Abstract

One of the most important and current future trends in the automotive industry is the development of Advanced Driver Assistance Systems (ADAS). Due to the ever-increasing complexity of ADAS, the safety validation of such systems is a major challenge. New methods have to be developed, as the previous certification and approval methods are not suitable for this use case. E.g. Monte-Carlo simulation, combined with Software-in-the-Loop (SiL) simulation may help to overcome this limit. In order to achieve this goal, scenario-based testing for the safety validation of highly automated driving systems, where specific traffic scenarios are parametrized, simulated and analyzed by a set of criteria, is the only possible approach to efficiently test and validate thousands of concrete scenarios. By using distribution functions for each input parameter in this approach, a safety statement can be given by approximating the probability of failure for each traffic scenario. This is done by determining the transition between the safe and unsafe region in the parameter space. This process heavily relies on data from real-world traffic scenarios to derive the necessary scenario information for testing. Thus, we present a methodology based on a qualitative modelling of the searched scenarios, by using universal pattern elements of an ontology. Furthermore, depending on the identified traffic scenarios, distribution functions are generated. Additionally, correlations between the extracted parameters are considered by using the Nataf Transformation. Then, a combined criterion is introduced, to describe the criticality of a simulated traffic scenario. This enhances a clear separation between the safe and unsafe region. Finally, an efficient and robust search strategy is proposed by combining dimension reduction technics, surrogate models with Neural Networks and advanced methods of the reliability analysis (Importance Sampling using multiple Design Points). It can be shown that the number of simulations carried out can be significantly reduced in comparison to Monte-Carlo.