A Multi-Ship Intelligent Collision Avoidance Decision Method Considering Intention Uncertainty

Abstract

A collision avoidance decision-making method that integrates intention recognition with predictive risk perception is proposed to address the uncertainty and dynamic complexity of multi-ship encounters. First, the probabilities of avoidance maneuvers are derived from COLREGs and maritime practice to establish predictive distributions of course alterations. Second, an avoidance-maneuver intention (AMI) confidence coefficient is formulated from evidence fused using Dempster–Shafer (DS) evidence theory to quantify the reliability of the target ship’s maneuver behavior. Third, a predictive risk-perception-based line-of-sight (PRP-LOS) method is developed by introducing a predicted-equivalent virtual ship position into a navigation risk field. Simulation results show that the proposed method accurately infers target-ship AMIs and effectively reduces collision risk while balancing safety and smoothness. These research results provide effective methodological support for improving autonomous decision-making capabilities.