Multi-Agent DRL Based Transmission Optimization for Distributed RIS-Aided Communication Systems

Abstract

In this paper, we investigate the distributed reconfigurable intelligent surface (RIS)-aided downlink multi-input single-output communication systems. Specifically, to make the environment stationary and reduce the complexity, a multi-agent deep deterministic policy gradient (MADDPG) based algorithm is proposed to optimize transmissions by only considering statistical channel state information (CSI). In the proposed method, the base station (BS) agent optimizes the BS beamforming vector, while the RIS agents are responsible for the corresponding RIS phase shift matrix exploiting centralized training with decentralized execution (CTDE) framework. Simulation results show that, with much lower computational complexity, the proposed algorithm has similar performance with the alternative optimization (AO) method.