High-Speed Aircraft Swarm Collaboration Method

Facing the future all-area fast-time response combat demand, the high-speed aircraft swarm system is bound to develop in the direction of intelligence. Considering the uninterpretable problem of reinforcement learning in practical applications, the safe reinforcement learning theory is introduced into the intelligent decision-making problem of aircraft swarm, and the improved soft actor-critic algorithm is proposed based on the Lyapunov stability theory. Firstly, a high-speed vehicle dynamics model is established with normal overload as the control quantity. Secondly, the task is divided into three parts: target striking, formation keeping, and threat zone avoidance, and the vehicle guidance reward function is designed. Finally, by comparing the training results with those of the actor-critic algorithm, it is concluded that the method can enable the agent to maintain a low safety cost throughout the training process.