Radar anti-jamming method based on multi-agent twin delayed deep deterministic policy gradient

With the rapid advancement of intelligent jamming techniques in modern electronic warfare, traditional anti-jamming methods relying on manually designed templates and static game-theoretic models are becoming increasingly inadequate for meeting real-time decision-making demands in dynamic electromagnetic environments. To address these limitations, this paper presents a novel cognitive radar anti-jamming decision-making framework based on the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) algorithm. By formulating the radar-jammer interaction as a Partially Observable Markov Decision Process (POMDP) and adopting a centralized training with decentralized execution (CTDE) paradigm, the proposed framework effectively tackles the challenges of environmental nonstationarity and partial observability inherent in cognitive radar-intelligent jammer adversarial scenarios. Comprehensive theoretical analysis and experimental results demonstrate that our multi-agent system outperforms existing methods in both decision-making efficiency and anti-jamming effectiveness.