Reinforcement learning-based three-dimensional cooperative guidance law

An innovative three-dimensional cooperative guidance law for striking stationary target, based on proximal policy optimization (PPO), is introduced. This novel guidance law directly correlates engagement state information with the navigation ratio of proportional navigation guidance (PNG). Initially, the cooperative homing guidance problem is transformed into a Markov decision process, and the reward function incorporates considerations for the zero-effort-miss (ZEM) and the consensus error of time-to-go. Subsequently, the cooperative homing guidance problem is transposed into the framework of reinforcement learning (RL). Ultimately, the effectiveness of the cooperative homing guidance solution is validated through numerical simulations, encompassing agent model training and Monte Carlo cases.