Crunching data gathered by devices called loop detectors, already buried at intersections and on most urban highways, computer algorithms written by Benjamin Coifman, a professor of civil and environmental engineering at Ohio State University, were able to determine that a traffic jam was forming within three and half minutes after vehicles began to slow.
During tests, the software helped California road crews recognize the onset of a traffic jam three times faster than without the application, according to Coifman. The extra time helped the crews clear accidents and restore traffic flow before other drivers could be delayed.
The software will be of particular interest to the technology industry because the cities with the greatest vehicle congestion also happen to be home to the bulk of tech companies. Traffic jams cost the average city almost $1 billion in lost work time and wasted fuel per year, according to the 2002 Urban Mobility study prepared by the Texas Transportation Institute. The study measured traffic in 75 major U.S. cities.
Silicon Valley definitely feels the crunch, with the San Francisco-Oakland area ranked No. 2 for most hours wasted annually per person during peak commute time, with 92 hours, and San Jose, Calif., appearing at No. 7, with 74 hours. The Seattle-Washington area, home of software powerhouse Microsoft, was ranked No. 3, with 84 hours spent in traffic annually per person. The Los Angeles area ranked No. 1, with 136 hours logged per driver.
"Traffic was one of the main reasons I wound up moving to within 10 miles of work," said Allen Bush of handheld maker Handspring. Bush has worked out four alternative routes he can use to try to beat the traffic to Handspring's Mountain View, Calif., office. "I would certainly find something like this system valuable."
Coifman began monitoring a three-mile stretch from Berkeley to Emeryville along Interstate 80 when he was a postdoctoral researcher at the University of California at Berkeley. He installed computer network hardware in control boxes along the highway and took in data from loop detectors every third of a mile.
He wrote algorithms that can capture a vehicle's length as it passes over a detector and use the data to identify the vehicle. When the same vehicle passes over the next loop, the computer calculates how fast the car is moving. The algorithms use the information to determine whether a traffic jam is forming.
The algorithms also take into account drivers' unpredictable behavior, such as changing lanes and exiting or entering from ramps. And the algorithms also consider rubbernecking--the delay caused by people slowing down to look at accidents along the road.
Coifman said detector systems that are currently in use wait for data from several detectors but that his algorithms can work from less information, speeding the process.
This, he said, is crucial, because the personal and financial costs grow more quickly the longer more people are stuck in traffic.
Coifman said his method needs only his software, a personal computer, and the loop detectors. He has been using Cellular Digital Packet Data modems--an old wireless networking technology--to transmit the information.
"People don't realize all the things that are going on behind the scenes to keep traffic flowing," said Coifman. "As (congestion) gets worse, this is going to be one more tool to help traffic engineers keep at the same level, at the very least."
Coifman's work was supported by the University of California, the California Department of Transportation, and the United States Department of Transportation. The findings of the work will appear in the March issue of the journal Transportation Research.
Coifman said he's discussing implementation of the setup with various states' transportation departments.