A spatial–temporal forensic analysis for inland–water ship collisions using AIS data

Ship collision is a major type of accident in waterway transportation, with considerable annual occurrence that always causes serious casualty. Among the evidences that can facilitate the investigation, AIS data recorded during the context of the collision are highly informative. This paper exploits the contextual AIS data in a coupled spatial–temporal perspective. With the discrete waypoints specified by the AIS data, the trajectories of the colliding ships are regenerated with three reasoning respects: (1) the conflict-resolution envisioned by each ship in the collision-inducing scenario when the two ships are drawing near; (2) to what extent the two ships can make situation assessment and the de facto maneuvers exerted to steer the ships; (3) how to evaluate the decision quality of the steering process and how to measure the consistency between the subjective endeavors and the objective results in the context of bilateral anti-collision maneuvers. According to the periodical nature of the AIS data, time is slotted into slices to make temporal analysis and to sketch out the decision sequences made by both ships. Fuzzy set method is adopted to estimate the quality of the decisions, and a Dempster–Shafer Theory (DST) based method is employed to fuse the discernment in time domain and in spatial domain, so that an overall impact of the behavior of both ships can be calculated. The current paper presents a microscopic analytical scheme for a two-vessel collision accident investigation, and the proposed scheme may be heuristic to yield a generalized result for a multi-vessel case.

► Using spline-based method to obtain the synchronized states of each vessel.
► Measuring the decision and maneuvering quality of vessels in the context of collision.
► Using Dempster–Shafer theory to model the uncertainty of decision rationality.
► Evidence combination is conducted in a spatial–temporal perspective to reach a more objective result.
► Using the ambiguity formula to calculate the uncertainty involved in the analysis.