Deep Learning-Based Cooperative CSI Feedback via Multiple Receiving Antennas in Massive MIMO

Abstract

Deep learning (DL)-based channel state information (CSI) feedback can provide high downlink CSI reconstruction accuracy for massive multiple-input multiple-output (MIMO) systems. However, existing DL-based CSI feedback methods assume that a single receiving antenna is deployed at the user equipment side so that they ignore the CSI correlation among multiple receiving antennas. In this paper, we propose a (m)ultiple (r)eceiving antennas cooperative CSI feedback network, named MRNet, with 3-dimensional convolutional layers to extract the correlation features among different receiving antennas. We further propose a layer reuse scheme for the improvement of storage efficiency of MRNet in the feedback system. Simulation results show that, via the cooperative CSI feedback of multiple receiving antennas, MRNet outperforms DL-based single receiving antenna CSI feedback approaches by at most 5.27dB in terms of the reconstruction accuracy. Moreover, MRNet is applicable to massive MIMO systems with different numbers of receiving antennas while the number of trainable parameters of the network is unchanged. The open source codes of MRNet are available at https://github.com/ch28/MRNet