Marine MAS: A multi-agent framework to support autonomous maritime system design, modeling, and assessment

Abstract

Autonomous shipping brings a great opportunity to 
improve the maritime traffic safety, reliability, and to 
reduce the cost, whose prototypes are expected to be 
industrialized in coming years. Current studies and 
projects focus on proof development of autonomous 
vessels, evaluation of autonomous vessel safety, and 
formulation of guidance for autonomous shipping paradigm. 
In system level, autonomous shipping is anticipated to 
realize advanced intelligence and autonomy in terms of 
planning, situation awareness and shipping safety, 
decision-making, etc. In this study, being aligned with the 
aforementioned aspects, a multi-agent system (MAS) 
framework, called “MarineMAS,” is proposed to aid the 
autonomous shipping modelling and evaluation. We 
emphasize the design perspective that especially focuses 
on critical knowledge transfer to make transition towards 
autonomous shipping, the core components to achieve 
system intelligence, and the potential technological set and 
experience gained to support efficient MAS modelling. A 
prototype system has been imple-mented from scratch to 
demonstrate the key features of the proposed MarineMAS 
frame-work. Through characterizing the multi-agent 
definitions and incorporating the corresponding evaluation 
logics, MarineMAS can be further developed and applied 
for various applications in autonomous shipping 
evaluation and management scenarios.