Deep Reinforcement Learning Assisted Trajectory Design for UAV Communication Systems Against Malicious Jamming

In this paper, we propose a novel trajectory design for the unmanned aerial vehicle (UAV) communication systems, where a single UAV moves along with multiple mobile terres-trial users to provide communication services, in the presence of malicious jamming. In our proposed design, we take the users' mobility and the jammer's adaptive jamming power into consideration and model the UAV's trajectory as a Markov decision process. Then, with the goal of optimizing the overall average throughput in our system, we design a twin-delayed deep deterministic policy gradient-based algorithm to judiciously optimize the trajectory of the UAV in a continuous manner, while simultaneously satisfying practical constraints on both the UAV's and users' mobility. We compare the performance of our proposed design with various baselines and show how it can improve the average throughput performance in the presence of malicious jamming and converge at a faster rate.