Data-Driven Analysis of Encounter Scenarios to Quantify COLREG-Compliant Autonomous Vessel Behavior

As the maritime industry advances toward increased autonomy, it becomes essential to ensure that autonomous systems incorporate reliable and predictable collision and grounding avoidance algorithms. The safe and effective development of Maritime Autonomous Surface Ships (MASS) requires a deep, quantitative understanding of how human mariners navigate in complex, real-world traffic situations. While the International Regulations for Preventing Collisions at Sea (COLREG) provide a necessary legal and operational framework, their intentionally flexible and ambiguous language delegates the specifics of collision avoidance to the professional judgment of experienced navigators. To accurately reflect how COLREG is applied in practice, a data-driven approach that analyzes historical navigation behavior to model real-world decision-making is essential. This paper presents a data-driven methodology for an automatic analysis of maritime encounter situations using historical Automatic Identification System (AIS) data to capture prevalent navigational practices. A case study is conducted in a specific region in Norway, identifying crossing encounters involving a ferry and another vessel, and the approach is designed to be scalable for application in other areas in future work. The identified encounter situations are further used to examine subsequent mariner maneuvers. In particular, we propose and compare several techniques for quantifying critical maneuver characteristics, including the predicted Distance at the Closest Point of Approach (DCPA), the resulting inter-vessel passing distance, and compliance with COLREG. Applying this comprehensive framework to an extensive regional AIS dataset yields aggregate statistics that establish an empirical baseline of human collision avoidance behavior. These results provide valuable quantitative insights into how COLREG are interpreted in real-world scenarios and contribute to the assessment and assurance of autonomous vessel behavior with respect to COLREG compliance. Ultimately, this work contributes to bridging the gap between regulatory intent and practical seamanship, enabling the development and assurance of autonomous vessel behavior that supports safe, predictable, and seamless interaction with human operated maritime traffic.

Keywords: AIS data, COLREG, collision avoidance, autonomy, evaluation, maneuver analysis.