The Safe AI Engineering project focuses on ensuring the safety of AI-controlled components for automated and connected driving. The project aims to develop a continuous, lifecycle-level safety methodology to ensure safety in dynamic environments. By integrating its results into established standards such as ISO 26262 and SOTIF, the project bridges the gap between AI verification and established safety assurance for the automotive industry.
The project focuses on camera-based detection of vulnerable road users (VRUs), such as pedestrians, using deep neural networks. Three increasingly complex scenarios are defined – from the detection of a static pedestrian to the handling of multiple dynamic VRUs in urban environments.