© 2020 Elsevier Ltd The rapid increase in the number of vehicles equipped with GPS devices has resulted in using automatic vehicle location (AVL) data as probes to identify traffic flow status as well as route travel speed on a very fine spatial-temporal scale. However, these traffic monitoring approaches heavily rely on the widely distributed probe vehicles in the network and the high frequency of these probe samples, which are rarely implemented in the real world. This study aims to analyze the applicability of providing accurate traffic flow information from four types of low-frequency AVL data. Each data source is applied for speed estimation to develop guidelines on GPS data requirements for travel speed estimation. First, the probe sample size of each data source on each target corridor is studied to reveal the road segments that have the potential for speed estimation, along with the GPS sampling frequency of each data source. Second, the impact of probe vehicle types, sample sizes, and GPS sampling frequency is analyzed. This study offers guidance in using GPS data to conduct speed estimation in different scenarios, which can be further implemented in a prototype software tool for estimating the real-time travel speed. This study has shown the applicability for speed estimation from four types of GPS data, where the transit bus GPS data provides the best mean speed estimation. The speed estimation results are compared with loop detector data on a test road segment to evaluate its accuracy. The comparison results show that given the current GPS data sample size and updating frequency, the transit bus GPS data can provide a reasonably accurate estimation of the traffic flow speed with a mean absolute speed difference of 6.96 km/h.
Available for download on Saturday, July 02, 2022