Energy Efficiency Maximization for RIS-Assisted MISO Symbiotic Radio Systems Based on Deep Reinforcement Learning

Abstract

In this letter, we consider a reconfigurable intelligent surface (RIS)-assisted multiple-input single-output symbiotic radio (SR) system. In order to investigate short-packet transmissions in SR system, both direct and backscatter signal packets are considered to have finite blocklength. According to different modulation schemes on RIS, we investigate the ON/OFF-based reflection and binary phase shift keying modulation schemes, and attain the average achievable rates of the SR system. Then, an energy efficiency maximization optimization problem is formulated by jointly optimizing transmit beamforming vector at the base station and the phase shift matrix at the RIS. Finally, we develop a novel deep reinforcement learning (DRL)-based framework to solve this non-convex optimization problem. Simulation results show that the proposed DRL-based framework can achieve higher energy efficiency and better convergence performance than the current DRL-based framework.