Partially centralized coordination method for collision avoidance of multiple ships via mediation

Journal article, 2026

Ship collision avoidance is a core issue in autonomous navigation. However, most research predominantly focus on two-ship or one-to-many scenarios from the perspective of a single own ship. Consequently, there remains a critical gap in resolving decision conflicts and uncoordinated maneuvers among multiple ships in complex and communication-restricted environments. This paper proposes a partially centralized coordination method via mediation based on the Distributed Constraint Optimization Problem (DCOP) framework. The approach follows a three-stage process. First, encounters are characterized as dynamic interaction topologies by calculating relevant motion parameters, considering COLREGs, and quantifying collision risks. Second, multi-ship coordination is formulated as a specific DCOP that mathematically integrates safety and economy into a specialized cost-constraint structure. Finally, the enhanced Optimal Asynchronous Partial Overlay (OptAPO) algorithm with a risk-driven mediation trigger is employed to adaptively balance distributed autonomy and localized centralization. Simulations conducted in symmetrical 6-ship and 8-ship encounter scenarios demonstrate that compared to reactive and incomplete algorithms, this method effectively overcomes decision conflicts and ensures safety. Furthermore, it achieves comparable global optimality to fully distributed algorithms while significantly reducing communication overhead. Ultimately, this approach provides an efficient, reliable, and deployable solution for multi-ship cooperative collision avoidance under constrained communication in complex encounter situations.