Integrating geometric and causation probability approaches into Dynamic Bayesian Networks for real-time collision risk prediction

Maritime transportation is vital for international trade, yet collision accidents continue to pose serious risks to navigational safety and global economic stability. This study develops a novel collision risk prediction model based on Dynamic Bayesian Networks (DBN), incorporating both geometric and causation probability approaches to realise real-time ship collision risk prediction and probabilistic risk assessment. Leveraging raw Automatic Identification System (AIS) data, the proposed model dynamically updates the probabilities of influential factors using Markov-chain-based transition analyses, mitigating uncertainties caused by noisy or incomplete data. In contrast to traditional deterministic models, the DBN captures mutual dependencies among dynamic risk factors, including variations in speed ratio, relative bearing, and temporal-spatial parameters such as Distance to Closest Point of Approach (DCPA), Time to Closest Point of Approach (TCPA) and relative distance. The model categorises collision risk into five discrete levels, ranging from very low to very high, providing decision-makers with actionable insights for real-time navigational safety. A key innovation lies in modelling these interdependencies among influential factors, which enables a holistic understanding of collision dynamics. Simulation results demonstrate that the DBN model outperforms traditional Collision Risk Index (CRI) approaches, particularly in accurately predicting complex collision scenarios and reflecting aggressive manoeuvres. This study presents a robust framework for maritime collision risk prediction, offering a foundation for enhancing navigational safety in increasingly congested and mixed-traffic environments involving the coexistence of manned and unmanned vessels.