Performance Analysis of RIS-Assisted Wireless Networks with Q-Learning-Based Phase Shift Optimization

Reconfigurable Intelligent Surfaces (RIS) is a promising technology for improving wireless communication reliability. This paper investigates a RIS-assisted downlink massive Multiple-Input Multiple-Output (MIMO) communication system that suffers severe blockage and interference between cells. We presume that the fading heavily weakened the direct link from the Base Station (BS) to the User Equipment (UE). Therefore, the path reflected by the RIS enhances the weakened direct link between the base station and user equipment by effectively integrating with the receiver. We utilize Q-learning algorithm to approximate the RIS phase shifts without relying on complex optimization methods. Thanks to Q-learning, the RIS controller learns the best phase configuration by interacting with the wireless propagation environment and maximizing the instantaneous Signal-to-Interference-plus-Noise Ratio (SINR). We perform Monte Carlo simulations to study the system capacity and user bit rate as a function of the number of RIS elements. The results show that the Q-learning-based RIS-assisted transmission outperforms random phase configurations and BS-UE direct communications.