Q-learning强化学习协同拦截制导律

Cooperative Interception Guidance Law Based on Reinforcement Learning of Q-learning

为实现多枚导弹协同拦截机动目标,提升拦截效能,提出了一种Q-learning强化学习协同拦截制导律。首先,基于逃逸域覆盖理论,建立了非线性多弹协同拦截模型。其次,以视线角速率为状态,依据脱靶量构造奖励函数,通过离线训练生成强化学习智能体,并结合传统比例制导控制方法,构建基于强化学习的变导引系数制导律,实时生成实现协同拦截的制导指令。最终,通过数值仿真验证了所提算法的有效性和优越性。

To achieve the cooperative interception of multiple missiles against a maneuvering target and improve the interception effectiveness, a cooperative interception guidance law is proposed through Q-learning technology. Firstly, based on escape domain covering theory, a nonlinear cooperative interception model is established. Then, a reward function is constructed by using miss distance with taking line-of-sight rate as the state, and a reinforcement learning agent is generated by offline training. At the same time, a variable coefficient guidance law based on reinforcement learning algorithm is designed by combining proportional navigation guidance law to generate guidance commands in real time. Finally, the effectiveness and superiority of the proposed algorithm are verified based on numerical simulation.