Abstract:The application of deep learning (DL)-based channel state information (CSI) feedback frameworks in massive multiple-input multiple-output (MIMO) systems has significantly improved reconstruction accuracy. However, the limited generalization of widely adopted autoencoder-based networks for CSI feedback challenges consistent performance under dynamic wireless channel conditions and varying communication overhead constraints. To enhance the robustness of DL-based CSI feedback across diverse channel scenarios, we propose a novel framework, ITUG, where the user equipment (UE) transmits only a selected portion of critical values in the CSI matrix, while a generative model deployed at the BS reconstructs the remaining values. Specifically, we introduce a scoring algorithm to identify important values based on amplitude and contrast, an encoding algorithm to convert these values into a bit stream for transmission using adaptive bit length and a modified Huffman codebook, and a Transformer-based generative network named TPMVNet to recover the untransmitted values based on the received important values. Experimental results demonstrate that the ITUG framework, equipped with a single TPMVNet, achieves superior reconstruction performance compared to several high-performance autoencoder models across various channel conditions.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to effectively reduce the communication overhead of channel state information (CSI) feedback in massive multiple - input multiple - output (Massive MIMO) systems while maintaining or improving the accuracy of CSI reconstruction. Specifically, the paper points out that the traditional deep - learning methods based on auto - encoders have the following problems when dealing with CSI feedback: 1. **Limited generalization ability**: Existing auto - encoder - based networks are sensitive to data changes and difficult to adapt to natural data drift in wireless communication, resulting in inconsistent performance in different environments. 2. **High communication overhead**: In frequency - division duplex (FDD) systems, the dimension of the CSI matrix is very high, and direct transmission will lead to huge communication overhead. 3. **Complexity of model training and deployment**: In order to improve the generalization ability of the model, a large amount of diverse data is required for training, which is often difficult to achieve in practical applications. To solve these problems, the paper proposes a new CSI feedback framework - ITUG (Important Transmitting and Unimportant Generating). The main innovation points of this framework are as follows: 1. **Selection and transmission of important values**: The user equipment (UE) only transmits a small number of key values in the CSI matrix, which are selected by a magnitude - and - contrast - based scoring algorithm (MCS). This method can significantly reduce the amount of transmitted data, thereby reducing communication overhead. 2. **Generation of untransmitted values**: The base station (BS) uses a Transformer - based generation model (TPMVNet) to reconstruct the untransmitted CSI values according to the received important values. This can improve the reconstruction accuracy of CSI while ensuring low communication overhead. 3. **Efficient encoding and decoding**: To further reduce the number of transmitted bits, the paper designs an adaptive quantization and modified Huffman coding (AQ - MH) method for converting important values and their positions into binary sequences for transmission. Through the above methods, the ITUG framework can achieve better reconstruction performance than existing high - performance auto - encoder models under various channel conditions while significantly reducing communication overhead.

A CSI Feedback Framework based on Transmitting the Important Values and Generating the Others

M-Net: A Lightweight Network Based on Multilayer Perceptron for Massive MIMO CSI Feedback

Viewing the MIMO Channel As Sequence Rather Than Image: A Seq2Seq Approach for Efficient CSI Feedback

Scalable Framework For Deep Learning based CSI Feedback

Deep Learning for Efficient CSI Feedback in Massive MIMO: Adapting to New Environments and Small Datasets

Multi-resolution CSI Feedback with Deep Learning in Massive MIMO System

Viewing Channel as Sequence Rather Than Image: A 2-D Seq2Seq Approach for Efficient MIMO-OFDM CSI Feedback

TransNet: Full Attention Network for CSI Feedback in FDD Massive MIMO System

Joint CSI Feedback and Prediction with Deep Learning in High-speed Scenarios

CSI-PPPNet: A One-Sided One-for-All Deep Learning Framework for Massive MIMO CSI Feedback

Multi-task Learning-based CSI Feedback Design in Multiple Scenarios

Lightweight Convolutional Neural Networks for CSI Feedback in Massive MIMO

A Low-Overhead Incorporation-Extrapolation based Few-Shot CSI Feedback Framework for Massive MIMO Systems

Changeable Rate and Novel Quantization for CSI Feedback Based on Deep Learning

An Efficient Deep Learning Framework for Low Rate Massive MIMO CSI Reporting

Deep Learning-based CSI Feedback for RIS-assisted Multi-user Systems

Binary Neural Network Aided CSI Feedback in Massive MIMO System

Deep Learning for CSI Feedback: One-Sided Model and Joint Multi-Module Learning Perspectives

Training Enhancement of Deep Learning Models for Massive MIMO CSI Feedback with Small Datasets

Optimizing Feedback User Selection for Standalone U-LTE Networks with Randomly Delayed CSI

Generalizing Deep Learning-Based CSI Feedback in Massive MIMO via ID-Photo-Inspired Preprocessing