Street wise

As rush hour has expanded far beyond 60 minutes, drivers' frustration has grown. And grown. Alternative-route seekers play the game, trying to shave off a minute here, a minute there, betting on the roads less taken.

But a bunch of MIT technology gurus would say, more often than not: Hit the highway, not the byway. And they've got the data to back it up.

Samuel Madden, Hari Balakrishnan, and their team of researchers have been collecting information about roads and their conditions for several years through a wireless sensor network deployed in a Cambridge livery company's fleet.

"We can understand the world through data collected by these sensors," Madden said. "We were interested in gathering data over a large geographic area, but the sensors are expensive, so you can't place them everywhere.

"We asked ourselves, 'What's the obvious thing that moves around?' " he said.

Answer: cars.

They've developed software, called CarTel, to test theoretical approaches to gathering and managing the sensors' data. In the process, they've discovered practical uses for the sensors: tracking speed and traffic at particular times of day along most of Greater Boston's highways, as well as road surface conditions and cars' fuel economy and performance.

Although it will take years for the technology to trickle down to consumers, Balakrishnan already has tested a real-world application for the software.

He had convinced himself that taking Interstate 93 north from the Massachusetts Institute of Technology to Winchester at rush hour was the slowest way home, but he didn't have numbers to back his belief. So he installed in his glove compartment a sensor kit built by the group to record his average time spent commuting over several weeks on each route he took.

The kit contained a device that connected to the auto's onboard diagnostic computer to record speed and mileage, as well as a wireless card that could transmit the data to a designated Internet address for later analysis.

The data proved him wrong. The interstate was faster by as much as 10 minutes for his 9-mile commute. Balakrishnan started taking the highway.

The general principle has since been supported by data collected by the CarTel system, which shows that unless there's an accident, it's usually faster to use a busy highway rather than side streets.

Balakrishnan, Madden, and their team originally started the mobile sensor network first to see if it could be done, and second to create a platform that could be useful to researchers gathering data about anything from air quality to noise pollution. But the group has stumbled across so many practical uses for the sensors, they even created an offshoot project called Pothole Patrol, or P{+2}.

For this project, CarTel paired with PlanetTran, a livery service that uses hybrid cars. CarTel placed sensor kits in each of the Cambridge-based fleet's 25 vehicles.

The kits, which cost about $450 each, contain global positioning devices and accelerometers, which measure movement in three directions: up/down, left/right, and front/back. Every time a car travels over a road anomaly, the accelerometer records the depth and width of the hole, sending the data to CarTel's website with geographic coordinates when it passes a wireless access point at MIT or elsewhere.

Back in the office, Madden analyzes the data and finds some of Boston's most uneven roads. Among them: Route 99 from Charlestown to Everett and the Callahan and Sumner tunnels to and from Logan Airport and Eastie.

Madden said the program has made him more aware of potholes as he bikes from his South End home to MIT, where he and Balakrishnan are professors in the department of computer science and electrical engineering.

"Every time I ride my bike somewhere, I pay attention" to potholes, he said. "But all drivers, especially in Boston, spend a lot of time being aggravated by the roads."

Madden hopes his research might help assuage some of that frustration in the future.

Early on in the project, he encouraged a few of the team's researchers to test out Pothole Patrol on a particularly bad patch of road near MIT. "We were trying to get some grad students to drive over this sinking storm drain," he said, "but they wouldn't do it, and now we're glad. It developed into a dramatic 6-foot-wide sinkhole."

He envisions that such information will eventually be useful for cities and states to maintain roads and establish priorities for repair. Municipalities could put sensors in selected vehicles to gather data over time and spot problems before they reach the sinkhole stage.

While PlanetTran helps the MIT researchers gather data, the MIT sensors have helped PlanetTran's owner, Seth Riney, manage his fleet with something like Harry Potter's Marauder's Map.

Riney can pull up a map on the Internet and see where each of the labeled cars in his fleet is at any moment. During an interview he watched as one moved along the Mass. Turnpike toward Logan.

"This helps us to operate efficiently," he said of the technology. He can assign prescheduled pickups to the driver who will get there fastest, which means each car can make more trips per day and each driver, more tips.

Using personal digital assistants, drivers heading to the airport can also retrieve such information as flight numbers, arrival times, and terminals, Riney said.

"A happy side effect of all this," he added, "is that passengers get WiFi in the cars, too."

Madden describes the partnership with PlanetTran as synergistic. But getting the wireless connection inside the vehicles was one of Madden's first challenges. The researchers discovered they could do this two ways: by using a cellular modem, such as in a mobile phone, or a wireless card that plugs into the sensor kit.

At first, he wasn't sure if it would be possible to use the WiFi networks scattered all over town, at such places as universities, businesses, and residences. But he and his team found the cars could transmit about 100 kilobytes of data, the equivalent of one small digital photograph, even while traveling at 40 miles per hour past one of these networks. The data then gets sent to a website that a user -- in this case, just the CarTel group -- can view.

Madden's team goes to great lengths not to usurp others' wireless service, by programming each car's minicomputer so it communicates only with a specific set of WiFi connections. "We try to be careful not to connect to every access point out there," he says. "We're wary of stealing bandwidth from a customer."

Once Madden figured out that the car sensors could transmit data, he began wondering how they could also receive data and for what purpose. He sees this as his team's next direction.

"Auto manufacturers are working on this," he said. And traffic advisory services such as SmarTraveler already do this -- by sending traffic alerts to cell phones. But he sees applications beyond gridlock warnings.

"You could have your e-mail read to you," he said, "or on a screen, be able to see the position of every car around you, even if you're driving behind a semi."

PlanetTran's Riney, who has a background in technology, said he believes intelligent vehicles are the wave of the future. "Having the car be self-aware, knowing the position of other cars around them and be Internet-enabled to communicate with each other," he said, "changes everything."

He has not accessed any data collected from the sensors to send traffic reports or analyze vehicle performance, but he hopes to do so in the future.

One student working on CarTel is already researching yet another application for the mobile sensor data: reliable route planning.

"Everybody who drives avoids particular roads at particular times of day," said Madden. But plotting out the best shortcut using a mapping service on the Internet isn't easy, because those services generally provide the shortest distance. Period.

Because Madden's system can track patterns over a long period of time, it can be used to find the most reliable route at any time of day. "We can ask more interesting questions," he said. "Not just what's the fastest route, but what's the fastest and most consistent route? Which roads will get me to my destination within 15 minutes 90 percent of the time?"

Those answers, he said, might be more valuable.