Research on Radar Countermeasure Knowledge Augmented Deep Adaptive Jamming Policy Decision

The absence of representative training data on actual radar countermeasure environment, leads to the low learning efficiency, policy gradient divergence, and poor generalization ability of jamming decision agent. To solve this problem, we propose a framework of deep adaptive jamming policy agent augmented with radar countermeasure knowledge. It can be formulated as optimal jamming decision path searching, by constructing a hierarchical jamming knowledge tree, and utilizing Monte Carlo Tree Search applied in deep reinforcement learning. Given the credit to radar countermeasure knowledge augmentation and jamming techniques hierarchical optimization, the simulation results demonstrate that compared with traditional q-learning and hierarchical reinforcement learning, the complexity of policy searching is significantly reduced, meanwhile the learning rate of adaptive jamming policy is highly acelerated.