Dynamic Risk Managment for Autonomous Vehicles in Pedestrian Zones

The safeguarding of highly automated commercial vehicle functions is based on a risk assessment of the system behavior in the intended operational environment. Current approaches rely on worst-case risk assessments that lead to unacceptable availability. A significant reason for this is overly conservative behavioral assumptions for the system users, people, and machines in the immediate environment. Dynamic risk assessment (DRA) approaches promise to remedy this situation by making behavioral assumptions based on the case. Thus, system behavior is based on the dynamic risk of the current situation rather than the maximum risk of all possible conditions. The goal of this work is the integration of models for dynamic behavior and risk prediction in automated driving functions that have been developed at Fraunhofer IESE and the RRLab at the TU Kaiserslautern. Further, it aims for an evaluation of the performance potential of the resulting overall model. Both simulation and real-vehicle environments in an autonomous shuttle bus are used as platforms for the review.

Dynamic approaches reduce the risks of autonomous systems have great potential in the commercial vehicle sector since no solutions have yet become established in this area. The current project is intended to provide the basis for further know-how development.