Safety Assurance for Highly and Fully Automated Driving

Autoren

Dr.-Ing. Rasmus Adler, Patrik Feth,
Fraunhofer IESE, Kaiserslautern

Zusammenfassung

Many prototype cars demonstrate the feasibility of highly and fully automated driving. However, many open issues remain with respect to the assurance of safety. The implementation of automated driving requires the application of approaches that lead
to models that cannot be analyzed and to unpredictable vehicle behavior. Traditional safety engineering has to be revised in line with these changes in development because the currently applied methodologies for fault avoidance/prevention, fault removal, fault tolerance, and fault forecasting are specific for the current development practice and will not fit new approaches such as deep learning or behavior-based control. We investigate why and how safety engineering has to be revised. In this context, we explain why the current functional safety standard ISO 26262 is not sufficient for assuring safety in the context of automated driving. Furthermore, we outline solution approaches for handling the new challenges and relate them to each other in order to give a big picture about safety in the area of automated driving and the usage of deep learning.

Mitglieder-Login