RIS Assisted Eavesdropping Avoidance for UAV Communication

Unmanned aerial vehicle (UAV) is a promising application in the field of wireless communications, where it can serve as a temporary base station to ground users. However, the wireless channel's characteristics make it susceptible to obstructions and eavesdropping. To mitigate these issues, reconfigurable intelligent surface (RIS) can be incorporated into UAV-based wireless communication systems to overcome line-of-sight (LoS) path blockages and increase system flexibility. This paper proposes a joint optimization scheme of UAV trajectory, RIS phase shift, and user scheduling to maximize the secure downlink transmission rate while accounting for obstacles and eavesdroppers. Yet the mixed-integer nature of the optimization problem with non convex properties, renders traditional approaches, such as successive convex approximation (SCA), inefficient. To this end, deep reinforcement learning (DRL) based algorithms are proposed to solve this problem effectively. Simulation results demonstrate that the proposed solution achieves a favorable secure downlink rate while completing the data transmission task.