Abstract—Link travel time is one of the most important variables in determining travel routes and start time of journeys. Link travel time is the basis of navigation and routing systems. Time dependent algorithms calculate fastest routes based on link travel time of transportation networks. By increasing use of portable receivers of global positioning systems, researchers are more interested in utilizing the data provided with such equipment in monitoring traffic states. In this research, travel time of arterial links is estimated in real time through the data received from Tehran’s transit buses AVL system. Link travel time estimation requires some analyses on spatial and temporal data. In this regard, the Holt-winters analysis is used for a short term travel time prediction. Traffic signal timing data are also involved in the computations. Three test vehicles are utilized to provide data required in validating proposed method and also estimating parameters of the model. Finally, estimated travel times are compared with the results of the baseline method. The RMSE of the proposed method indicates the accuracy of the travel time prediction provided by the proposed method. This accuracy indicates the efficiency of the proposed method in the link travel time estimation, which could be a promising outcome for using buses as traffic probes. Keywords—Bus probes, Link travel time, AVL system, Holt- Winters time series analysis. I. INTRODUCTION S schedules become more accurate, the importance of planning is more evident than ever. A large part of people’s time is wasted in crowded transportation networks that not only leads to increased fuel consumption, but also is a treat to community health. Many location based services are designed to help people with their travel plans. While accessing up to date and reliable information of traffic state throughout huge transportation networks is an integral part of such services, developing novel and automatic methods of traffic monitoring is necessary in such a situation. Some methods have been already used for arterial link travel time estimation. Taylor et al. has divided the current methods into two main categories [ 1 ]. Rouzbeh Forouzandeh 1 is MSc student in GIS, Department of Geomatic engineering, University of Tehran, Iran (corresponding author’s phone: 00989370862917; e-mail: roozbehforoozandeh@gmail.com). Navid Khademi 2 is with the Department of Civil Engineering, University of Tehran, Iran (e-mail: Navid.khademi@ut.ac.ir). Farshad Hakimpour 3 is with the Geomatic Engineering Department, University of Tehran, Iran (e-mail: fhakimpour@ut.ac.ir). The first category estimate link travel time based on fixed detectors. Such methods are known as site base methods. The second category of are those in which link travel time is estimated through utilizing the position of moving vehicles in traffic flows. These methods are based on the data provided by the probe or floating vehicles. Probe based methods can be categorized as GPS based, Bluetooth based, and RFID based methods. Probe vehicles data have already been used in estimating link travel time. Van Zuylen et al. modeled delays at traffic signals in various traffic conditions by using probe vehicle data [ 2 ]. They estimated delays at traffic signals by calculating parameters of delay distribution functions through rebuilding traffic light queues. In their study, the Genetic Algorithm method along with the maximum likelihood estimator was employed to determine the optimum parameters of the delay distribution functions at traffic signals. Hofleitner et al. introduced a new approach for arterial link travel time estimation by using GPS data gathered from taxicabs of California [ 3 ]. They divided link travel time into two parts including travel time of free flow speed and delays related to traffic signals. The researchers modeled distribution function of travel time and then estimated the parameters by using GPS data received from taxi fleet. They assumed that traffic flow has a uniform distribution while modeling link travel time distribution. The results indicated that Hofleitner’s proposed approach has a better performance compared with older methods. Along with cars, buses have also been used as traffic probes. Nobuhiro et al. used bus probes to investigate the travel time variability [ 4 ]. They acquired travel time distribution functions by analyzing AVL data of transit buses. There are also some other researches devoted to bus probes [6], but they rarely address the link travel time estimation. This paper proposes a novel method using transit buses for monitoring traffic state of arterials. In many cities, transit buses are equipped with AVL 1 systems by which geographical position of buses is measured and recorded over time. This data can be used for the real time monitoring of traffic states. The proposed method in this paper will be introduced in the following sections. Models and results will also be discussed in the subsequent sections. 1 Automatic Vehicle Location Real-Time Link Travel Time Estimation: Using Buses as Traffic Probes Rouzbeh Forouzandeh 1 , Navid Khademi 2 , and Farshad Hakimpour 3 A Int'l Journal of Research in Chemical, Metallurgical and Civil Engg. (IJRCMCE) Vol. 4, Issue 1 (2017) ISSN 2349-1442 EISSN 2349-1450 https://doi.org/10.15242/IJRCMCE.DIR1116402 60