Track-Optimization of Relay UAV based on Deep Reinforcement Learning

Future wars will be more technologically advanced and information-driven. Unmanned Aerial Vehicle (UAV) relay communications will play a crucial role in the battlefield. The paper aims to address the deployment of RUAVs on the battlefield by designing trajectories that consider air users, ground users, and potential threats. The goal is to ensure that RUAVs maximize the uplink data transfer rate while ensuring safety. Given the complexity of the battlefield environment, we have designed a neural network structure with predictive capabilities and proposed a DDQN-based RUAV trajectory optimization algorithm. Our experiments have demonstrated that our algorithm learns environmental information more quickly, achieves greater returns, and exhibits faster and more stable convergence performance in battlefield environments than traditional methods.