Reinforcement Learning Algorithm for Organizational Decision Making in Air Defense Kill Networks

When facing high-frequency disturbances and dynamic uncertain environments, existing decision models are difficult to adapt to rapidly changing battlefield demands, and real-time optimization of the kill network scheme based on the battlefield situation is necessary. Although there has been some progress in current research, most of them have transformed the decision-making problem of the kill network into a Weapon Target Assignment (WTA) problem, without considering the dynamic intrinsic mechanism of kill network interception or the impact of dynamic disturbances on the battlefield situation and kill network. To this end, this article proposes a reinforcement learning based method for organizing and decision-making in a distributed air defense kill network. A dynamic representation model of the naval fleet is established, considering iterative optimization of the state under disturbance, and solved through reinforcement learning. The final interception scheme that conforms to the combat mechanism of the kill network is obtained, which satisfies the cooperative air defense mechanism of the combat platform and improves the adaptability of the air defense system to disturbances. The experimental results show that compared with the killing network scheme obtained by traditional commander decision-making methods, this method can optimize and adjust the scheme in real time based on disturbances, and the overall interception probability is higher than the result of commander decision-making.