Tufts engineers want to use flying robots to stave off infrastructure disasters

Usman Khan and Babak Moaveni
Usman Khan, left, and Babak Moaveni examine a prototype flying robot that could help download data for bridge inspections. Photo: Alonso Nichols

At first, it was just another rush hour in downtown Minneapolis. Thousands of drivers inched along in the August heat, sighing as traffic reached a standstill on the I-35 westbound bridge, a major artery spanning the Mississippi River.

Suddenly, with a sickening rumble, the 40-year-old steel bridge gave way, plunging more than 10 stories into the water below. In a moment, the span had transformed from a solid structure into a tangled mess of steel, concrete and cars, leaving 13 dead and scores more critically injured.

According to the American Society of Civil Engineers, this sort of gross structural failure, which took place in 2007, could happen again. Every four years, the group issues a report card for American infrastructure—and on the latest one, in 2013, bridges around the country scored an unimpressive C+.

Of 607,000 bridges in the country, more than 151,000—roughly 25 percent—are now considered obsolete or “structurally deficient,” meaning that they need urgent maintenance or replacement, and must be monitored on a regular basis to ensure their safety.

That monitoring is done visually by teams of engineers as they dangle beneath the bridge on ropes or reach up from a cherry picker. It’s slow, dangerous work, and due to time and funding restraints, it often takes place only once every other year.

Babak Moaveni, an assistant professor at the School of Engineering who studies structural health monitoring, thinks that the current monitoring approach needs to be improved. Even experienced crews can miss subtle changes in a structure if they conduct these infrequent inspections visually, he says.

Virtual Inspector

That’s why he and his colleague, assistant professor Usman Khan, are proposing a new approach to track the health of a bridge around the clock. In his lab at Tufts, Moaveni and his students are developing a sort of “electronic inspector”—a system of autonomous, computerized sensors designed to be attached permanently to beams and joints in a bridge. They would be able to record and analyze vibrations in the structure instantaneously, at all hours, and alert authorities if anything seems out of the ordinary.

Moaveni hopes that these “smart” sensors would pick up tiny but significant changes in a structure that would otherwise be missed. Once installed, they would also be cheaper to operate than teams of inspectors, and would allow authorities to home in on specific parts of a bridge that need closer attention.

Right now, he’s in the early stages of testing the idea with help from a team of graduate and undergraduate students. They have installed a network of prototype sensors on a footbridge leading to Dowling Hall on the Medford/Somerville campus, and in the course of their research, have tested the bridge’s integrity against nearly 5,000 pounds of concrete weights added to its surface—a load well within the bridge’s limits but great enough to simulate structural damage. The team has been monitoring vibration levels using an on-site computer.

The point of this research, Moaveni says, is to develop mathematical algorithms that can automatically tell the difference between normal vibrations and “abnormal” ones. “Right now, if a bridge has severe damage, we’re pretty confident we can detect that accurately. If there is only a small amount of damage, though, it’s less obvious,” he says. “The challenge is building the system so it picks up small anomalies without crying wolf.”

Going Wireless

Moaveni is testing the system using a web of cables that link the sensors beneath the Dowling Hall bridge. To make it truly cost-effective, though, he’ll ultimately need to go wireless. Cables, after all, are expensive to run through an existing bridge, and are prone to failure in extreme weather.

The trouble is, the sensors in a wireless system would have to rely on batteries for power, so they’d need to be built to last as long as possible on a single charge. Moaveni thinks it would be straightforward to create a sensor that turns on only for a few minutes each hour, or once every other day, using power in tiny sips instead of gulps while gathering a wealth of useful data. Getting that data back off the sensor, however, would remain a huge energy drain, he says.

Compared to the tiny trickle of electricity needed to keep a sensor running, transmitting information over wi-fi or a cellular network eats up huge amounts of power, just like it does on your cell phone. “Your cell phone battery usually lasts a day or two before you need to recharge it. We’d need to create sensors that last up to five years,” says Usman Khan, an assistant professor of electrical and computer engineering at Tufts, who is working with Moaveni on the project.

Checking on the Fly

Together, the two researchers are exploring a decidedly futuristic solution. They envision using drones—tiny autonomous flying robots—that could hover around the bridge, take photographs of its structures, collect data from vibration sensors and ferry all that information back to a central “base station” to unload it. It sounds like something out of Blade Runner, but the National Science Foundation is betting it’ll work. Earlier this year, the organization awarded Khan its prestigious CAREER grant, a $400,000 research grant that recognizes the work of young faculty with promising ideas (Moaveni received the award himself in 2013).

In theory, using flying robots to visually inspect a bridge and shuttle data would be an elegant way to extend the range of Moaveni’s sensor network. In practice, however, Khan says there are plenty of problems to overcome.

First off, there’s the issue of navigation. In the open air, the drones could easily use GPS to know their location, but while flying under a metal bridge or inside a tunnel, that system isn’t reliable. The robots would have to fly blind. “It’s a huge challenge to get a robot where it needs to go without a reliable reference point. How do you find your way when the North Star is missing?” Khan asks.

There’s also the issue of communication between drones. On larger structures, like the Tobin Bridge in Boston, it could take days or weeks—an impractical amount of time—for a single tiny drone to visit all the sensor points installed, so Khan and Moaveni want to use multiple drones working in tandem. In order to do that, however, the machines would need to talk to each other somehow, avoid each other in mid-air, and divvy up sensors to target so they’re not covering the same ground.

“When you have that many drones involved, it poses a really complex mathematical problem,” Khan says. “Right now, we’re working on developing algorithms in the lab, so we’re hoping to test this with multiple robots in the near future.”

Although the pair envision future bridges surrounded by swarms of drones, Moaveni says the system might be used effectively with a single drone in the meantime. And despite his focus on such machines, he insists that structural engineering—and, in this case, maintenance—will always remain a uniquely human pursuit. His robots, no matter how complex, will simply be tools.

“Even a system like this won’t be able to stop a bridge from collapsing,” says Moaveni. “It’ll only be able to tell if changes in the structure are taking place. The real judgment call will still be made by engineers.”