Anti-Interception Guidance for Hypersonic Glide Vehicle: A Deep Reinforcement Learning Approach

Anti-interception guidance can enhance a hypersonic glide vehicle (HGV) compard to multiple interceptors. In general, anti-interception for aircraft be divided into procedural guidance, fly-around and active evading guidance. However, these methods cannot applied an HGV’s unknown real-time process due limited intelligence information or on-board computing abilities. this paper, approach based on deep reinforcement learning (DRL) is proposed. First, the penetration conceptualized as generalized three-body adversarial optimal (GTAO) problem. The problem then modelled Markov decision (MDP), DRL scheme consisting of actor-critic architecture designed solve this. Reusing same sample batch during training results in fewer serious estimation errors critic network (CN), which provides better gradients immature actor (AN). We propose new mechanismcalled repetitive (RBT). addition, data test confirm that RBT improve traditional DDPG-based-methodes.