Detection and analysis of corner case scenarios at a signalized urban intersection.

One of the major challenges in automated driving is ensuring that the system can handle all possible driving scenarios, including rare and critical ones, also referred to as corner case scenarios. For the validation of automated driving functions, it is necessary to test the corner cases in simulation environments. However, the effectiveness of simulation-based testing depends on the availability of realistic test data that accurately reflect real-world scenarios. This work aims to detect, cluster, and analyze rare and critical traffic scenarios based on real-world traffic data from an urban intersection and prepare the data for usage in simulation environments. The scenarios are detected by filtering hard braking maneuvers, red light violations, and near misses under adverse weather conditions. A long-term analysis of trajectory, weather, and traffic light data was conducted to find these rare scenarios. Our results show that 24 hard braking maneuvers are included in our dataset with a duration of half a year. They occur due to failure to yield, emergency vehicle operations, and a red light violation. Some of the scenarios include crashes, lateral evasive maneuvers, or are under adverse weather conditions like fog. Altogether, we provide methods to extract corner case scenarios based on multiple data sources and reveal diverse types of corner case scenarios at an urban intersection. In addition, we analyze the behavior of road users in critical scenarios and show influencing factors to avoid crashes. By combining and converting the data to an industry standard for simulation we provide realistic test cases for the validation of automated vehicles. Therefore, the results are relevant for both, traffic safety researchers to learn from road user behavior in these rare scenarios and developers of automated driving systems to test their functions.