: include: A coding unit, used for inputting the CSI to be fed back and the shared environment information of the area to which the terminal belongs into a CSI coding model to obtain compressed CSI; The first sending unit is configured to send the compressed CSI to a network device.

56. A channel state information (CSI) feedback device, characterized in that: include: A second receiving unit, configured to receive the compressed CSI sent by the terminal; The decoding unit is used to input the compressed CSI into a CSI decoding model to obtain the CSI fed back by the terminal.

57. A training device for a machine learning model, characterized in that: include: A second training unit is used to train a machine learning model based on a channel state information CSI data set of each sub-area within the coverage area of ​​the network device; The machine learning model is a two-layer autoencoder for jointly encoding and jointly decoding CSIs of multiple terminals in the same sub-area.

58. A non-transitory readable storage medium, characterized in that The non-transitory readable storage medium stores a computer program, and the computer program is used to enable a processor to execute the method according to any one of claims 1 to 5.

59. A non-transitory readable storage medium, characterized in that: The non-transitory readable storage medium stores a computer program, and the computer program is used to enable a processor to execute the method according to any one of claims 6 to 17.

60. A non-transitory readable storage medium, characterized in that The non-transitory readable storage medium stores a computer program, and the computer program is used to cause a processor to execute the method according to any one of claims 18 to 27.

Images