基于强化学习的雷达抗复合干扰  

Anti-composite Jamming Technology Based on Reinforcement Learning

作　　者：许佰涛 刘冬利[2] 侯建强 李祎帆 XU Baitao;LIU Dongli;HOU Jianqiang;LI Yifan(Midshipmen Group Five,Dalian Naval Academy,Dalian 116018;Department of Information Operation,Dalian Naval Academy,Dalian 116018)

机构地区：[1]海军大连舰艇学院学员五大队,大连116018 [2]海军大连舰艇学院信息系统系,大连116018

出　　处：《舰船电子工程》2022年第10期83-86,92,共5页Ship Electronic Engineering

基　　金：军内科研项目(编号:DJYJNKY2019-018)资助。

摘　　要：针对传统雷达抗复合干扰决策难、时间长、难以预测的问题,提出了一种基于先验知识强化学习的雷达抗复合干扰的方法。该方法将雷达单一干扰效益,拓展到复合干扰效益,进一步得到复合干扰转化概率矩阵,通过强化学习得到动作价值矩阵,确定最优抗复合干扰策略。实验仿真发现:与传统的Q-learning相比,基于先验条件的雷达智能抗干扰决策更加贴近战场情况,可以较为准确地预测下接下来干扰状态,为指挥员提供决策依据。Aiming at the problem that the traditional radar anti-jamming decision-making is difficult,long and unpredictable,this paper proposes a radar anti-composite interference method based on prior knowledge and reinforcement learning. This method extends the single interference benefit of radar to the compound interference benefit,further obtains the probability matrix of compound interference conversion,obtains the action value matrix through reinforcement learning,and determines the optimal anti-composite interference strategy. Experimental simulation found that compared with traditional Q-learning,radar intelligent anti-jamming decision based on prior conditions is closer to the battlefield situation,which can predict the next interference state more accurately,provide a decision-making basis for commanders.

关 键 词：强化学习 抗复合干扰 复合干扰转移概率 动作价值矩阵 

分 类 号：TN974[电子电信—信号与信息处理]