关键词: 高超声速飞行器, 再入协同制导, 纵程解析解, 深度强化学习, 深度Q-学习网络

Abstract: A cooperative guidance law based on deep Q-learning network (DQN) algorithm for lateral maneuver of hypersonic vehicles is proposed to meet the requirement of cooperative saturation attack. The longitudinal and lateral guidance laws are deigned. The longitudinal lift-to-drag ratio is calculated to obtain the modulus of bank angle based on the high-precision longitudinal analytical solution. The lateral bank reversal logic is abstracted as a Markov decision process (MDP), and the reinforcement learning method can be used. A lateral intelligent maneuver decision-making device based on DQN algorithm is designed. The intelligent agent can be generated by offline training according to the mission requirements, and is called online to generate the sign of bank angle. Simulated results show that the proposed guidance law can autonomously generate bank angle reversals on-line, strictly meets the requirements of flight time constraints and energy management, and has high guidance accuracy and good mission adaptability. Compared with the coordinative reentry guidance law based on three-dimensional analytical solutions, the proposed guidance law is used to fully achieve the lateral maneuverability of vehicle and make it have higher penetration potential.

Key words: hypersonicvehicle, cooperativereentryguidance, downrangeanalyticalsolution, deepreinforcementlearning, deepQ-learningnetwork