Estimation of travel time distributions for urban roads using GPS trajectories of vehicles: a case of Athens, Greece

• Published in: Personal and ubiquitous computing Vol. 25; no. 1; pp. 237 - 246
• Main Authors: Adnan, Muhammad, Gazder, Uneb, Yasar, Ansar-ul-Haque, Bellemans, Tom, Kureshi, Ibad
• Format: Journal Article
• Language: English
• Published: London Springer London 01.02.2021; Springer Nature B.V
• Subjects: Availability; Computer Science; Cost benefit analysis; Datasets; Empirical analysis; Global positioning systems; GPS; Mobile Computing; Original Article; Personal Computing; Planning; Project management; Roads; Roads & highways; Statistical distributions; Time of use; Traffic assignment; Traffic control; Traffic management; Traffic planning; Transportation planning; Travel; Travel time; User Interfaces and Human Computer Interaction; Variability; Vehicles; Greece; GPS-based big data; Statistical distribution; Travel time distribution; Value of travel time variability; Athens city
• ISSN: 1617-4909; 1617-4917
• DOI: 10.1007/s00779-020-01369-4

Investigating travel time distribution and associated variability is important for a variety of transport planning, traffic management, and control projects. Studies that investigated travel time distribution tend to be limited to explore changes in characteristics of distribution with respect to space and time of day. Given the availability of big dataset that contains seven different types of vehicle trajectories in the city of Athens for around 56,000 trips which are traversing on more than 1.8 million road links, this study presents the detailed investigation of travel time distribution in different spatiotemporal settings. The study considered four different types of urban roads and six time intervals along with consideration of weekdays and weekends. The empirical investigation employed Kruskal-Wallis, Chi-square, and Kolmogorov-Smirnov tests to fit travel time data into seven unimodal statistical distributions that are found in the literature to describe travel time distribution. It is found that lognormal distribution outperformed other distribution, and all of the considered categories of travel time data are well-fitted to this distribution. Additionally, parameters of lognormal distribution for different categories of travel time data are not significantly different from each other, which led to the conclusion that travel time distribution is roughly independent of space and time, which is in agreement with a few earlier studies that are limited in their scope especially in relation with availability of data. With this important finding, this study estimate values of travel time variability for different classes of individuals employing a standard approach that requires time of day independent standardized distribution of travel time. It is estimated that for Athens population value of travel time variability is approximately half of the value of travel time. This is useful to carry out cost-benefit analyses for mobility-related projects in Athens, Greece.