Deep Reinforcement Learning for Interference Suppression in RIS-Aided High-Speed Railway Networks

Abstract

This paper investigates the reconfigurable intelligent surface (RIS)-aided high-speed railway (HSR) network, where one RIS is deployed nearby the onboard mobile relay (MR) to suppress the external interference in HSR system. In order to enhance the HSR network capacity against the interference, we formulate an optimization problem for designing the phase shifts at the RIS. Since the HSR environment is time-varying and complicated, the optimization problem is challenging to settle. Inspired by the recent advances of artificial intelligence (AI), we propose a deep reinforcement learning (DRL)-based scheme to design the RIS phase shifts. Simulation results show that 1) deploying the RIS nearby the onboard MR is strongly facilitative of suppressing the interference; 2) the proposed DRL scheme can achieve better capacity than the baseline schemes.