Joint Beamforming and Trajectory Optimization for UAV-Assisted Double IRS Secure Transmission System: A Deep Reinforcement Learning Approach

Abstract

Unmanned aerial vehicle (UAV) integrated with intelligent reflecting surface (IRS) has excellent potential to improve air-to-ground communication performance. However, the openness of the air-to-ground channel makes secure information transmission a challenging issue. In this paper, we first propose a UAV-assisted double IRS secure transmission system, where one IRS is carried by UAV, and the other is deployed near the base station. The collaborative beamforming gain provided by the cascaded reflection link overcomes the performance bottleneck of a single IRS in the air. Secondly, with the aim of maximizing the sum secrecy rate, a problem is formulated for jointly optimizing the trajectory of UAV and the phase shift of each IRS. As the formulated problem is a non-convex optimization problem and varies dynamically with the environment, a deep reinforcement learning-based secure transmission approach is presented to continuously achieve the optimal phase shift and trajectory of UAV in dynamic environments. Furthermore, we reduce the dimension of the action space by dividing the IRS into several sub-surfaces, each of which shares the same phase shift. The simulation results demonstrate that the secrecy rate of the UAV-assisted double IRS secure transmission system can be significantly improved by the proposed approach.