Ship collision avoidance behaviour identification and analysis method

In the field of intelligent ships, enabling ships to navigate with human-like decision-making capabilities has long been a research focus. However, few studies have objectively demonstrated or systematically mined the existence of ship collision avoidance behaviour (CAB) from real-world data. This study proposes a spatial-temporal characterisation framework for ship CAB based on extracting and analysing near-collision scenarios from historical Automatic Identification System (AIS) data. Firstly, collision avoidance manoeuvres are identified using a Change Point Detection (CPD) technique, in which a cost function is formulated to distinguish the manoeuvring behaviour from normal sailing behaviour, and critical encounter scenarios are extracted through encounter situation analysis. Second, contributions of collision avoidance characteristic indicators are determined through Principal Component Analysis (PCA), taking into account the requirements of the International Regulations for Preventing Collisions at Sea (COLREGs). The proposed framework is applied to the Øresund waterway between Denmark and Sweden, where the spatial-temporal distribution of critical encounters and behavioural patterns is examined. Results of the case study reveal the importance of relative bearing and phase in determining the timing of collision avoidance for both give-way and stand-on ships. This study provides a novel perspective for understanding ship navigation behaviour, enhancing situational awareness, and supporting the quantitative characterisation of collision risk.

Keywords: Ship Collision Avoidance Behaviour, Change Point Detection, Principal Component Analysis.