DYNAMIC ORIGIN-DESTINATION MATRIX ESTIMATION USING GOOGLE MAPS TRAVEL TIME DATA CASE STUDY: BANDUNG CITY

• Main Author: Irfan, Muhammad
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/62387
• Institution: Institut Teknologi Bandung

Traditionally, Demand Origin-Destination Matrix obtained through road user surveys is very expensive and may face problems of sampling bias or data recording errors. The increasing use of Big Data in the transportation sector in recent years allows direct observation of traffic flows on various roads. One of the transportation travel time provider platforms is Google through the Google Maps API which can provide real time travel time data on various roads. With the right method, traffic flows on different roads and at different times will reflect transportation needs and allow for more effective and efficient management of urban traffic planning. Therefore, dynamic O-D Matrix estimation was carried out using travel time data from the Google Maps API by taking a case study in the city of Bandung. The analysis method in this study uses the first two stages of the fourstep model, namely the trip generation and trip distribution. Based on observations, the peak of traffic activity in the city of Bandung occurs at 08.00 WIB and 17.00 WIB for Monday and Friday, and at 15.00 WIB for Saturday and Sunday. Based on the estimation results of trip generation, the largest number of trips occurred on Monday at 17.00 WIB with 723,003 trips and the smallest number of trips occurred on Sunday at 15.00 WIB with 354,639 trips. The average number of trips from the four days is estimated at 484,658 trips. Meanwhile, based on the estimation results of the trip distribution, the movement pattern on weekdays tends to be more concentrated in the western part of Bandung City, while on weekends the relative intensity of the movement pattern is evenly distributed throughout the Bandung City area. From the results of statistical tests with the Root Mean Square Error (RMSE) and Geoffrey E. Havers (GEH) indicators, the type of trip distribution estimation modeling with the smallest error is the Unconstrained Gravity Model with the rank resistance function. The average RMSE value of the model is 370, then the average value of GEH is 2.40 so that the modeling results can be categorized as good because the value is less than 5.0.