Generation and complexity analysis of ship encounter scenarios using AIS data for collision avoidance algorithm testing

Abstract

Collision avoidance algorithms are critical for enhancing ship navigation safety. To ensure the reliability of these algorithms, comprehensive testing is imperative prior to their deployment. A pivotal aspect of such testing is the strategic formulation of testing scenarios, which plays a vital role in securing the trust of testing phase results. This study proposes a systematic method for generating ship encounter scenarios and analyzing their complexity using Automatic Identification System (AIS) data, to support the implementation of collision avoidance algorithm testing. Firstly, real-world ship encounter scenarios are extracted from AIS data by considering the spatial and temporal dependencies among ships. Subsequently, a complexity evaluation model based on encounter topology evolution is designed to assess the complexity of these scenarios. Finally, detailed strategies for using these scenarios in collision avoidance algorithm testing are explored based on the complexity evaluation results. Real AIS data from the waters off Ningbo-Zhoushan Port is collected to validate the performance of the proposed method. Research findings indicate that the proposed method successfully separates numerous ship encounter scenarios from historical AIS data and comprehensively evaluates their complexity. Based on this foundation, valuable guidance is provided for the application of extracted scenarios in collision avoidance algorithm testing.