Developing robust traffic navigation scenarios for autonomous ship testing: an integrated approach to scenario extraction, characterization, and sampling in complex waters

The fast development of Maritime Autonomous Surface Ships (MASSs) marks a significant advancement in the shipping industry, offering enhanced performance, cost-effectiveness, and environmental sustainability. This study aims to develop a holistic methodology for extracting, characterizing, and sampling maritime traffic navigation scenarios to support the testing and validation of MASSs, ensuring their reliability before widespread deployment. It begins with a precise scenario extraction technique that can effectively capture dynamic interactions among ships over time, enhanced by an in-depth analysis of ship motion dynamics, geographical complexities, and spatial-temporal nested interdependencies. Subsequently, the extracted scenarios are characterized using newly created metrics and advanced models in a structured and integrated manner. This allows for the classification and parameterization of ship motion patterns, conflict complexities, and encounter types, thereby enhancing the interpretability of traffic co-behaviors. Finally, a hierarchical greedy sampling strategy is developed to adaptively select representative scenarios from a sufficiently realistic set, striking a balance between comprehensive scenario coverage and efficiency in MASS testing. Extensive experimental analyses validate the efficacy of the proposed methodology. It precisely identifies real evolutionary multi-ship encounter situations, finely characterizes scenarios to support the encoding, explaining, and understanding of dynamic traffic behaviors, and systematically selects representative scenarios by incorporating multiple selection principles. Consequently, this methodology makes new contributions to the pioneering development of an accurate, interpretable, and representative set of real-world traffic navigation scenarios for autonomous testing. This is crucial for assessing and validating the advancements in MASSs and their emerging functionalities, thereby promoting highly and fully automated navigation.