Modeling the decision-making trigger for autonomous ship collision avoidance via imitating human social cognitive process

Artikel i vetenskaplig tidskrift, 2026

Maritime Autonomous Surface Ships (MASS) require collision avoidance decision triggers that align with human cognitive patterns to ensure safe coexistence with manned vessels. Current threshold-based triggers suffer from instability caused by the uncertainty of ship relative motion and instantaneous parameter fluctuations, and the simple comparison manner lacks the cognitive characteristics of human decision-making. This paper proposes VDDM-CADT (Variable Drift Diffusion Model for Collision Avoidance Decision Trigger), which transforms instantaneous collision risk parameters (DDV, TDV, distance) into accumulated evidence over time. Drawing from cognitive neuroscience, the model employs a drift-diffusion process where evidence accumulates until reaching variable decision boundaries shaped by time pressure. Analysis of three simulation cases demonstrates that VDDM-CADT produces stable triggers under parameter fluctuations and ship motion, with decision timings showing high consistency with experienced ship officers. Unlike conventional threshold methods, this evidence-based approach exhibits the social cognitive characteristic inherent to human judgment, potentially enabling MASS to demonstrate socially interpretable behavior patterns during encounters with conventional vessels.