RL Based UAV Trajectory Optimization to Avoid Jammers and Optimize Wireless Connectivity

Drones have found extensive use in military applications owing to their inconspicuousness and easy deployment. Major challenges in the UAV space today include: (a) Maintaining seamless connectivity with the ground station to maintain command and control (C2) link and ensure continuous streaming of high quality data and (b) avoiding enroute RF jammers that can saturate the RF receiver and inhibit communication. Additionally, GNSS jammers can render drones completely nonfunctional by not allowing the drone to communicate with GPS satellites to ascertain their location. This work addresses these two challenges by developing a reinforcement learning (RL) algorithm that is able to optimize the wireless connectivity while avoiding jammers along the drone route. Notably, the algorithm ensures minimal deviation from the planned path in the presence of jammers. The training process for the drone can be completely accomplished pre-flight. The drone is only required to carry a lookup table (Q-table) during flight thereby reducing the computing requirements onboard the drone.