Towards Decision Support in Vessel Guidance Using Multi-Agent Modelling

This paper presents a decision support system for marine vehicle collision avoidance that utilizes agent-based modelling. It generates waypoints, consecutive strategies of heading changes, and if necessary, speed changes to avoid risky collision situations in multi-vessel encounters. The global collision risk metric is defined as a weighted sum of cost functions, which are computed for each target vessel based on factors such as distance and relative velocity. An evolutionary game theory algorithm based on the replicator dynamics concept is applied to determine the best strategy using competitive agents. The proposed method aims to optimize vessel trajectories considering the risk of collision and deviated path length. The feasibility of the approach is demonstrated using simulation in the NetLogo modelling tool and provides insights into how to define an appropriate model for a scalable agent-based application for vessel guidance algorithm verification.