Joint Optimization of Reconfigurable Intelligent Surfaces and Base Station Beamforming in MISO System Based on Deep Reinforcement Learning

To solve the communication system throughput demand due to the rapid growth of communication devices and the blocking problem of millimeter-wave (mmWave) communication, reconfigurable intelligent surface (RIS) technology is used to improve communication quality. In the RIS-assisted multiple-users wireless communication system, we investigate the joint optimization problem of the base station (BS) beamforming and the RIS phase control to maximize the sum rate of the system. To be more practical, we consider the RIS cell's gain and the antenna's radiation pattern in the RIS-assisted wireless channel model. We apply a deep deterministic policy gradient (DDPG) algorithm based on deep reinforcement learning (DRL) to solve the nonconvex joint optimization problem. The parameters that affect the convergence effect of the proposed algorithm are discussed. Simulation results show that the phase discretization of the RIS cell decreases the sum rate. Furthermore, we find the DDPG algorithm obtains a higher system sum rate than the benchmark algorithm.