基 于 强 化 学 习 的 多 机 协 同 超 视 距 空 战 决 策 算 法

As the modern air combat environment grows increasingly complex and dynamic，the need for rapid and effective decision-making methods has become urgent. This paper proposes a multi-aircraft cooperative beyond-visual-range air combat decision-making algorithm based on long and short-term memory (LSTM) and multi-agent deep deterministic policy gradient (MADDPG) to address the challenge of collaborative confrontation of multiple unmanned aerial vehicles (UAVs). First, a beyond-visual-range air combat environment is established, including the UAV movement model, the radar detection zone model, and the missile attack zone model. Second, the multi-aircraft collaborative beyond-visual-range air combat decision-making algorithm is proposed. This algorithm includes a centralized-training distributed-execution framework and a state space of the collaborative air combat system to handle synchronous decision-making across multiple UAVs，a learning rate decay mechanism to enhance network convergence speed and stability, an improved network based on LSTM to strengthen tactical feature extraction, and a decay-factor-based reward function to improve cooperative confrontation performance. Experimental results demonstrate that the proposed algorithm equips UAVs with effective collaborative attacking and defensive capabilities, while exhibiting strong stability and convergence.