The challenge of proximity: Evaluation of map matching results in real road networks

Abstract

The increasing availability of vehicle tracking data enabled by global positioning system (GPS) trackers provides researchers with more data to study human mobility. To study how people move in the road network, it is a crucial step to match vehicle tracking data to corresponding road segments, which is defined as map matching. Map matching toolkits that are currently used in practice usually produce results with matching errors. These errors are more prominent in a dense road network, where road segments are close to each other. This study evaluates which parts of the road network are prone to map matching errors and designs a model to detect the errors. Detecting the errors is the initial step before fixing them and producing a more accurate map matching result.
This study is based on a real trajectory dataset and the road network data is from Open Street Map (OSM).