The collision avoidance system is one of the core systems of MASS (Maritime Autonomous Surface Ships). The collision avoidance system was validated using scenario-based experiments. However, the scenarios for the validation were designed based on COLREG (International Regulations for Preventing Collisions at Sea) or are arbitrary. Therefore, the purpose of this study is to identify and systematize objective navigation situation scenarios for the validation of autonomous ship collision avoidance algorithms. A data-driven approach was applied to collect 12-month Automatic Identification System data in the west sea of Korea, to extract the ship’s trajectory, and to hierarchically cluster the data according to navigation situations. Consequently, we obtained the hierarchy of navigation situations and the frequency of each navigation situation for ships that sailed the west coast of Korea during one year. The results are expected to be applied to develop a collision avoidance test environment for MASS.

The reliability of collision avoidance systems for Maritime Autonomous Surface Ships is one of the most critical factors for their safety. In particular, since many ship collisions occur in coastal areas, it is crucial to ensure the reliability of collision avoidance algorithms in geographically limited coastal waters. However, studies on maritime autonomous surface ships collision avoidance algorithms mainly focus on the traffic factor despite the importance of the geographic factor. Therefore, this study presents a methodology for establishing a practical collision avoidance system test bed, considering the geographic environment. The proposed methodology is a data-driven approach that objectively categorizes collision risk situations by extracting these risks using Automatic Identification System (AIS) and Electronic Navigational Chart (ENC) data, followed by clustering algorithms. Consequently, the research results present a direction for establishing test beds from the perspective of geographic and traffic factors.

In the maritime industry, the systematic validation of collision avoidance systems of autonomous ships is becoming an increasingly important issue with the development of autonomous ships. The development of collision avoidance systems for autonomous ships faces inherent risks of programming errors and has mostly been tested in limited scenarios. Despite efforts to verify these systems through scenario testing, these scenarios do not fully represent the complex nature of real-world navigation, limiting full system verification and reliability. Therefore, this study proposed a method for analyzing collision risk situations extracted from AIS data through graph-based modeling and establishing validation scenarios. This methodology categorizes collision risk scenarios according to their centrality and frequency and demonstrates how simple collision risk situations gradually evolve into harsh situations.

In the second autonomy stage in autonomous ships, reducing the number of bridge operators may increase the risk of accidents due to human error, especially the loss of situational awareness in collision risk situations. Because autonomous and remote-control systems are not yet well developed, bridge operators play a critical role in preventing collision accidents. However, most recent studies have focused on the technical development of autonomous ships, paying only a little attention to the monitoring of bridge operators. Similar to driver-monitoring systems in autonomous vehicles, there is a need to monitor the situational awareness of bridge operators on ships. In this study, we developed a model for evaluating the collision risk situation awareness (CRSA) of bridge operators based on their behavior. The results show that the model is effective in predicting the CRSA of bridge operators and is promising for practical applications.