Testing Scenarios Construction for Connected and Automated Vehicles Based on Dynamic Trajectory Clustering Method

Abstract

In order to verify the safety of connected and automated vehicles (CAVs), it is necessary to conduct comprehensive testing in virtual simulation environments, closed facilities, and public roads. Scenario-based testing is an important method to test CAVs, and the key step of this method is constructing testing scenarios. Powered Two-wheelers (PTWs) riders are more vulnerable than other road users due to the lack of protection. Since they often drive at high speeds and uncertain routes, the driving scenarios containing two-wheelers are more dangerous for CAVs. In this study, we first extracted 164 cases of crashes between passenger cars and two-wheelers from the China In-depth mobility safety study-traffic accident (CIMSS-TA) database. Then, the pre-crash scenarios are reconstructed case by case with the obtained variables. Next, we utilize the crash reconstruction software PC-Crash to reconstruct each collision and get the pre-crash trajectory for each participant. Finally, the trajectory data are clustered and analyzed by the k-medoids algorithm, and six typical crash scenarios between passenger cars and two-wheelers are obtained. The research results will contribute to constructing high-risk test scenarios for CAVs.