Adaptive Lightweight CNN-Based CSI Feedback for Massive MIMO Systems

Abstract

Massive multiple-input multiple-output (MIMO) is one of the most promising technologies for a user equipment (UE) to achieve a high data rate. However, massive MIMO requires channel state information (CSI) at the transmitter and the CSI overhead fed back by UEs exponentially increases as the number of antennas increases. In the last years, many studies have been conducted to solve the problem of enormous CSI feedback overhead by utilizing deep learning. In this letter, we propose an adaptive lightweight convolutional neural network (CNN) in the deep learning-based MIMO CSI feedback. The proposed network adaptively finds the compression ratio to be used in the network and reduces the computational complexity of the network. Simulation results show that the proposed lightweight CNN significantly reduces the computational complexity in comparison with the conventional CsiNet while achieving the equivalent performance; and moreover the proposed network converges faster.