Vision-only fully automated driving in dynamic mixed-traffic scenarios

Abstract

In this work an overview of the local mo-tion planning and dynamic perception framework withinthe V-Charge project is presented. This framework en-ables the V-Charge car to autonomously navigate in dy-namic mixed-traffic scenarios. Other traffic participantsare detected, classified and tracked from a combinationof stereo and wide-angle monocular cameras. Predictionsof their future movements are generated utilizing infras-tructure information. Safe motion plans are acquired witha system-compliant sampling-based local motion planner.We show the navigation performance of this vision-onlyautonomous vehicle in both simulation and real-world ex-periments.