1 April 2004—There's something unusual about Randall Guensler's 1999 white Ford Explorer, but you wouldn't know it if you passed him in traffic. Hidden under the carpet beneath a rear seat is a metal box containing electronics and a Global Positioning System (GPS) unit wired to the vehicle's engine-diagnostic computer. The box tracks the SUV's position and speed and sends that information to a remote server over a cellphone network.

A professor of civil and environmental engineering at the Georgia Institute of Technology in Atlanta, Guensler wants to get rid of the congestion that increasingly clogs the city's metropolitan area. To understand better how such traffic knots form, he and a team of transportation engineers have used the GPS-enabled monitoring device to "bug" about 500 vehicles owned by drivers who, for the sake of science and fewer traffic jams, were willing to participate in the experiment. Whenever a driver is behind the wheel, the monitor gathers data, which is transmitted to a central server at Georgia Tech during off-peak hours.

This monitoring campaign, which will go on until August, represents a major advance in traffic analysis. It provides more data at less cost than any previous scheme, and it gathers that data from all points in the road network instead of just a few fixed spots. When completed, it will be the most comprehensive data set about traffic patterns and driving behavior in the United States—a gold mine for transportation engineers. "There's never been a data set this rich," Guensler gushes. "It's amazing all the different things that we'll be able to see."

Next month, Guensler will meet with officers of the Georgia Department of Transportation, in Atlanta, a cosponsor of the project with the U.S. Federal Highway Administration, to discuss how the agency can use the data to improve traffic management, transportation modeling and simulation, and air quality assessment, among other activities. "I am very impressed by what the project has done up to now," says Rick Deaver, chief of R and D at the Georgia Department of Transportation's Office of Materials and Research, in Forest Park.

Current traffic monitoring systems are expensive and require constant maintenance. For example, wire loops buried in the asphalt that detect vehicles as they pass, although one of the main technologies used to track traffic, fail constantly and require expensive machinery to cut into the pavement. Cameras connected to computers that analyze images and determine the volume of traffic are also expensive and still lack reliability. As a result, traffic monitoring on arterials—heavily traveled roadways that are not highways—is often meager. Local agencies monitoring arterials usually don't have enough money to deploy extensive monitoring systems.

"Right now it's very difficult to gather data on arterials," says Steve R. Gordon, a researcher at Oak Ridge National Laboratory's Center for Transportation Analysis, in Knoxville, Tenn. He says that getting information directly from cars as is being done in the Georgia Tech project would be an effective way to monitor these major roadways. "It's really important if it works," Gordon says. For monitoring arterials, "being able to track vehicles without having to deploy infrastructure would be a terrific boost."

Traffic management agencies would then be able to better control traffic by changing signal timing, rapidly diverting traffic from congested routes, and providing travel advisories using electronic signs, he says.

In fact, Guensler points out that by simply improving signal timing, traffic flows better and traffic jams can be prevented in certain areas. "It's one of the most cost-effective ways to reduce congestion, minimize emission, and reduce fuel consumption," he says.

At a cost of around US $700–$800, the Georgia Tech-designed monitoring device is basically a simple computer equipped with a GPS unit and connected to the car's speedometer or, in the case of models made in1996 or later, to the onboard engine-diagnostics computer. When hooked to the onboard computer, the device picks up extra data, such as engine speed, throttle position, manifold pressure, and other information, that indicate how a person is driving and how much pollution that vehicle is producing.

So far, the team has collected statistics on about 400 000 vehicle trips and hopes to have data on close to one million trips when the experiment is over. That means one million sets of second-by-second position, speed, acceleration, and other data from trips all over the Atlanta region. Researchers will be able to see what kind of households make more trips, how many hours a person spends behind the wheel on average, how long the same trip takes during different times of the day, and even how often participants drive above speed limits.

To select participants, researchers first defined several household types—based on characteristics like income, household size, and vehicle ownership. The goal was to recruit the right types of participants, in the right proportion, geographically distributed all over the area so that they were representative of the actual household distribution in the Atlanta region. Working with this microcosm of Atlanta drivers, researchers should be able to find statistically significant results about the whole population.

Another important aspect of the project was to protect participants' privacy. Researchers conduct the experiment much like a medical study. All data collected is encrypted and stored in secure systems, and researchers work with data sets that, while containing information on detailed vehicle activity, cannot be traced back to an individual driver or address.

A major challenge for a traffic monitoring system that relies on data transmitted by individual vehicles is guaranteeing reliable communication. Transmission in real time using a cellular network might prove precarious, especially for a large number of vehicles. If hundreds of cars try to transmit data simultaneously, they could cause a network traffic jam. To find a solution to this and other potential problems, Guensler teamed up with computer science professor Richard Fujimoto, also at Georgia Tech. Fujimoto plans to solve the problem by simulating different network scenarios. "We're combining traffic simulations with wireless network simulations," he says.

One alternative to having every car transmit its data directly to a cellphone base station is using Wi-Fi wireless technology to transmit data packets from cars to other monitor-equipped vehicles nearby. Data packets would hop from car to car until they reach a vehicle close to a cellphone tower or to a receiver installed by the roadside. Vehicles would then form an ad hoc mobile network, consisting of moving nodes that are constantly joining and leaving the network, Fujimoto says. His simulation study, backed by the U.S. National Science Foundation, in Washington, D.C., is trying to answer such questions as: how many cars are needed to form a usable network? And how reliable would the network be?

How to deploy such ad hoc networks is one of the main challenges in computer networking today. Communication in a wireless car network might be subject to somewhat unexpected factors, such as a truck that suddenly blocks a transmission between two vehicles. Nevertheless, if simulations show that such a system could work, a fleet of "bugged" vehicles may one day provide traffic management agencies with information that will ultimately change the way they control the flow of vehicles through and around cities. Whether that's the way out of congestion is another question.