Project Outline

The efficient and effective safety assessment of automated driving functions is currently one of the most relevant research topics. For this purpose, widespread concepts from industry and research use scenario-based approaches in which the relevant characteristics of a traffic situation are filtered in order to store them in databases using a suitable scheme. The data can come from different sources (field tests, accident databases, simulation, UAV, etc.). A uniform database and standardized system are currently being researched and developed. Based on fka's work in the PEGASUS project, fka and its scientific cooperation partner, the Institute for Automotive Engineering (ika) at RWTH Aachen University, have been working for a year now on a research project (82.0729/219) commissioned by the Federal Highway Research Institute.

The goal of the two-year project is to derive a universally applicable concept for describing traffic situations within the driving domain controlled-access highway. Thereby, driving on the highway is to be divided into disjunctive elements, so-called basic scenarios, which depend on various factors that describe the situational traffic context. In this way, it should be possible to construct both simple and complex scenarios. The central result of the project is a codebook that describes the generally applicable scheme for the classification of scenarios in analogy to the accident type classification.

Solution Approach and First Results

To meet the requirements for a standardized and extensible system, the fka and ika experts are developing a methodology based on a building block logic. With the help of the literature and our expert knowledge, three superior categories were identified: driving states, transitions and superimposed interactions. Special care was taken to achieve a suitable level of granularity in order to make the application in scenario databases as practicable as possible. The building blocks included allow a large set of relevant scenarios to be constructed in the domain of controlled-access highways, while the resulting scenario catalog remains extensible.

Within the project, the compatibility of the building block approach with the concept for the definition of collision-relevant scenarios developed in the PEGASUS project is demonstrated. For the detailed consideration of scenarios during a lane change, the PEGASUS concept was extended by an explicit description of collision-relevant scenarios during a lane change.

For the first validation of the whole concept, an extensive workshop with more than 25 experts from different countries and domains has already been conducted virtually on Dec. 16, 2020. Since there was a very positive response in the context of this, we would like to make the results of the workshop available to all interested parties here. Fill out the form below to receive the workshop summary.

With the help of the newly gained knowledge from the workshop, the concept will be validated on the basis of data in the next step. For this purpose, the classification scheme will be implemented in software in order to identify the scenarios on the basis of real world data, for example the highD dataset from levelXdata by fka. Finally, the codebook will be created, which describes the classification scheme by means of an action guideline as well as formulates requirements for data collection in order to detect both basic and complex scenarios. Thus, this project creates the basis for a representative assessment of traffic situations as well as their systematic analysis .