From Simulation to Semi-Physical Validation: An Intelligent Jammer-Assisted Radar Anti-Jamming Evolution Method

Adaptive waveform decision-making remains a key challenge in radar anti-jamming, especially when radar systems need to evolve their strategies through dynamic confrontation with intelligent jammers. This paper presents an intelligent jammer-assisted radar anti-jamming evolution method based on the AlphaZero framework. The radar and jammer are modeled as two competing agents, and their waveform-level interaction is formulated as a sequential decision-making game. By combining self-play learning with Monte Carlo Tree Search, the proposed framework guides the generation of high-value interaction samples and improves the training efficiency of the radar agent. As a result, the radar agent can progressively optimize its anti-jamming strategy and enhance its decision-making capability during adversarial interactions. To further evaluate its practical feasibility, a semi-physical hardware-in-the-loop validation platform is developed. Experimental results show that the proposed method accelerates the convergence of the radar agent, improves the utilization efficiency of valuable interaction samples, and bridges the gap between algorithmic simulation and practical radar–jamming system implementation.