Maritime Autonomous Surface Ships (MASS) are deemed as the future of maritime transport. Although showing attractiveness in terms of the solutions to emerging challenges such as carbon emission and insufficient labor caused by black swan events such as COVID-19, the applications of MASS have revealed problems in practice, among which MASS navigation safety presents a prioritized concern. To ensure safety, rational route planning for MASS is evident as the most critical step to avoiding any relevant collision accidents. This paper aims to develop a holistic framework for the unsupervised route planning of MASS using machine learning methods based on Automatic Identification System (AIS) data, including the coherent steps of new feature measurement, pattern extraction, and route planning algorithms. Historical AIS data from manned ships are trained to extract and generate movement patterns. The route planning for MASS is derived from the movement patterns according to a dynamic optimization method and a feature extraction algorithm. Numerical experiments are constructed on real AIS data to demonstrate the effectiveness of the proposed method in solving the route planning for different types of MASS.

海上自主水面船舶 (MASS) 被视为海上运输的未来。虽然在解决诸如COVID-19等黑天鹅事件导致的碳排放和劳动力不足等新兴挑战方面表现出吸引力，但MASS的应用在实践中暴露出问题，其中MASS航行安全是一个优先考虑的问题。为了确保安全，MASS 的合理路线规划显然是避免任何相关碰撞事故的最关键步骤。本文旨在使用基于自动识别系统 (AIS) 数据的机器学习方法为 MASS 的无监督路线规划开发一个整体框架，包括新特征测量、模式提取和路线规划算法的连贯步骤。来自载人船只的历史 AIS 数据经过训练以提取和生成运动模式。MASS 的路线规划是根据动态优化方法和特征提取算法从运动模式中导出的。在真实的 AIS 数据上构建了数值实验，以证明所提出的方法在解决不同类型 MASS 的路径规划方面的有效性。