A long-duration study of user-trained 802.11 localization

Abstract

We present an indoor wireless localization system that is capable of room-level localization based solely on 802.11 network signal strengths and user-supplied training data. Our system naturally gathers dense data in places that users frequent while ignoring unvisited areas. By utilizing users, we create a comprehensive localization system that requires little off-line operation and no access to private locations to train. We have operated the system for over a year with more than 200 users working on a variety of laptops. To encourage use, we have implemented a live map that shows user locations in real-time, allowing for quick and easy friend-finding and lost-laptop recovery abilities. Through the system's life we have collected over 8,700 training points and performed over 1,000,000 localizations. We find that the system can localize to within 10 meters in 94% of cases. © 2009 Springer Berlin Heidelberg.