Detection and Analysis of Critical Interactions in Illegal U-turns at an Urban Intersection

Kurzfassung

Before Advanced Driver Assistance Systems (ADAS) can guide vehicles through real-world traffic, it has to be ensured that they will operate reliably in normal, but particularly in rare and critical situations such as traffic conflicts or near misses under all circumstances and conditions. To test the ADAS functions in rare critical situations, this study aims to gather knowledge about such situations i.e. detect them at an urban intersection, analyze the road user behavior and describe relevant kinematic patterns based on an aggregated long-term analysis. To limit the number of possible situations, we focus on interactions between illegally U-turning motorized road users (MRU) and vulnerable road users (VRU). Since trajectory and video data of traffic violations are rare, the relevant trajectories of MRUs and VRUs need to be identified first. Therefore, virtual loops are employed, which are placed at the expected starts and ends of the trajectories. All trajectories that intersect both, the start and end loop, are extracted from the dataset. Then, the resulting trajectories have to be evaluated regarding driving paths, interaction, and criticality. For this purpose, the surrogate measure of safety "post encroachment time" (PET) is applied. Afterward, available scene videos are used to evaluate the PET-triggered situations as critical or uncritical encounters. Finally, descriptive and inferential statistical methods are applied to kinematic data of those trajectory pairs to identify relevant behavioral patterns of the road users. The examined dataset was recorded at the Application Platform for Intelligent Mobility Research Intersection of the German Aerospace Center in Brunswick, Germany. Applying the beforementioned methodology to the dataset yielded the detection of relevant interactions. The kinematic patterns of the interactions that were assessed as critical close encounters were further analyzed to derive situational patterns. Based on this analysis it can be shown that the reason for critical situations was that the U-turning MRU had to leave the intersection. Thus, we can validate that the road safety for vehicles leaving the intersection in an unallowed direction can become critical. To understand these situations in detail they are described in the following. The U-turning MRUs use the lane of the left turning vehicle and have to let the oncoming traffic pass before they can execute their turning maneuver. While the median U-turn curve radius is 7.6 m other curve radii vary between 2.8 and 22.3 m. Some U-turning vehicles that enter the intersection during the red phase of the VRU are waiting so long for the oncoming vehicles to pass that the traffic light for the VRUs is already switching to green when the U-turning vehicle leaves the intersection. Based on the PET-triggered situations and their video scenes we could identify and evaluate critical U-turn situations. Our analysis showed, that these situations occur when the vehicles had to wait a long time at the intersection and had to leave it at a time when the traffic lights gave the right of way to the VRUs that were crossing the lane. In a conclusion, tailored preventive measures such as vehicle-to-infrastructure communication could reduce criticality in such U-turn situations because the vehicles would then be aware of the traffic light state. The research leading to these results is funded by the German Federal Ministry for Economic Affairs and Climate Action within the project Methoden und Maßnahmen zur Absicherung von KI basierten Wahrnehmungsfunktionen für das automatisierte Fahren (KI-Absicherung). The authors would like to thank the consortium for the successful cooperation.