Incorporation of System Identification Based Ship Manoeuvring Model into Real-time Collision Risk Analysis

Collision risk assessment is a critical component of navigational safety, particularly in the context of emerging autonomous and semi-autonomous vessels. Although numerous collision risk models have been proposed, many existing approaches rely primarily on kinematic information derived from AIS data and commonly assume constant speed and straight-line motion for target ships (TS). Such assumptions limit the ability to capture realistic manoeuvring behaviour and introduce significant uncertainty in risk prediction. This study proposes an integrated framework that incorporates a data-driven ship manoeuvring model into real-time collision risk analysis. A system identification-based manoeuvring model, developed using a Dynamic Bayesian Network (DBN), is employed to predict future motion states, including position, speed, and heading. These predicted states are subsequently integrated into a DBN-based collision risk assessment model through CPA-based evaluation, enabling anticipatory risk prediction at one-minute intervals. Simulation studies are conducted using the KCS benchmark vessel to investigate both own ship (OS) and TS intention scenarios. The results demonstrate that the OS can quantitatively evaluate the impact of different rudder actions on collision risk and make informed, COLREG-compliant manoeuvring decisions in advance. Furthermore, the TS intention analysis highlights how considering simple control-related assumptions, such as rudder actions, can substantially influence future encounter geometry and collision risk evolution, even when such parameters are not directly available from AIS data. Overall, the proposed framework provides an intention-aware and uncertainty-informed collision risk assessment capability, offering valuable insights for the design of future autonomous navigation and decision-support systems.

Keywords: Collision risk assessment, Ship manoeuvring modelling, Dynamic Bayesian Networks, Intention analysis, Motion uncertainty, Maritime transportation safety.