Potential Use of GPS Data for Calibrating Travel Demand Models

The ability of travel demand models to accurately represent travel behavior is largely dependent on the availability of reliable and up to date data. Most transportation agencies have access to traditional network data (i.e., geometry, traffic control, traffic counts, etc). However, there are areas of serious data deficiencies, including: trip-making behavior and network performance data to support model calibration. In fact, recent travel survey data are rarely available, especially for the smaller Metropolitan Planning Organizations (MPOs). As for traffic data, they are often developed in absence of careful consideration of their potential use in the travel demand model calibration. There have been several efforts to automate some data collection functions for supporting planning models. These efforts generally rely on Global Positioning System (GPS)-based travel data collection devices, replacing the traditional paper-based trip diaries. These devices provide a relatively easy user interface while increasing data accuracy (spatial and temporal trip characteristics). An added benefit of these systems is their ability to generate additional data such as speed, travel time, and delay about network routes. This information is very valuable in supporting travel demand model calibration and measuring the performance of the system. This presentation reports on a pilot study to establish route-specific travel time performance measures using one GPS unit. The main focus of the study was not the travel behavior characteristics of the driver, but rather to measure the performance of key routes under seasonal, time-of-day, and day-of-week variations. Four volunteer drivers in the Fargo, North Dakota and Moorhead, Minnesota area who each took turns using the GPS to track their daily trip to and from work were used for the pilot. The data were downloaded weekly and transferred into a Geographic Information System (GIS) for easy viewing and data management. Route information such as travel time, speed, and congestion variations were clearly identified by the unit. Sample data and analysis are shown, demonstrating the value of this method to enhance travel demand model calibration, even with a small number of participants.