Reinforcement Learning Based Radar Anti-Jamming Strategy Design against a Non-Stationary Jammer

Abstract

In modern electronic warfare, the jamming strategy is more complex than before and the jammer is capable of changing its strategy. Traditional anti-jamming strategy learning methods apply reinforcement learning (RL) and cannot handle the non-stationary jamming strategy. To address that issue, we propose a method that combines RL and supervised learning (SL) to design an anti-jamming strategy for frequency-agile (FA) radar. Firstly, we consider three jamming strategies, which include changes in the duration and type of jamming. And the non-stationary jammer is described by the uncertainty of the jamming strategy. Secondly, by analyzing the non-stationary characteristics of the jamming strategy, an anti-jamming algorithm of joint RL and SL is proposed, and we give the specific process of the algorithm. Finally, the simulation experiments demonstrate that the proposed method can effectively deal with the dynamically changing mainlobe interference, and the anti-jamming performance of the proposed method is robust.