Multi-Agent Reinforcement Learning for Multiple Rogue Drone Interception

Unmanned aerial vehicles (UAVs) are increasingly being utilized for a wide variety of applications. However, malicious or illegal UAV (drone) activity poses great challenges for public safety. To address such challenges, this work proposes a framework based on reinforcement learning (RL) in which multiple UAVs cooperatively jam multiple rogue drones in flight in order to safely disable their operation. The main objective is to select mobility and power level control actions for each UAV to best jam the rogue drones, while also accounting for the interference power received by surrounding communication systems. Simulation experiments are conducted to evaluate the performance of the proposed approach, demonstrating its effectiveness and advantages as compared to a centralized solution.