Urban principal traffic flow analysis based on taxi trajectories mining

Abstract

The understanding of urban traffic pattern can benefit the urban operation a lot, including the traffic forecasting, traffic jam resolution, emergency response and future infrastructure planning. In modern cities, thousands of taxicabs equipped with GPS can be considered as a large number of ubiquitous mobile probes traversing and sensing in the urban area, whose trajectories will bring great insight into the urban traffic management. Thus, in this paper we investigate the urban traffic pattern based on the taxi trajectories, especially the principal Origin- Destination traffic flow (OD flow) extraction. Focusing on the picking-up and dropping-off events, the issue is solved by a spatiotemporal densitybased clustering method. The OD flow analysis is formulated as a 4-D node clustering problem and the relative distance function between two OD flows is defined, including a clustering preference factor which is adjustable according to the observation scale favor. Finally, we conduct the method on the taxi trajectory dataset generated by 28,000 taxicabs in Beijing from May 1st to May 30th, 2009 to evaluate its performance and interpret some underlying insights of the time-resolved results.