Model-free reinforcement learning-based radar beam multistage management method

In this paper, considering the multi-stage radar detection against the repeater jamming, the multi-stage beam management for fire control radar (FCR) is investigated. Based on model-free reinforcement learning, a multi-stage dwell time optimization method is proposed to solve this beam management problem. First, a Markov decision process (MDP) under unknown environment model is established to describe the multi-stage confrontation process. Meanwhile, the search to lock-on time of the FCR is used as the evaluation of detection performance. Then, in response to the unknown environment model in the dwell time optimization, the model-free reinforcement learning framework is formulated. On this basis, a Q-learning based multi-stage dwell time optimization method is proposed. Finally, numerical results demonstrate the effectiveness and robustness of the proposed method in managing the multi-stage radar beam.