Identification of key risk ships for MASS collision avoidance in mixed traffic through axiomatic fuzzy set-based semantic clustering

In mixed traffic involving conventional ships and Maritime Autonomous Surface Ships (MASS), autonomous collision avoidance requires attention not only to ships presenting high collision risks, but also to those that may conflict with the own ship and whose unstable motion may significantly affect trajectory prediction and collision-avoidance decisions. Therefore, their risk characteristics and identification rationale need to be represented using fine-grained risk semantics, thereby providing an interpretable basis for autonomous collision avoidance. To this end, this study first develops a 4D-OPTICS based method to screen candidate ships in dense encounter scenarios. By excluding spatially proximate ships with weak relative-motion risk, the method narrows the candidate set for subsequent identification. An Axiomatic Fuzzy Set (AFS)-based semantic extraction model is then developed to transform relative-motion risk indicators into fine-grained semantic representations of spatial proximity, temporal urgency, and motion-trend uncertainty. Subsequently, a hierarchical identification model is constructed based on AFS fuzzy clustering, incorporating ship-level semantic verification and the give-way and stand-on responsibilities prescribed by the COLREGs to distinguish priority attention ships from dynamic monitoring ships. Numerical simulations and real-world AIS data collected in Yingkou Port demonstrate the effectiveness of the proposed framework. The framework establishes an interpretable basis, grounded in risk semantics and rule-based criteria, for identifying and prioritizing ships in multi-ship encounter scenarios involving MASS.