Intelligent Maneuver Strategy for a Hypersonic Pursuit-Evasion Game Based on Deep Reinforcement Learning

In order to improve the problem of overly relying on situational information, high computational power requirements, and weak adaptability traditional maneuver methods used by hypersonic vehicles (HV), an intelligent strategy combining deep reinforcement learning (DRL) neural network (DNN) is proposed solve pursuit–evasion (PE) game under tough head-on situations. The twin delayed deterministic (TD3) gradient algorithm utilized explore potential instructions, DNN fit broaden application scenarios, generated with initial situation both pursuit evasion sides as input overload HV output. addition, experience pool classification training convergence rate TD3 algorithm. A set reward functions designed achieve adaptive adjustment miss distance energy consumption different simulation results verify feasibility effectiveness above in dealing PE difficult situations, improvement strategies are validated well.