On Reliability of Routes Computed Based on Crowdsourced Points of Interest

Abstract

Today’s routing services provide routes that meet different needs and preferences of different users. To compute desired (optimal) routes, these services must support databases that contain accurate origin and destination locations, a high accuracy road network database, and an optimization algorithm. Origin and destination (O/D) locations are commonly collected by professional, commercial, and crowd sources via manual or automatic (geocoding) approaches. The routes computed for the same pairs of locations (O/D) obtained from different sources may be different. Considering the increased interest in collecting points of interest (POIs) through crowdsourcing, in this chapter, we address this research question: Are the routes computed using crowdsourced POIs (as O/D) reliable? To address this question, we conducted experiments where routes (shortest and fastest) computed using crowdsourced POIs (e.g., through OpenStreetMap) were compared with the routes computed using POIs obtained from professional and commercial sources. Metrics including route length, travel time, Euclidian distance, and number of road segments were used in the comparisons. The results reveal that, in general, though there are no significant differences between the routes, differences usually occur at or near origins and destinations.