Algorithmic strategies for adapting to environmental changes in 802.11 location fingerprinting

Ubiquitous and accurate indoor positioning represents a key capability of an infrastructure that enables indoor location-based services. At the same time, such positioning has yet to be achieved. Much research uses commercial, off-the-shelf 802.11 (Wi-Fi) hardware for indoor positioning. In particular, the dominant fingerprinting technique uses a database (called a radio map) of manually collected Wi-Fi signal strengths and is able to achieve positioning accuracies that enable a wide range of location-based services. However, a major weakness of fingerprinting occurs when changes occur in the indoor environment that cause the signal propagation patterns and thus signal strength to change. Under such circumstances, a radio map collected at one time is unable to offer accurate positioning at all times. We propose a data-centric approach to achieving accurate positioning in changing environments. Unlike previous work, our approach does not require the deployment of special sensors that capture current signal strength phenomena, but rather lends itself towards ubiquitous indoor positioning. An empirical comparison of our proposals against conventional, static radio maps demonstrates very significant improvements in positioning accuracy in changing environments.