Path planning for UUV using explainable AI and deep reinforcement learning techniques

Currently, an increasing number of deep learning algorithms are being applied to UUV path planning research, demonstrating powerful performance and high efficiency. However, the black-box nature poses significant challenges to the interpretability of AI-driven decisions, and this lack of interpretability may lead to various impacts such as potential safety risks. This paper conducts interpretability research on UUV path planning using deep reinforcement learning by employing two explainable AI methods: feature analysis and policy distillation.