Predicting Seafarer Control Transfer Intentions for Safe Human-Machine Collaboration in Maritime Autonomous Surface Ships

With the rapid development of autonomous ships, the role of seafarers onboard Maritime Autonomous Surface Ships (MASS) is in significant transformation. Among onboard operations in autonomous ships, the control transfer between seafarers and systems serves as a critical process to ensure MASS safety. However, research gaps remain concerning when and how seafarers takeover control from autonomous systems. This study investigates seafarers' control transfer intentions in hazardous navigation scenarios by integrating computer vision techniques with behavioral analysis in real-world sailing experiments. The experimental framework synchronously captures seafarer behavior videos and vessel status data onboard autonomous ship Yujiaotou 001. Computer vision methods are employed to extract key behavioral features, including body posture and hand movements. By integrating behavioral features with vessel trajectory data, this study identifies key indicators of seafarers' intention to shift and develops an intention prediction model for onboard seafarers during control transfer. The findings provide theoretical and empirical foundations for enhancing maritime education and training for MASS, as well as improving safety management in autonomous ship operations.

Keywords: MASS, human-machine collaboration, control transfer, takeover, handover.