A Review on Path Planning for Autonomous Underwater Vehicles: From Models, Classical Methods, and Learning‐Based Intelligence Perspectives

ABSTRACT Autonomous underwater vehicles (AUVs) are pivotal for advancing ocean exploration, environmental monitoring, and deep‐sea resource utilization. However, their operational effectiveness is fundamentally constrained by the harsh and dynamic underwater environment. This review systematically synthesizes the state of the art in AUV path planning, integrating three core dimensions: high‐fidelity underwater environment modeling, evolution from classical to intelligent path planning algorithms, and the emerging role of machine learning, particularly reinforcement learning, in enabling adaptive decision‐making. We dissect recent advances in AUV motion dynamics modeling, terrain and current field reconstruction, and multi‐AUV collaborative task allocation. Furthermore, we critically evaluate traditional graph‐based and sampling‐based planners, alongside bio‐inspired and learning‐based approaches, highlighting their scalability, real‐time performance, and robustness under uncertainty. In addition, this review highlights key challenges, including computational constraints, energy efficiency, and environmental unpredictability, and proposes the future direction of meta learning to enhance autonomy, multimodal awareness fusion, and edge computing support planning. This review establishes a unified framework to guide the next generation of AUV path planning systems to achieve true autonomy and intelligence in complex ocean tasks.