Energy Harvesting Design for Cooperative Reconfigurable Intelligent Surface with Multi-Agent Deep Reinforcement Learning

Cooperative reconfigurable intelligent surface (Cooperative- RIS)-assisted communication systems have been proven to provide more smart computing and communication ability than single-RIS or multi-distributed-RIS systems regarding beamforming gain and channel diversity. However, the increased number of deployed RISs brings heavier energy consumption, which makes the cooperative- RIS system less sustainable. The majority of previous works mainly study energy harvesting (EH) for single- RIS systems which cannot be directly applied to the cooperative- RIS systems. In this paper, we study the EH for cooperative- RIS-assisted communication systems. Specifically, we first model the cooperative beamforming and EH optimization as a non-convex problem. Besides, we propose a multi-agent deep deterministic policy gradient (MADDPG)-based EH efficiency maximization scheme for cooperative RISs while satisfying an acceptable sum rate. Finally, experimental results show the effectiveness of our proposed scheme.