RALEIGH, N.C. — Everyone knows the feeling: You enter a hospital to visit a sick friend. The receptionist points and tells you the number, but soon you’re lost in a maze of identical white hallways, the room numbers seeming to increase and decrease randomly. You are lost.
Romit Roy Choudhury, a computer engineering professor at Duke’s Pratt School of Engineering, has developed a system that could prevent this feeling. Called UnLoc, it makes use of existing technology in your smartphone to do indoors what GPS can do outdoors.
GPS, short for “global positioning system,” works by beaming a signal from your phone to a satellite in space. The satellite uses the signal to pinpoint your position, then sends your location to your phone using another signal.
But GPS doesn’t work in buildings. The signals aren’t strong enough to get through roofs. Plus, GPS is accurate only to about 15 feet, and that precision won’t cut it indoors. And in a multi-story building, GPS wouldn’t know what floor you’re on.
He Wang, a graduate student in Choudhury’s lab, has been working on UnLoc as part of his Ph.D. He says the technology, unveiled last month at the Mobisys Conference in the U.K, takes advantage of existing stuff already in your smartphone.
Your smartphone can tell you where you are going, even without GPS. It comes equipped with motion sensors such as a compass to show direction, and an accelerometer that records spikes with each step you take.
“The phone can count steps to get an estimate of distance traveled,” Wang said. But over time, this estimate becomes slightly inaccurate and must be corrected. Knowing the distance to a fixed landmark would provide the needed course correction.
“Just like when your friend gives you directions to her house: ‘You know you’re almost there if you pass a big store on your left,’” Wang said.
UnLoc uses two kinds of landmarks.
There are “seed landmarks,” which are physical features with obvious signals, such as stairways, elevators and corners. The accelerometer in your phone can recognize when you’re on an elevator; there will be two shocks, one at the start and one at the stop, the same ones we feel in our bones.
The second sort of landmarks are called “organic,” invisible signatures we can’t see or feel but are very real. Your phone, in addition to the motion sensors, has a Wi-Fi antenna and a magnetometer. If there are lots of electronics in a room you walk by, your phone’s magnetometer can pick up the room’s magnetic flavor.
Or there are areas where certain combinations of Wi-Fi networks are seen, like an Internet fingerprint for that spot.
Wi-Fi fingerprinting of locations is not new. It’s usually called wardriving. People record the unique Wi-Fi features of an area, and map in detail where the pattern occurs, creating a database.
But the UnLoc team’s technology differs from wardriving, because the system learns landmarks with each user who passes nearby. Eventually the location of the given landmark becomes more accurate, using the data collected from the crowd. The system learns.
In turn, once the system knows the location of a landmark, it can be used to guide people.
Since the system can learn, it requires no reprogramming, if, for instance, a wall is knocked down. It would learn on its own. It can also identify temporary structures, such as a long line for the elevator every day around 5 p.m.
The team has tested the system in a mall and two Duke engineering buildings, showing accuracy to 5 feet, which far outperforms GPS.
It also eats less battery than GPS, a notorious hog that requires bouncing signals from the Earth to a satellite.
Users concerned about privacy may opt to simply load the current building map on their device, and not send their data back to teach the system.
There is no set date for when UnLoc will be available for consumers. But the competition is fierce, with behemoths such as Google, Microsoft and Apple also pursuing indoor location technologies, so the UnLoc team expects fast development. The Duke researchers developed UnLoc in collaboration with a team from the Egypt-Japan University of Science and Technology.
In addition to indoor navigation, the technology also breeds possibilities for entirely new applications.
“If you walk by a store in the mall, it can send you a coupon,” Wang said. “Or even if you walk by a specific product in the grocery aisle.”