Method and apparatus for CSI feedback performed by online learning-based UE-driven autoencoder

1 . A method of a terminal in a wireless communication system, the method comprising:

transmitting, to a base station, training capability information relating to artificial intelligence (AI) model training of an autoencoder configured to compress and reconstruct feedback information for a channel state information-reference signal (CSI-RS);

receiving, from the base station, information on a training completion time point and decoder information corresponding to the training capability information, wherein the decoder information includes decoder model information and decoder structure information;

generating a training dataset for the autoencoder based on at least one received CSI-RS;

training the autoencoder based on the decoder information, the information on the training completion time point, and the generated training dataset;

transmitting, to the base station, training result information of the autoencoder; and

in case that the training result information indicates a failure of autoencoder training, receiving, from the base station, information indicating a replacement of a decoder model or a replacement of a decoder structure.

2 . The method of claim 1 , wherein the training capability information comprises at least one of latency time information, information on a calculation capability available for training, or trainable model size information.

3 . The method of claim 1 , wherein the information on the training completion time point comprises at least one of CSI compression accuracy information, CSI reconstruction accuracy information relating to a minimum sum-rate, CSI reconstruction accuracy information for a minimum user perceived throughput (UPT), CSI reconstruction accuracy information corresponding to a discrete Fourier transform (DFT)-beam-based type 2 codebook level, CSI reconstruction accuracy information based on a number of CSI feedback bits of the terminal, a maximum latency time, or an indication indicating end of training in response to convergence of a training result.

4 . The method of claim 1 , wherein the training result information comprises at least one of a trained weight of a decoder, a one-bit signal indicating a success or a failure of training, CSI compression accuracy information relating to performed training, CSI reconstruction accuracy information relating to performed training, a number of epochs consumed in training, a time consumed for training, or a number of pieces of data used in training.

5 . The method of claim 4 , further comprising, in case that the one-bit signal indicates a failure of training, receiving, from the base station, at least one of an indication for increment of a number of pieces of training data or an indication for a decoder replacement, or re-training or additionally training information of the autoencoder based on an increased number of pieces of the training data or replaced decoder information.

6 . The method of claim 1 , further comprising:

transmitting a request for a pre-training dataset to the base station; and

receiving at least one pre-training dataset from the base station,

wherein the autoencoder is trained based on a sum of the generated training dataset and the pre-training dataset.

7 . The method of claim 6 , wherein the request for the pre-training dataset comprises request number information of the pre-training dataset, and

wherein the request number information is determined, by the terminal, based on a reference numerical value required for training the autoencoder and the training capability information.

8 . The method of claim 6 , wherein the pre-training dataset received from the base station is a dataset selected, by the base station, based on location information of the terminal.

9 . A terminal in a wireless communication system, the terminal comprising:

a transceiver configured to transmit and receive a signal; and

a controller coupled to the transceiver, the controller configured to:

transmit, to a base station, training capability information relating to artificial intelligence (AI) model training of an autoencoder configured to compress and reconstruct feedback information for a channel state information-reference signal (CSI-RS);

receive, from the base station, information on a training completion time point and decoder information corresponding to the training capability information;

generate a training dataset for the autoencoder, based on at least one received CSI-RS;

train the autoencoder, based on the decoder information, the information on the training completion time point, and the generated training dataset;

transmit, to the base station, training result information of the autoencoder; and

in case that the training result information indicates a failure of autoencoder training, receiving, from the base station, information associated with a replacement of a decoder model or a replacement of a decoder structure.

10 . The terminal of claim 9 , wherein the controller is further configured to:

transmit a request for a pre-training dataset to the base station; and

receive at least one pre-training dataset from the base station, and

wherein the autoencoder is trained based on a sum of the generated training dataset and the pre-training dataset.

11 . A method of a base station in a wireless communication system, the method comprising:

receiving, from a terminal, training capability information relating to artificial intelligence (AI) model training of an autoencoder configured to compress and reconstruct feedback information for a channel state information-reference signal (CSI-RS);

determining decoder information, based on the training capability information and transmitting, to the terminal, the determined decoder information and information on a training completion time point;

transmitting at least one CSI-RS to the terminal;

receiving training result information of the autoencoder from the terminal; and

in case that the training result information indicates a failure of autoencoder training, determining a replacement of a decoder model or a replacement of a decoder structure, and transmitting, to the terminal, information indicating the replacement of a decoder model or the replacement of a decoder structure.

12 . The method of claim 11 , wherein the training capability information comprises at least one of latency time information, information on a calculation capability available for training, or trainable model size information.

13 . The method of claim 11 , wherein the information on the training completion time point comprises at least one of CSI compression accuracy information, CSI reconstruction accuracy information relating to a minimum sum-rate, CSI reconstruction accuracy information for a minimum user perceived throughput (UPT), CSI reconstruction accuracy information corresponding to a discrete Fourier transform (DFT)-beam-based type 2 codebook level, CSI reconstruction accuracy information based on a number of CSI feedback bits of the terminal, a maximum latency time, or an indication indicating end of training in response to convergence of a training result.

14 . The method of claim 11 , wherein the training result information comprises at least one of a trained weight of a decoder, a one-bit signal indicating a success or a failure of training, CSI compression accuracy information relating to performed training, CSI reconstruction accuracy information relating to performed training, a number of epochs consumed in training, a time consumed for training, or a number of pieces of data used in training.

15 . The method of claim 14 , further comprising, in case that the one-bit signal indicates a failure of training, transmitting, to the terminal, at least one of an indication for increment of a number of pieces of training data or an indication for a decoder replacement.

16 . The method of claim 11 , further comprising:

receiving a request for a pre-training dataset from the terminal; and

transmitting at least one pre-training dataset to the terminal,

wherein the autoencoder is trained based on a sum of the training dataset generated by the terminal and the pre-training dataset.

17 . The method of claim 16 , wherein the request for the pre-training dataset comprises request number information of the pre-training dataset, and

wherein the request number information is determined, by the terminal, based on a reference numerical value required for training the autoencoder, and the training capability information.

18 . The method of claim 16 , wherein the pre-training dataset transmitted to the terminal is a dataset selected, by the base station, based on location information of the terminal.

19 . A base station in a wireless communication system, the base station comprising:

a transceiver configured to transmit and receive a signal; and

a controller coupled to the transceiver, the controller configured to:

receive, from a terminal, training capability information relating to artificial intelligence (AI) model training of an autoencoder configured to compress and reconstruct feedback information for a channel state information-reference signal (CSI-RS);

determine decoder information, based on the training capability information and transmit, to the terminal, the determined decoder information and information on a training completion time point;

receive training result information of the autoencoder from the terminal; and

in case that the training result information indicates a failure of autoencoder training, determine a replacement of a decoder model or a replacement of a decoder structure, and transmit, to the terminal, information indicating the replacement of a decoder model or the replacement of a decoder structure.

20 . The base station of claim 19 , wherein the controller is further configured to:

receive a request for a pre-training dataset from the terminal; and

transmit at least one pre-training dataset to the terminal, and

wherein the autoencoder is trained based on a sum of a training dataset generated by the terminal and the pre-training dataset.