Abstract: In the modern electronic warfare environment, the jamming styles faced by radars have become increasingly complex and diverse, showing a high degree of dynamics, and traditional anti-jamming decisions relying on manual experience are difficult to deal with effectively. Aiming at the requirement of radar dynamic adaptive anti-jamming decision-making, this paper proposes a radar intelligent anti-jamming dynamic decision-making method based on reinforcement learning Q-Learning. With the goal of “reducing the jamming threat level and quickly converging to low-threat state”, this anti-jamming decision-making method first quantifies the radar jamming threat level and constructs the jamming state transition rules based on the anti-jamming benefit matrix, then designs a reward function with the reduction of jamming threat level as the core orientation, and finally realizes the autonomous learning and dynamic optimization of anti-jamming strategies under the Q-Learning framework. Simulation experiments show that this method can achieve stable convergence in the scenario of 10 jamming states and 9 anti-jamming measures. The overall success rate under 10 initial states reaches more than 92%, and the average number of transfer steps from high-threat states to the target state is controlled within 8 steps. It is significantly superior to fixed strategies and random strategies in transfer efficiency and decision-making success rate, which verifies the effectiveness and superiority of the method and provides an effective technical path for the intellectualization of radar anti-jamming decision-making.