An intelligent decision-making method for anti-jamming communication based on deep reinforcement learning

一种基于深度强化学习的通信抗干扰智能决策方法

Received: 27 August 2020

Abstract

In order to solve the problem of intelligent anti-jamming decision-making in battlefield communication, this paper designs an intelligent decision-making method for communication anti-jamming based on deep reinforcement learning. Introducing experience replay and dynamic epsilon mechanism based on PHC under the framework of DQN algorithm, a dynamic epsilon-DQN intelligent decision-making method is proposed. The algorithm can better select the value of epsilon according to the state of the decision network and improve the convergence speed and decision success rate. During the decision-making process, the jamming signals of all communication frequencies are detected, and the results are input into the decision-making algorithm as jamming discriminant information, so that we can effectively avoid being jammed under the condition of no prior jamming information. The experimental results show that the proposed method adapts to various communication models, has a fast decision-making speed, and the average success rate of the convergent algorithm can reach more than 95%, which has a great advantage over the existing decision-making methods.

摘要

为解决战场通信智能抗干扰决策问题，设计了一种基于深度强化学习的通信抗干扰决策方法。该方法在DQN算法架构下引入经验回放和基于爬山策略（PHC）的动态ε机制，提出动态ε-DQN智能决策算法，该算法能够根据决策网络状态更优地选择ε值，提高收敛速度和决策成功率。在决策过程中，对所有通信频率是否存在干扰信号进行检测，将结果作为干扰判别信息输入决策算法，使算法可在无先验干扰信息条件下智能决策通信频率，在尽量保证通信不中断的前提下，有效躲避干扰。实验结果表明，所提方法适应多种通信模型，决策速度较快，算法收敛后的平均成功率可达95%以上，较已有方法具有较大优势。