SEE Maximization in RIS-Aided Network With RSMA: A PPO-SCF Method

Abstract

This letter investigates a reconfigurable intelligent surface (RIS)-aided multiple-input single-output (MISO) network with rate splitting multiple access (RSMA), where some specific applications serve paying users (PUs) while non-paying users (NUs) are regarded as potential eavesdroppers. An optimization problem is formulated to maximize the system’s secrecy energy efficiency (SEE), subject to the constraints on the achievable secrecy rate requirement of PUs, the system power budget and the secrecy decoding rate of the common stream, where the beamforming vectors, the artificial noise (AN) vectors, the secrecy common message rates, and the discrete phase shifts are jointly optimized. Due to the non-convexity of the problem with mixed continuous and discrete variables, a method called proximal policy optimization with smooth clipping function (PPO-SCF) is proposed, where a zero-penalty term is presented and integrated into the reward function to ensure the constraints are satisfied. Simulation results show that the proposed PPO-SCF outperforms the baselines in terms of SEE. Furthermore, the impact of numbers of RIS phase-shift elements and the spatial distribution of RIS coordinates on the SEE performance are also discussed.