Research on Radar Anti-Jamming Algorithm Based on Improved Counterfactual Regret Minimization

Aiming at the confrontation problem between radar and intelligent jamming machine in modern electronic warfare, this paper mainly studies the optimization of radar anti-jamming strategy based on game theory, and proposes a Prioritized Experience Replay Deep Dueling Counterfactual Regret Minimization algorithm. Firstly, the game between radar and jammer is modeled as an Extensive-Form Game, using a game tree to search for Nash equilibrium strategies through multiple rounds of interaction; Secondly, deep learning is combined with the Counterfactual Regret Minimization algorithm. A deep neural network is used to replace the traditional table storage, so as to solve the computational problem caused by the huge state space in radar confrontation. Subsequently, dueling network structure is introduced into the value network, which decouples the state value and action advantage. This improvement enhances the sensitivity of radar in making decisions on key frequency points. Finally, a prioritized experience replay mechanism is adopted to replay high-value samples first, accelerating the convergence of the radar strategy. The experimental results show that the proposed algorithm can effectively improve anti-interference performance compared to traditional algorithms.