Missile guidance with assisted deep reinforcement learning for head-on interception of maneuvering target

Abstract In missile guidance, pursuit performance is seriously degraded due to the uncertainty and randomness in target maneuverability, detection delay, environmental noise. many methods, accurately estimating acceleration of or time-to-go needed intercept maneuvering target, which hard an environment with uncertainty. this paper, we propose assisted deep reinforcement learning (ARL) algorithm optimize neural network-based guidance controller for head-on interception. Based on relative velocity, distance, angle, ARL can control achieve large terminal angle. To reduce influence uncertainty, predicts target’s as auxiliary supervised task. The task improves ability agent extract information from observations. exploit agent’s good trajectories, presents Gaussian self-imitation make mean action distribution approach actions. Compared vanilla learning, exploration continuous control. Simulation results validate that outperforms traditional methods proximal policy optimization higher hit rate larger angle simulation noise, maneuverable target.