Occurrence Estimation for the Classification and Prioritization of Concrete Scenarios in the Context of Virtual Scenario-Based Validation of Vehicles

This article deals with the creation and concretization of a scenario catalogue in the context of scenario-based validation of highly automated driving functions. Due to the possible combinatorics of the various driving situations and influencing parameters in the open-world context, it is not possible to simply execute all possible combinations. Instead, new methods are needed that can make a representative statement about the respective logical scenario, but at the same time attempt to reduce the degree of execution. This article aims to examine the usefulness of using (top-view) recordings of real world traffic to extract the occurrence of appearance of certain, predetermined scenario parameters of a logical scenario. This information on the occurrence is used to prioritize simulation efforts for more efficient and better evaluation of driving functions in the context of autonomous driving. To this end, a methodology for exposure estimation is to be developed as part of this work. This approach uses analogies to existing methodologies (e.g. Automotive Safety Integrity Level (ASIL) defined in ISO 26262 [1]) and transfers them to the scenario-based context to attach an exposure level to each concrete scenario. An existing public dataset of top-view camera recordings of real traffic scenarios was analysed and evaluated with respect to pre-defined scenario-specific parameters. Finally, the results will be evaluated and validated with an exemplary logical scenario.