Intelligent Penetration Guidance for Hypersonic Missiles via Reinforcement Learning and Optimal Control

To enhance the penetration capability and strike accuracy of missiles in scenarios involving multiple interceptors, this paper proposes a penetration guidance strategy based on deep reinforcement learning (DRL). First, the BANG-BANG optimal penetration strategy is derived as an expert policy by optimizing the maximum miss distance between the attacking missile and the interceptor missiles as a performance metric. Subsequently, a Markov Decision Process (MDP) model for missile penetration guidance is established, and a multi-objective reward function is designed to integrate penetration success rate, miss distance, and energy consumption. Furthermore, a penetration strategy learning network based on Generative Adversarial Imitation Learning (GAIL) and Proximal Policy Optimization (PPO) is constructed and trained. Simulation results demonstrate that the proposed strategy exhibits high training efficiency and enables superior decision-making in complex adversarial scenarios.