Research on Guidance Law with Constraint Attack Angle Based on Reinforcement Learning

Abstract

Aiming at the problem of designing guidance law to attack surface target at a specific angle, a reinforcement learning guidance law model is constructed by using depth deterministic strategy gradient algorithm, and the model states, reward rules and guidance environment are designed. By setting different initial conditions and attack angles, the guidance law model is trained and the guidance law is stable. The reinforcement learning guidance law can make the missile hit the fixed target with constraint attack angle and it can hit the surface target with low speed within small attack angle error. The simulation results show that compared with the optimal guidance law with constraint attack angle, the attack angle convergence rate of the enhanced learning guidance law with constraint attack angle is faster, the acceleration change is smoother, the acceleration value at the end of guidance is smaller, the ability to adapt to battlefield conditions is much better.

Keywords

• |guidance law|reinforcement learning|ddpg|constraint attack angle|markova|intelligence algorithm