Safe Perception: On Relevance of Objects for Vehicle Safety
Cornelius Buerkle, Fabian Oboril, Julio Jarquin, Frederik Pasch, Kay-Ulrich Scholl · 2021
Automated vehicles are a great promise for the future of transportation systems. However, safety assurance is still a major roadblock for the mass deployment of such vehicles. This is a great challenge especially for the perception systems that have to handle a diversity of environmental conditions and road users. Up to now, perception systems treat the complete environment and all traffic participants in the same way, although only a subset of all objects has an influence on vehicle safety. To close this gap, we present in this work a comprehensive definition of safety-relevant objects, and for the most critical area around the vehicle, the safety-relevant area. Using these definitions, we demonstrate that a common object detection system is not able to detect all safety-relevant objects, which will make new, safer approaches necessary in future.