Deep reinforcement learning enabled UAV-IRS-assisted secure mobile edge computing network

The deployment of intelligent reflecting surfaces (IRS) on dynamically moving unmanned aerial vehicles (UAVs) can enhance the communication performance of mobile edge computing (MEC), improve the system flexibility, and alleviate eavesdropping on air-ground channels. In this paper, an IRS-equipped unmanned aerial vehicle (UAV)-assisted secure MEC network is proposed. By jointly optimizing the Relay-UAV stopping point, IRS-UAV stopping point, IRS reflection coefficients and the task offloading ratio, the objective of our proposed optimization scheme is to minimize the transmission delay and computing delay while considering the secure transmission performance. To solve this non-convex optimization problem with coupled variables, we propose an intelligent optimization algorithm based on dueling double deep Q networks (D3QN)-deep deterministic policy gradient (DDPG) that can efficiently explore the trajectories and a great number of the IRS reflection elements. Simulation results demonstrate that the intelligent algorithm exhibits good convergence and our proposed scheme can achieve a good balance between system consumption and secrecy rate.