Topic: Estimating Traffic Speed Using Cellular Phone Data
Presenter: Xianfeng Song
Traffic speed is an important parameter of traffic flow, which refers to the average speed of vehicles on a roadway segment over a period of time. It could be measured by using inductive loops or cameras, but only subject to critical crossroads or segments due to the cost of installations and management. In recent years, using vehicles with cellular phone as probes has been a hot topic to estimate traffic flow on roadways since it takes full advance of existing cellular systems to detecting cellular phones in motion. In general, the base stations of cellular phone system are sparsely distributed in urban area in comparison of dense road networks. For a vehicle moving from one cell to another, it is difficult to find its real path and get its speed for calculating traffic speed. This paper proposed a set of models to detect phones in motion and estimate traffic speed. First the analysis of phone calling events was carried out. The static events containing a ping-pong handover effect are filtered out and the events of motions are thus left for traffic speed estimation. Next, an algorithm of Mixed-Integer Linear Problems (MILP) was used to match the paths of vehicles to road networks and further calculate its moving speed on roadways. Finally, the traffic speed on a road segment was calculated by the statistics of the speed of vehicles on the segment. The above approach was implemented using an Open Source GIS solution (Python Plus NetworkX and PostgreSQL). The experiment results show that the estimate of traffic speed is reasonable in comparison of the measurement data by cameras located in the study area of Hangzhou City, China.