An UAV Interception Method based on Guidance Law inspired Reinforcement Learning

This paper presents a reinforcement learning (RL)based unmanned aerial vehicle (UAV) interception system, addressing the critical challenge of intercepting adversarial UAVs in dynamic and uncertain environments. UAVS have important roles in both the military and civilian sectors, including reconnaissance, surveillance and disaster response The integration of artificial intelligence, particularly RL, has enhanced UAV capabilities in autonomous navigation, target recognition, and path planning. This work formulates UAV interception as a guidance-based decision-making problem and employs the Proximal Policy Optimization (PPO) algorithm to train agents to generate optimal velocities for interception. A novel reward function inspired by traditional guidance laws is proposed to accelerate training convergence. In order to validate the approach proposed in this paper, an RL simulation environment, a velocity increment implementation for realistic UAV dynamics, and an effective reward function are designed using Unreal Engine 4 and AirSim. Simulation experiments demonstrate the feasibility and effectiveness of the proposed RL-based interception method.