A Probabilistic Clustering-Based Indoor Location Determination System

We present an indoor location determination system based on signalstrength probability distributions for tackling the noisy wireless
channel and clustering to reduce computation requirements. We provide
two implementation techniques, namely, Joint Clustering and Incremental
Triangulation and describe their tradeoffs in terms of location
determination accuracy and computation requirement. Both techniques have
been incorporated in two implemented context-aware systems: User
Positioning System and the Rover System, both running on Compaq iPAQ
Pocket PC's with Familiar distribution of Linux for PDA's. The results
obtained show that both techniques give the user location with over 90%
accuracy to within 7 feet with very low computation requirements, hence
enabling a set of context-aware applications.
Also UMIACS-TR-2002-30