Identifying relevant traffic situations from simulation data for testing ADAS and fully automated vehicles - A multilayer model concept

This work presents a multilayer model concept for identifying and rating relevant and unknown traffic situations from simulation data. This concept ensures full testing and validation of Advanced Driver Assistance Systems (ADAS) and fully automated vehicles. Following consistent terminology, this paper first discusses known definitions of general terms including scene, situation and scenario. Subsequently, well-known criticality metrics are summed up and assessed with regard to their potential to test ADAS and fully automated vehicles. As far as we know, the discussed criticality metrics are not applicable to identify all traffic situations which are relevant for fully automated driving and ADAS. To overcome this limitation the proposed multilayer model concept first filters potentially relevant situations. This is done by generating manoeuvre spaces and taking further information of a scene into account. Then, a grade of influence on the target vehicle is calculated to rate situations. Besides introducing the concept, the pre-filtering algorithm will be demonstrated using an interactive simulation tool.