一种基于深度强化学习的动态自适应干扰功率分配方法  

A Dynamic Adaptive Jamming Power Allocation Method Based on Deep Reinforcement Learning

作　　者：彭翔 许华[1] 蒋磊[1] 张悦 饶宁 PENG Xiang;XU Hua;JIANG Lei;ZHANG Yue;RAO Ning(Information and Navigation School,Air Force Engineering University,Xi'an,Shaanxi 710077,China)

机构地区：[1]空军工程大学信息与导航学院,陕西西安710077

出　　处：《电子学报》2023年第5期1223-1234,共12页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61906156)。

摘　　要：针对传统干扰功率分配方法在干扰目标策略未知的情况下容易造成资源浪费和干扰效费比低的问题,本文提出一种基于深度强化学习的动态自适应干扰功率分配方法 .在目标通信功率及功率控制策略完全未知的情况下,该方法将空间分布的侦察节点的观测值作为连续状态输入,利用深度强化学习方法进行干扰功率的辅助决策,可通过对目标策略的有效学习实现自适应稳定干扰.为进一步提升算法性能,本文设计了基于时序误差的优先经验回放机制和自适应探索策略.仿真结果表明,所提方法在与传统干扰功率分配方法干扰效果相当的情况下可节约42.5%的功率资源,提升了干扰效费比,且成功率和功率损耗皆优于对比的智能算法.To solve the problem that traditional jamming power allocation methods are prone to waste resources and low jamming effectiveness-cost-ratio when the jamming target strategy is unknown,a dynamic adaptive jamming power al-location method based on deep reinforcement learning is proposed.When the communication power of the target and its power control strategy is completely unknown,the method takes the observation values of spatially distributed reconnais-sance nodes as continuous state input and uses the deep reinforcement learning method to assist the decision-making of jam-ming power.It can achieve the adaptive stable jamming by the effective learning of target strategy.To further improve the performance of the algorithm,a prioritized experience replay mechanism based on temporal-difference error and an adap-tive exploration strategy are designed.The simulation results show the proposed method can save 42.5%of power resources and improve the jamming effectiveness-cost-ratio when the jamming effect is equivalent to that of the traditional jamming power distribution method.The success rate and power cost of the proposed algorithm are better than those of the compara-tive intelligent algorithms.

关 键 词：电子对抗 通信对抗 干扰资源分配 干扰决策 功率分配 深度强化学习 优先经验回放 

分 类 号：TN975[电子电信—信号与信息处理]