This content is not included in your SAE MOBILUS subscription, or you are not logged in.

Virtual Assessment of Automated Driving: Methodology, Challenges, and Lessons Learned

Journal Article
ISSN: 2574-0741, e-ISSN: 2574-075X
Published December 18, 2019 by SAE International in United States
Virtual Assessment of Automated Driving: Methodology, Challenges, and Lessons Learned
Citation: Wagner, S., Knoll, A., Groh, K., Kühbeck, T. et al., "Virtual Assessment of Automated Driving: Methodology, Challenges, and Lessons Learned," SAE Intl. J CAV 2(4):263-277, 2019,
Language: English


Automated driving as one of the most anticipated technologies is approaching its market release in the near future. Since several years, the research in the automotive industry is largely focused on its development and presents well-engineered prototypes. The many aspects of this development do not only concern the function and its components itself, but also the proof of safety and assessment for its market release. It is clear that previous methods used for the release of Advanced Driver Assistance Systems are not applicable. In contrast to already released systems, automated driving is not restricted to a certain field of application in terms of driving scenarios it has to take action in. This results in an infeasible amount of required testing and unforeseeable scenarios the function can face throughout its lifetime. In this article, we show a scenario-based approach that promises to overcome those challenges. In contrast to previous methods, it includes virtual test domains in a verified way to diminish the demand for real-world testing. Local verification of certain scenarios from real-world testing enables virtual variation in a local test space, and thus contributes to the test volume. The approach is implemented, evaluated, and shows first promising results. An intentional evaluation against a corner case-a scenario where the use of virtual test domains is challenged-from the set of scenarios reveals current weak spots. These are transformed into lessons learned, proposed solutions, and future work for enhancements. With the aid of these, the suggested methodology is a promising candidate to overcome the challenges of evaluating automated driving functions.