DRL-based Resource Allocation for Aerial IoT Systems With No-Fly-Zones

In this work, we investigate the fairness and security of uncrewed aerial vehicle (UAV) communication systems with a flying eavesdropper in a scenario with multiple no-fly zones (NFZs). To improve the security of communication between the aerial base station and the terrestrial users, a UAV is employed as a jammer, transmitting artificial noise to degrade the eavesdropping channel's quality. The three-dimensional flight trajectory and transmit power of UAVs and user scheduling are jointly optimized to maximize the cumulative fair secrecy rate (CFSR) of the considered system. Considering the complexity and non-convexity of the formulated problem, an algorithm based on heterogeneous multi-agent deep reinforcement learning is proposed. This algorithm is meticulously designed to adhere to constraints related to flight space, power, safe distance, and NFZs while maximizing CFSR. Simulation results demonstrate that the proposed algorithm effectively avoids NFZs and achieves tradeoff between fairness and security compared to benchmark schemes, highlighting its potential for enhancing UAV communication systems' performance.