Map-matching based on driver behavior model and massive trajectories

Map-matching is the ground work of ITS (Intelligent Transportation System) related researches. In this paper, we explore how to infer the most probable path from low-sampling-rate trajectory based on drivers´ behavior model and large-scale GPS data collected by taxis. We make contribution in three ways. Firstly, we introduce SF-Feature to quantitatively measure taxi drivers´ preferences of roads with respect to sections of trip. Secondly, based on SF-Feature and our observation that drivers prefer to the roads they are familiar with and frequently use, drivers´ route-planning behavior is formulated to an optimization problem constrained by travel-time and travel-length. Thirdly, an efficient greedy algorithm is employed to seek out the solution. Finally, we compare our map-matching algorithm SF-Matching to other state-of-art low-sampling-rate algorithms. The experiment demonstrates that SF-Matching is superior to the others.