Despite big advances in technologies like GPS and RFID, finding your way through crowded shopping malls and train stations isn’t really any easier than it was five years ago. Companies like Google and Broadcom are working on the problem, but there still isn’t a universal solution that provides the kind of accuracy needed for indoor localization to really be useful. Well, Duke University researcher Romit Roy Choudhury is working on an application called UnLoc (for "unsupervized localization") that uses recursion, filtering, and "invisible landmarks" to work out your indoor location down to 1.6 meters (about 63 inches) — and the accuracy is improving.

Invisible landmarks are things like 3G and Wi-Fi dead zones, and motion signatures from elevators or stairwells, and UnLoc uses them much in much the same way humans do — as points of reference. Your current position is estimated using a filtering algorithm that figures out where you "should" be based on readings from your phone’s sensors, and then updates its estimate as you run into new landmarks. As He Wang, the project’s lead Ph.D. student points out, "the best part of the application is that it is recursive, which means that it starts with zero knowledge but ‘learns’ over time."