Deep Reinforcement Learning-Based Cooperative Guidance Law with Multi-Constraint

To enhance the destructive capabilities of small air-to-surface missiles, it is necessary to address the challenges related to terminal attack angle constraints and multi-missile coordinated attacks, we propose a cooperative guidance law with multi-constraint based on deep reinforcement learning in this paper. Firstly, an intelligent control term is introduced into the biased proportional navigation law with angle constraints. Secondly, a reward function is designed to reduce terminal attack angle error and flight time error, ensuring that the missile successfully reaches the target within the expected flight time, while simultaneously converging the attack angle to the desired value. Finally, the deep deterministic policy gradient algorithm is employed for the purpose of training and generating intelligent guidance instructions. Simulation results substantiate the efficacy of the devised guidance method in achieving accurate target strikes, both for individual missiles and in scenarios involving multiple missiles. And exhibiting better stability and guidance accuracy compared to the traditional biased proportional navigation law.