Simulation-based safety testing of Automated Driving System (ADS) is a cost-effective and safer alternative to field tests. It allows for comprehensive evaluation of self-driving cars in virtual environments. However, ADS encounters many scenarios based on scenery, traffic and road objects, environment, road geometry, and maneuver. It is practically impossible to test every scenario using a simulator.

In this paper, we propose a process for prioritizing and selecting scenarios from the existing lists of scenarios. The aim is to refine the scope of tested scenarios and focus on the most representative and critical ones for evaluating ADS safety. Our process excludes scenarios not covered by the ADS Operational Design Domain (ODD). The remaining scenarios are then categorized into distinct groups based on similarity in the critical actions performed by the ego vehicle or the target object. The scenario groups frequently encountered in actual driving situations are prioritized using real-world data. Scenarios that may not be feasible to implement due to the limitations of the simulation environment are filtered out and kept in a separate list in reserve. Within each scenario group, scenarios are prioritized considering the frequency of occurrence of scenario elements in road accident datasets. Finally, the scenario with the highest score in each prioritized group is selected for testing ADS via simulation.

We apply our process to two pre-existing scenarios catalogs provided by the Land Transport Authority of Singapore and the Department of Transportation. The total number of scenarios in both selected catalogs is 111. After applying our process, we prioritized and selected six scenario groups containing 51 scenarios for testing ADS in the CARLA simulator. Our process for scenario prioritization and selection in simulation-based safety testing of ADS is a structured and organized approach, contributing to the development and trustworthiness of autonomous vehicles.