Holistic Driving Scenario Concept for Urban Traffic

Abstract

Scenario-based approaches are important for verifying and validating automated driving systems due to the complexity of traffic, which cannot be covered by on-road tests. A challenge is the definition of a scenario catalog that is not too abstract to address this complexity, but still keeps a manageable number of driving scenarios. This paper presents a semi-formalized approach to derive a scenario catalog, resulting in less than 300 base-scenarios, for use in a driving scenario database for safety validation of automated driving. The approach defines abstract concepts as collections of characteristics to distinguish different driving scenarios. Within high-level categories, certain concepts are combined, either as full factorial of characteristics, or by considering, that the combination of characteristics implies other characteristics. Superclasses are also defined to group scenarios with common characteristics. The approach efficiently represents real-world interactions and can cover many existing catalogs within predefined constraints.