Collective Perception Services (CPS) provided by roadside infrastructure is a promising way to extend the environment perception range of vehicles beyond their embedded sensors. However, sharing the perceived lists of objects can result in unacceptably high communication overhead, when the number of objects is large. We propose a Layered Costmap based novel perception sharing protocol to enable an overhead-efficient CPS, specifically in cases with many objects, such as crowded intersections. We demonstrate a CARLA-based collaborative autonomous vehicle simulation platform with realistic vehicular communications model, and evaluate costmap-based and object-list-based CPS in a T-junction scenario. Simulation results show that CPS from roadside infrastructure can significantly increase the perception coverage by 2.2x for T-junction scenario compared to on-board sensors’ perception. Moreover, layered costmap-based CPS outperforms object list-based CPS in efficiency enabling more than one order of magnitude overhead reduction for 40 or higher perceived objects.
Demo: Infrastructure Assisted Efficient Collective Perception Service for Connected Vehicles
2021-11-10
1054415 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Terrain Perception for DEMO III
NTRS | 2000
|Terrain perception for DEMO III
IEEE | 2000
|Terrain Perception for DEMO III
British Library Conference Proceedings | 2000
|