DRL-Based IRS-Assisted Secure Hybrid Visible Light and mmWave Communications

Abstract

This paper explores a new advancement in physical layer security (PLS) techniques, focusing on the integration of Intelligent Reflecting Surfaces (IRS). This work centers on developing an intelligent hybrid system combining communication lines using millimeter wave (mmWave) and Visible light communication (VLC). The system comprises four VLC access points with light fixtures, reinforced by a mirror array sheet, and a mmWave access point with antennas, supported by a reflecting unit sheet. Within the system, both sheets function as IRS. The aim is to enhance the secrecy capacity (SC) of the system by optimizing the beamforming weights at the VLC fixtures, the beamforming weights at the mmWave AP, the mirror array configurations, and the phase shift vector while meeting specific power constraints. Given the numerous variables and the dynamic nature of user mobility, traditional optimization techniques may be inadequate for improving SC. To address this complexity optimally, we propose a deep reinforcement learning (DRL) approach based on the deep deterministic policy gradient (DDPG) technique. The DDPG algorithm can adapt to channel variations due to user movement and high-dimensional factors. Furthermore, it intelligently selects the optimal technique to improve SC, whether VLC or RF. Simulation results confirm the efficacy of our approach in enhancing the SC for the authorized receiver, particularly in mmWave connections.

Keywords

• Visible light communications
• millimeter wave
• deep learning
• machine learning
• deep reinforcement learning
• intelligent reflecting surfaces