Structured Metric Approach for Marine Collision Avoidance Algorithms

This technical report presents a structured metric approach for evaluating marine collision avoidance algorithms, particularly in the context of autonomous ships. The report outlines the challenges posed by the rapid advancement of autonomous ship technologies, which have outpaced the development of legislative frameworks. It emphasizes the need for a well-defined process to certify new algorithms and systems onboard. The authors introduce a set of testing scenarios designed for evaluating path-planning algorithms, focusing on the determination of applicable COLREG rules and the computation of evasive routes. The report details the formulation of dedicated metrics to quantitatively compare the performance of various collision avoidance strategies, facilitating a human-based compliance evaluation of COLREG adherence. The effectiveness of these metrics is demonstrated through testing an existing algorithm, and the report concludes with suggestions for enhancing the robustness of the testing scenarios to improve the evaluation process.