Generating test scenarios for assessing automated driving systems using scenario fusion

This thesis presents the find-unify-synthesize-evaluate for representativity (FUSE4Rep) process model, a novel approach to the safety evaluation of automated driving systems (ADS). Designed to make road traffic safer by preventing accidents, ADS must demonstrate a higher level of safety than human drivers. FUSE4Rep addresses the challenge of unifying divergent information from sources such as police accident data and video-based traffic observations to ensure a comprehensive scenario representation. Through scenario fusion, the process synthesises diverse traffic data into a representative scenario catalogue, enabling a thorough assessment of ADS over a wide scenario space. Using statistical matching, it derives and varies logical scenarios to cover potential real-world conditions in stochastic simulations. A case study shows how German police accident data and video-based observations are used to create a fused scenario catalogue, demonstrating the practical application of FUSE4Rep. As part of the comprehensive “Dresden Method” for ADS evaluation, this approach provides a reliable framework for the development of safer ADS and contributes to improved road safety.