A Learnable Optimization and Regularization Approach to Massive MIMO CSI Feedback

Channel state information (CSI) plays a critical role in achieving the potential benefits of massive multiple input output (MIMO) systems. In frequency division duplex (FDD) MIMO systems, base station (BS) relies on sustained and accurate CSI feedback from users. However, due to large number antennas users being served overhead can become bottleneck. this paper, we propose model-driven deep learning method for feedback, called learnable optimization regularization algorithm (LORA). Instead using l 1 -norm as term, LORA introduces module that adapts characteristics automatically. The conventional Iterative Shrinkage-Thresholding Algorithm (ISTA) is unfolded into neural network, which learn both process term by end-to-end training. We show improves accuracy speed. Besides, novel quantization corresponding training scheme are proposed, it shown operate successfully at different bit rates, providing flexibility terms overhead. Various realistic scenarios considered demonstrate effectiveness robustness through numerical simulations.