基于双层强化学习的多功能雷达认知干扰决策方法

Multi-function radar cognitive jamming decision-making method based on two-layer reinforcement learning

针对认知对抗中干扰方难以获取雷达方正确先验知识、面对复杂模型求解最优干扰策略困难的问题,提出了基于双层强化学习的多功能雷达认知干扰决策算法,第1层强化学习验证先验知识是否正确,并决定是否更新先验知识;第2层强化学习基于更新的先验知识进行强化学习,生成Q矩阵指导干扰方进行干扰决策。为了提高双层强化学习算法的干扰决策效率以及干扰决策正确率,对Q-learning算法的动作选择策略和收益函数设置方法进行了改进。仿真实验表明,在错误先验知识的情况下,该算法可以解得正确的最优干扰策略。相比于单层强化学习,该算法提高了干扰方适应复杂电磁环境的能力,使得强化学习在多功能雷达认知干扰决策中更具应用价值。

In cognitive confrontation,it is difficult for the jammer to obtain correct prior knowledge of the radar and solve the optimal jamming strategy orienting to complex models.Therefore,a multi-function radar cognitive jamming decision-making algorithm based on two-layer reinforcement learning is proposed to solve above problem.The first�layer reinforcement learning verifies whether the prior knowledge is correct and decides whether to update the prior knowledge.The second layer of reinforcement learning performs reinforcement learning based on updated prior knowledge,and generates a Q matrix to guide the jammer to make jamming decisions.The strategy of action selection and benefit function setting method of the Q-learning algorithm are improved,so as to improve the efficiency of jamming decision-making and the accuracy of jamming decision-making.Simulations show that the algorithm can solve the correct optimal jamming strategy under the condition of false prior knowledge.Compared with single-layer reinforcement learning,the algorithm improves the ability of the jammer to adapt to complex electromagnetic environment,which makes reinforcement learning more valuable in multi-function radar cognitive jamming decision�making.