Passive Anti-Jamming Decision-Making Based on Deep Reinforcement Learning

Applying reinforcement learning to the selection of anti-jamming measures in multi-processing domains is an extremely attractive solution. However, current research mainly focuses on design of network optimization, with insufficient attention paid to the rational construction of scenarios, making it difficult to achieve engineering implementation. This paper aims to address this issue by adding jamming parameter information to the discrete space of jamming state and expanding it into a continuous state space, which is in line with the influencing factors of anti-jamming measures. On this basis, based on the radar signal processing process, an anti-jamming index with target detection performance is designed, which is suitable for practical application scenarios. Simulation shows that compared to traditional manual anti-jamming strategies, the proposed method greatly improves the accuracy of selecting anti-jamming measures and analyzes the main factors that affect the accuracy.