DRL-Based Robust Transmission for Sub-Connected Active RIS-Assisted Communications

This work investigates the performance of a sub-connected active reconfigurable intelligent surface (RIS)-assisted communications under imperfect channel state information (CSI). To provide robust transmission, we formulate an energy efficiency (EE) maximization problem that jointly optimizes the transmit precoder at the base station (BS) and the beamforming matrix at the RIS under the consideration of a norm-bounded CSI error model. Due to the non-convex nature of the problem, we solve it using deep reinforcement learning (DRL)-based deep deterministic policy gradient (DDPG) and proximal policy optimization (PPO) approaches to determine the optimal transmit precoder and beamforming matrix. We demonstrate the convergence and robustness of the proposed algorithms via extensive simulations. Furthermore, we highlight the impact of several key system parameters, such as the number of total elements, number of required amplifiers, and maximum available transmit power at BS, on the performance of the considered system. Finally, we compare the performance of the proposed algorithms to the traditional analytical optimization methods and validate their efficacy.