Energy-Efficiency Maximization of Multiple RISs-Enabled Communication Networks by Deep Reinforcement Learning

Abstract

Reconfigurable Intelligent Surfaces (RISs) have become an emerging paradigm to improve the average sum-rate, enhance energy efficiency and extend coverage areas in wireless communications. In this paper, a multiple RISs-enabled energy-efficient downlink communication system is investigated. Then, to maximize energy efficiency for the proposed system, the joint optimization problem of user-RIS association, reflective elements ON/OFF states, phase shift, and transmit power is formulated. However, as the formulated problem is mixed-integer, non-convex, and NP-hard, it is challenging to solve in polynomial time. To overcome the challenge, by using the Block Coordinate Descent (BCD) method, the formulated problem is decomposed into two sub-problems: 1) joint user-RIS association, reflective elements ON/OFF states, and phase shift problem, and 2) power control problem. Then, the deep reinforcement learning (DRL) algorithm and convex optimization technique are deployed in order to solve the decomposed sub-problems alternatively to find close optimal solutions. Finally, comprehensive simulation results are established to demonstrate the effectiveness of our proposed algorithms.