Unified Identification Method for Different Types of Traffic Conflicts Based on Vehicle Projection

Abstract

As a common method for identifying traffic risks, the traffic conflict technique needs to be more uniform in identifying different types of traffic conflicts and comparing the effectiveness of different conflict identification methods. To establish a unified identification method for traffic conflicts, this study describes the conditions under which vehicles collide through vehicle projection and constructs the Traffic Conflict of Unified Identification (TU) based on the time of vehicle collision to estimate the degree of risk of collision between vehicles. Using HighD high‐resolution traffic flow trajectory data with traffic conflict counts and cumulative exposure collision times, the differences and causes of different methods for all‐type and subtype conflict recognition were comparatively analyzed. From the perspective of a single traffic conflict, the correlation between conflict severity and the intensity of risk‐avoidance behaviors is proposed to compare and analyze the capability of the identification methods. The results show that (1) the unified identification method can better identify traffic conflicts without distinguishing the type of traffic conflicts, reducing the misjudgment of following conflicts and omission of lane changing conflicts to a greater extent, and being less affected by the predetermination of conflict types. (2) Compared with the traditional time to collision (TTC), TU has a better correlation between conflict severity and the intensity of vehicle‐avoidance behaviors, regardless of whether the conflict severity is characterized by the minimum collision time or the cumulative exposure collision time. The unified identification method improves the accuracy of traffic conflict identification, expands the scenario applicability of traffic conflict technology, and provides a new way for real‐time traffic conflict technology risk prediction for autonomous driving.