Cyclist Safety in Urban Automated Driving - Sub-Microscopic HIL Simulation

Abstract

The protection of Vulnerable Road Users (VRUs) is a major challenge in urban areas. This does not only apply to autonomous vehicles but to manually driven cars as well. This work deals with the impact of fully automated vehicles on the safety of cyclists in urban areas. We show the impact of different maximum sensor detection ranges on the safety of VRUs. The resulting driving comfort of the ego AV is analyzed. To ensure realistic results we combine a microscopic traffic simulation with a sub-microscopic simulation for the ego car. Additionally we occlude cyclists close to the intersection. This setup includes a Hardware in the Loop (HIL) system where we are running an open source automated driving planning and control system on a vehicle computer in real time. Thus we achieve an accurate simulation architecture which can evaluate urban scenarios in the most realistic way, using the current state of the art simulation software. Our findings point towards a high impact of occlusions on the safety of cyclists in urban intersections, while the maximum detection range of the AVs sensor systems has a minor influence in the considered scenarios.