Identifying Critical Test Scenarios for Lane Keeping Assistance System Using Analytic Hierarchy Process and Hierarchical Clustering

Lane keeping assistance system (LKAS) is critical components of autonomous driving technologies, and their reliable operation is crucial to ensure passenger safety and maintain traffic flow. The validation of LKAS's safety and performance requires extensive testing using diverse and representative test scenarios. However, identifying the most critical test scenarios from a large data is a significant challenge. Inspired by Scenarios Engineering, the present study introduces a novel approach that includes being the first to combine analytic hierarchy process (AHP) and hierarchical clustering (HC) to identify critical test scenarios for the comprehensive evaluation of LKAS. The proposed approach first utilizes the AHP to rank the key factors that contribute to the failure of the LKAS. Then, by taking the key factors as clustering parameters, the HC method is employed to cluster LKAS failure scenarios with similar characteristics. The experimental results suggest that our method is capable of effectively clustering 200 LKAS failure scenarios into 9 categories of critical, reasonable, and reliable LKAS test scenarios. Furthermore, these 9 categories of critical test scenarios can provide guidance for the development and testing of autonomous driving and serve as a foundation for establishing a testing and evaluation system for LKAS in China and around the world.