Scalability and Evaluation of Crowdsourced Traffic-Sign Mapping

Determine how traffic-sign map quality scales with the number of contributing vehicles, how multi-vehicle traffic-sign detections should be optimally combined, and how crowdsourced mapping systems should be evaluated using metrics that jointly account for localization and detection performance.

Background

The paper discusses crowdsourced updating of high-definition maps using observations collected by connected vehicle fleets. Traffic signs are discrete landmarks whose map quality depends not only on positional accuracy but also on correct semantic classification, false-positive control, and recovery of missed detections.

Existing approaches have demonstrated the feasibility of crowdsourced map construction, but the paper identifies unresolved issues concerning the relationship between fleet size and map quality, the optimal aggregation of heterogeneous multi-vehicle observations, and the design of a unified evaluation measure. The proposed simulation framework and the GOSPA_M metric are introduced to investigate these issues, but the broader questions are presented as open rather than fully resolved.