ADVERTISEMENT

Mixed Signals: Smart Phone Sensors Recruited to Deliver Indoor GPS

Duke University researchers are developing a mobile app that uses wi-fi antennas, cellular radios and other detectors to guide smart phone users
gps,navigation,smartphone



Courtesy of Duke University, via YouTube

Global positioning system (GPS) devices may not always provide spot-on directions, but they do provide drivers, cyclists and hikers with convenient access to digital map data of every square meter of the planet shadowed by satellites. Step indoors and you will find that same GPS receiver becomes an expensive paperweight.

Indoor GPS has been in the works for at least a decade, but the plethora of interfering signals from wi-fi, ultrasound, cellular and other devices make it difficult for GPS units to come up with an accurate reading. Whereas a GPS discrepancy of a few meters while navigating city streets makes little difference to a driver, that same margin of error inside a big-box electronics store or hospital is likely to send users down the wrong aisle or hallway.

Instead of focusing on a single type of signal to map indoor areas, researchers at Duke University and Egypt-Japan University of Science and Technology are developing software to help smartphone users find their way by gathering information from a number of different signal types. Their UnLoc system—short for unsupervised indoor localization—gathers signal data using wi-fi antennas, cellular radios, compasses, gyroscopes and accelerometers.

UnLoc tags each of these signals as a virtual landmark. One example would be an elevator whose distinct pattern of movement can be detected by a smartphone's accelerometer, according to the researchers, who presented their UnLoc system on June 27 at the 10th International Conference on Mobile Systems, Applications, and Services (MobiSys 2012) in Low Wood Bay, England (pdf). Similarly, a particular corridor might be configured with a unique set of wi-fi access points that the smartphone can read. UnLoc envisions these kinds of signatures as internal landmarks within a building, according to the researchers, who are led by Romit Roy Choudhury, an associate professor of computer engineering at Duke. Landmark information could then be stored in individual phones or shared as part of a larger database that maps indoor environments in more detail.

The researchers have demonstrated how UnLoc works in a YouTube video. In the video, Alex Mariakakis, a Duke undergraduate student who participated in the research, navigates an on-campus building using a Samsung Nexus S Android phone. As Mariakakis walks the building's corridors, UnLoc superimposes a dot marking his progress on a map of the building's floor plan, which is stored in the phone. The researchers say that UnLoc, which can track a user's location even without a preinstalled floor plan, typically located between 10 and 20 landmarks per floor in the buildings where they tested and was accurate to within about 1.7 meters.

As mobile devices have become more powerful, a number of academics and technology companies are developing indoor localization capabilities. For instance, Google, Microsoft and Stanford University are participating in the WiFiSLAM project, which focuses specifically on using wi-fi signals—as opposed to the variety of signals UnLoc purports to detect—to pinpoint a smartphone user's location to an accuracy of within 2.5 meters.

Indoor localization is expected to be particularly useful in health care settings. Norway-based Sonitor Technologies sells an ultrasound-based indoor positioning system (IPS). Patients in a health care facility wear wristbands that emit ultrasonic waves, and microphones placed throughout the facility pick up the high-frequency sound. Since walls and doors effectively confine the signals to a room, Sonitor's IPS avoids signal confusion with any radio-frequency identification tags or wi-fi hot spots used in the same building.

Rights & Permissions
Share this Article:

Comments

You must sign in or register as a ScientificAmerican.com member to submit a comment.
Scientific American Holiday Sale

Black Friday/Cyber Monday Blow-Out Sale

Enter code:
HOLIDAY 2014
at checkout

Get 20% off now! >

X

Email this Article

X