Data-driven multi-agent system for maritime traffic safety management

Abstract

In this study, we propose a data-driven multi-agent system (D-MAS), in which dynamic interactions among different maritime agents, representing shore station-vessel, vessel-vessel, and virtual agent-physical agent, are modeled, to enhance maritime traffic safety management. The multi-agent framework is proposed through integrating learning, forecasting and planning technologies. Our prototype multi-agent system provides solutions to accommodate maritime data analysis and simulation for maritime traffic dynamics and safety evaluation and management. The prototype system developed on top of a set of universal APIs enables adaption to different algorithms for simulation and modeling. Based on multi-agent characterization, we have presented agent-based situational awareness, traffic forecasting, intelligent route planning that are of high importance for maritime traffic safety. The framework proposed can be further developed and applied for applications in maritime safety management.