Indoor Navigation with a Smartphone Fusing Inertial and WiFi Data via Factor Graph Optimization

Mobile devices are getting more capable every year, allowing a variety of new applications, such like supporting pedestrian navigation in GPS-denied environments. In this paper we deal with the problem of combining in real-time dead reckoning data from the inertial sensors of a smartphone, and the WiFi signal fingerprints, which enable to detect the already visited places and therefore to correct the user’s trajectory. While both these techniques have been used before for indoor navigation with smartphones, the key contribution is the new method for including the localization constraints stemming from the highly uncertain WiFi fingerprints into a graphical problem representation (factor graph), which is then optimized in real-time on the smartphone. This method results in an Android-based personal navigation system that works robustly with only few locations of the WiFi access points known in advance, avoiding the need to survey WiFi signal in the whole area. The presented approach has been evaluated in public buildings, achieving localization accuracy which is sufficient for both pedestrian navigation and location-aware applications on a smartphone.