Beware of the limitations of using toll tags in order to calculate travel time on limited access roadways and arterials.
Experience from iFlorida Model Deployment
Throughout this process, lessons learned regarding the operation of a toll tag network for estimating travel times were documented. These lessons learned are summarized below.
- Beware of limitations of using toll tags to calculate travel times. FDOT identified and/or experienced a number of factors that limited the effectiveness of the toll tag network for producing travel times. Those considering using toll tag based travel time networks should consider these factors in their designs. Some of the factors FDOT identified and problems they experienced included:
- Limited market penetration for toll tag readers. Before deploying a toll tag network, FDOT conducted tests indicating that about 20 percent of vehicles on Orlando arterials had toll tag readers, indicating that toll tag readers would generate a large number of reads for the iFlorida arterials.
- Misaligned toll tag readers. After deployment, FDOT found that some toll tag readers did not produce as many reads as expected. FDOT discovered misaligned antennae and obstructions between the antennae and the roadway were often the root cause of this problem.
- Duplicate tag reads. On arterials, vehicles often passed under the toll tag readers at low speeds or were stopped under the reader. This often resulted in duplicate reads of the same tag.
- Vehicle diversions. FDOT found that many tags read at the start of a travel time segment did not match any tags read exiting the segment. This was likely caused by vehicles diverting onto other roads before exiting the segment. The fraction of vehicles diverting seemed to increase as the travel time segment length increased.
- Vehicle stops. An analysis of travel time observations indicated that the mean observed travel time was typically significantly higher than the median observed travel time. This was likely because some vehicles made stops between the time they entered and exited a travel time segment, introducing a high bias into the travel time observations. Methods were needed to filter out these high travel time outliers.
- Toll tag reader failures. FDOT experienced a large number of reader failures early in the deployment and struggled to maintain high availability of the readers throughout the project. At peak performance, about 90 percent of readers were operational, though the percent of operational readers could be much lower.
- Clock mis-synchronization. When first deployed, the internal clocks on many of the toll tag readers were not synchronized with a standard clock, which prevented use of the toll tag reader data for computing travel times.
- Transmission of archived tag reads. The FDOT toll tag reader system maintained an archive of toll tag data at each reader. If the transmission of the tag information to the toll tag server failed, the reader would later re-transmit all of the tag reads that had previously failed to transmit. At times, this transmission required so much network bandwidth that the transmission of real-time toll tag reads from other readers was delayed, which prevented the use of the real-time data to generate real-time travel time estimates.
- Adjust travel time estimation parameters for arterials. The algorithm that FDOT used to generate travel times from the toll tag reader data was originally developed for use on limited access highways. FDOT discovered that some of the parameter settings needed to be different to work well on arterials. For example, diverting traffic on arterials meant that the observed average travel time was typically higher than the actual travel time, which was not the case on limited access highways. The algorithm for excluding outliers is, therefore, more important on arterials than for limited access highways. The toll tag matching efficiency was much higher for limited access roads than for arterials and much lower for long arterials than for short ones. For short arterial segments (about 1 mile in length), about 50 percent of entering vehicles were later observed exiting the segment. For long arterial segments (about 5 miles in length), this percentage dropped to less than 20 percent.
- Include tag-reads from turning vehicles for more accurate travel time estimates. Including tag reads from turning vehicles can increase the number of travel time observations generated. The FDOT algorithms used data from a single reader to supply tag reads for both the entrance into and exit from a travel time segment. Depending on the exact placement of the readers relative to the intersection, more matches can be obtained by allowing data from multiple readers to be used. (For example, if a reader supplying tag reads at the entrance of a travel time segment is placed upstream from the starting intersection, then it will not record tag reads from turning vehicles that enter the segment. Including tag reads from readers monitoring the intersecting road can provide reads from these turning vehicles.) The evaluation team tested such an algorithm and found a significant increase in the number of travel time observations recorded.
- Monitor the maintenance of toll tag readers as soon as each has been installed and functioning. The iFlorida toll tag readers were deployed over a four month period, and FDOT did not begin actively monitoring the reader status until all the tag readers had been installed and deployment was complete. In the intervening period, many readers had failed, which created a large demand for reader repairs that FDOT was not able to fulfill.
Author: Robert Haas (SAC); Mark Carter (SAIC); Eric Perry (SAIC); Jeff Trombly (SAIC); Elisabeth Bedsole (SAIC): Rich Margiotta (Cambridge Systematics)
Published By: United States Department of Transportation Federal Highway Administration 1200 New Jersey Avenue, SE Washington, DC 20590
Source Date: 01/30/2009
EDL Number: 14480URL: http://ntl.bts.gov/lib/31000/31000/31051/14480.htm
Average User Rating
CCTV, closed circuit television cameras, road monitoring, sensors, vehicle detector, traffic detection, traffic monitoring, congestion monitoring