Link Travel Speed Estimation Using Transit GPS Data

Abstract

This thesis studies the travel speed estimation at road link level using sparse transit GPS data. A link travel speed estimation method is first proposed, which estimates link travel speeds by inferring the timestamp when probe vehicles are passing specific locations along roadways. A field test is conducted on an urban freeway to evaluate the performance of transit bus-based link travel speed estimation using this method, and the impact of probe vehicle type and GPS update interval upon the estimation accuracy are analyzed. The test results suggest that the proposed method can provide reliable link travel speed estimates, with a mean absolute speed difference of 7.0 km/h compared to loop detectors. This approach assumes that the travel speed between two consecutive GPS points of a probe is similar, which is not reasonable when applying to urban arterials, as the existence of intersections and bus stops makes the travel time between two neighbor GPS points not as homogeneous as on freeways. Therefore, a link travel time allocation method for individual probes is developed to overcome this difficulty. The proposed travel time allocation method decomposes travel time into several parts and uses probability functions to estimate travel times of traversed links. A field test is conducted on an urban road corridor to evaluate the performance of the link travel time allocation method using transit bus probe data, which is compared with the previously mentioned link travel speed estimation method. The results show that the proposed link travel time allocation method can improve the estimation accuracy of individual link travel times, especially under congestion condition. It can also provide a good estimation of the travel delay caused by vehicles stopping at intersections or bus stops, which can be used to analyze the dwell time of transit probes at bus stops in the future.