Study On Meta-reinforcement Learning Guidance Law For Intercepting Maneuvering Targets

With the continuous improvement of the performance of high-tech attack guidance weapons,higher requirements are put forward for defensive interception guidance system.This paper studies the guidance law of intercepting maneuvering target based on meta-reinforcement learning method to improve the intercepting performance of missile defense system,which has important theoretical significance and application value.The main work of this paper is as follows:1.Based on Markov Decision Process,this paper describes the terminal guidance of missile interception,designs the forms of state,action and reward function,and establishes the reinforcement learning interception guidance model to adapt to the complex maneuvering of target.2.The zero-effort-miss distance was introduced as the reward function,and the step-updating training method was designed according to the importance sampling principle of PPO(Proximal Policy Optimization)-Clip.A guidance law of reinforcement learning interception based on PPO-Clip was proposed,and the effectiveness of the algorithm was verified by simulation experiments.3.A dual-loop optimization strategy combining MAML(Model-Agnostic MetaLearning)and PPO-Clip algorithm was designed,and a meta-reinforcement learning interception guidance law based on MAML+PPO-Clip algorithm was proposed,which improved the generalization ability of the algorithm.The effectiveness of the method was verified by simulation.