Human-Governed Agentic AI for Maritime Crew Change Management: A Multi-Agent Decision Architecture for Coordinating Travel, Documentation, Port Access, Relief Risk, and Welfare

Maritime crew change management remains fragmented across crewing departments, travel providers, port agents, vessel operators, documentation teams, and compliance functions. Although digital platforms and predictive models can improve visibility, many crew change failures still arise from poor coordination among interdependent decisions rather than from a lack of information alone. This study develops a human-governed agentic AI architecture for maritime crew change management. The proposed architecture uses specialized decision agents for crew availability, documentation readiness, travel planning, port access, relief-risk monitoring, welfare escalation, carbon-aware itinerary selection, and emergency replacement. These agents exchange structured decision states and produce coordinated recommendations under human supervisory control. A governance layer is introduced to prevent uncontrolled automation by enforcing approval gates, confidence thresholds, audit trails, escalation rules, and constraint validation. The architecture is evaluated through a ۳۰-day computational scenario involving ۸ vessels, ۱۸ relief requirements, ۳۱ candidate replacements, ۹ joining ports, and ۶۴ travel-documentation-port coordination cases. The agentic AI architecture is compared with a conventional sequential workflow and a predictive-alert workflow. Results show that the agentic workflow reduces unresolved crew change conflicts by ۴۸.۶%, lowers average decision response time from ۱۴.۲ hours to ۵.۸ hours, reduces high-risk relief cases from ۱۱ to ۴, and improves on-time relief completion from ۷۷.۸% to ۹۴.۴%. The results also show that human approval gates prevent ۶ potentially unsafe autonomous recommendations, confirming the importance of governed autonomy in crew change decision-making. The study contributes to maritime workforce management by shifting crew change research from isolated prediction and optimization models toward coordinated, multi-agent decision orchestration. It also provides a practical foundation for deploying agentic AI in safety-sensitive maritime operations without removing human accountability.

کلیدواژه ها:

Agentic AI ، Maritime crew change ، Multi-agent decision support ، Crew rotation ، Human-in-the-loop governance ، Seafarer relief ، Maritime workforce management ، Crew logistics ، Decision orchestration ، Responsible AI