Feasibility Study of Urban Road Traffic State Estimation Based on Taxi GPS Data

GPS-equipped probe vehicle has become an effective way to collect information for real-time traffic state estimation. Taxis on urban road network are very suitable as this type of vehicles because of their low cost, wide coverage and easy data access. However, they have some special operation features, varying from other kinds of vehicles. Therefore, a doubt whether the GPS data from taxis can precisely reflect traffic state is put forward. This paper validated the effectiveness of traffic state estimation based on taxi GPS data by experimental analysis, which is carried out on the data collected from 8,013 GPS-equipped taxis travelling on urban road network for 15 days. We tested three widely applied GPS-based traffic speed estimation methods (average method, integral method and fitting method). Firstly, Pearson correlation coefficient was applied to calculate the relationships between true and estimate traffic state. Secondly, we compared the estimation accuracy in different segments, passing taxi number and collected points. Based on the strategies concluded from these experiments, we presented and validated three optimized methods to improve the traffic state performance. These experiments confirm that it is effectiveness of applying taxis as probe vehicles, and the optimization methods is helpful to reduce influence of special operation features.