Aerial Reconfigurable Intelligent Surface-Assisted Secrecy Energy-Efficient Communication Based on Deep Reinforcement Learning

This paper investigates the issue of physical layer security (PLS) in aerial reconfigurable intelligent surface (ARIS)-assisted millimeter-wave (mmWave) communications, with a specific focus on maximizing the secrecy energy efficiency (SEE) for all legitimate users while considering the presence of potential eavesdroppers. Our objective is to jointly optimize the active beamforming of the base station (BS), passive beamforming of ARIS, and ARIS flight trajectory to maximize SEE. However, the channel state information (CSI) is intricately coupled with the ARIS trajectory, thereby significantly augmenting both computational complexity and design challenges. Deep reinforcement learning (DRL) can make realtime decision at each time slot by effectively interacting with the dynamically evolving environment. Therefore, to address these sophisticated challenges effectively, we propose a dual-proximal policy optimization (D-PPO) algorithm that decouples continuous optimization variables. Simulation results demonstrate that our proposed algorithm outperforms traditional dual-deep deterministic policy gradient (D-DDPG) algorithm in achieving greater SEE.