Terminal guidance law of missile based on Soft Actor-critic with impact angle constraint

Considering the decrease in interception success rate caused by the dynamic model of maneuvering targets in aerodynamic environments, while also ensuring the satisfaction of the desired impact angle constraint at the moment of impact, this paper proposes a guidance method based on Soft Actor-critic (SAC) algorithms. The impact angle constraint problem between the missile and the target is transformed into a line-of-sight angle constraint problem. A reward function is designed based on the relative motion status of the missile and target during the guidance process and the desired line-of-sight angle error to guide the interception missile to hit the target while satisfying the impact angle constraint. Finally, a SAC bias proportional navigation law (SACBPN) is obtained by training the reinforcement learning agent. Numerical simulations are conducted under both steady motion and maneuvering conditions of the target. Monte Carlo simulation results show that the proposed deep reinforcement learning guidance law SACBPN is significantly superior to the bias proportional navigation law (BPN), effectively reducing the miss distance and impact angle error at the moment of impact. This indicates that the proposed guidance method is more suitable for intercepting moving targets in aerodynamic environments.