Analysing Experimental Results Obtained when Applying Search-based Testing to Verify Automated Driving Functions

Assuring safety in case of automated and autonomous driving is of uttermost importance requiring exhaustive search for critical scenarios allowing to reveal faults in current implementations. Different approaches like search-based testing have been already used to come up with test cases that allow to detect situations where the automated or autonomous driving function reacts in an unwanted way. In this paper, we contribute to the corresponding research and provide an in-depth analysis of results obtained using search-based testing applied to two different automatic emergency braking systems. We primarily focus on answering the question regarding the number of parameters required to find crashes. An answer to this question has implications for practice as well as for other testing techniques like combinatorial testing, where there is a need to identify the combinatorial strength in advance for assuring the detection of faults. Our analysis revealed important parameters of tests and we observed that interactions of at least 4 parameters are required to obtain critical scenarios of a high probability.