Crowdsourcing turning-restrictions from map-matched trajectories

摘要

The availability of GPS-enabled devices has generated massive amounts of GPS tracking data produced by vehicles traversing the road-network. While initially used for improving traffic estimation and routing, only recently has this data been used for map-construction efforts. This work focuses on the specific aspect of identifying turning restrictions in the underlying road-network graph. We propose a novel, efficient and straightforward method to deduce turning restrictions for OpenStreetMap data, by mining historic map-matched trajectories from an existing fleet-management service. Our extensive experimental evaluation and verification process utilizing online map-services, satellite imagery, street view and public map-data APIs proves the efficiency and reliability of the proposed method.