Intelligent decision and planning for unmanned surface vehicle: A review of machine learning techniques

With the increasing demand for unmanned surface vehicles (USVs) in fields such as marine environmental monitoring, resource exploration, and emergency rescue, the development of intelligent decision and planning technologies has become critical. However, the complexity and dynamic nature of marine environments pose significant challenges to traditional methods in practical applications. In recent years, the rapid advancement of machine learning (ML) has offered novel solutions for the intelligent decision and planning of USVs. This paper systematically reviews the research progress in USV decision and planning based on ML. First, it reviews the classification of USV autonomy levels and the historical development of ML in unmanned marine systems. Then, the paper proposes and elaborates on the "ML-MDP" framework (a ML-based Mission planning, Dynamic decision, and Path planning framework) for USVs, analyzing the latest research outcomes in these areas and explores the suitability of various ML algorithms in addressing these challenges. Finally, the paper analyzes the challenges faced by ML in USV applications and its future development directions. This review aims to provide a valuable reference for researchers in related fields, highlighting the potential of ML in marine unmanned systems and promoting advancements in USV intelligence.