Curriculum learning-based missile guidance law for intercepting maneuvering targets with high-speed

In this paper, a missile guidance strategy is designed by using curriculum-based reinforcement learning for intercepting high-speed maneuvering targets. To enhance the exploration capabilities of agents within the state space, an exploration ability index is introduced. The proposed index, integrated with the entropy of the Soft Actor-Critic algorithm, dynamically adjusts the training process to augment exploration. In the proposed curriculum-based reinforcement learning, multiple sub-tasks are designed to progressively train the agent from simple to complex scenarios. This incremental learning framework enables the agent to efficiently master the task, expediting the overall training process. Through comprehensive simulations, we validate the efficacy of the proposed guidance law. Comparing the results obtained in other reinforcement learning methods, our findings highlight the advantages of curriculum learning in enhancing the efficiency and effectiveness of missile guidance systems.