Identification of hazards in urban driving situations using accident categorizations

In contrast to human drivers, automated vehicles (AV) can observe its complete surrounding in every moment and take precise measurements. The decision-making process and computation of feasible trajectories are very rapid and actuation can happen almost without any delay. Therefore, research suggests, that automated vehicles will contribute to make traffic safer. The deployment in the market, though, requires assurance of the safety of an automated system through validation and verification. A unified process to do so does not yet exist. In particular, when the use cases of the system comprise complex situations such as those occurring in urban traffic, the process might become very complex. A first step to design AVs and approach the verification of safety is the identification of hazards. This paper introduces a description of urban driving situations similar to that of accident categorizations used in traffic safety research. Driving situations as well as real accidents from an urban area are described in this format. With the help of Data Matching, the relevant attributes of the accident categories are matched to the driving situations and thus hazards are identified for the respective situations. Additionally, a comparison with the real data can help to rate the hazards.