Mobile device positioning using learning and cooperation

We present a probabilistic graphical framework for mobile device positioning. We study the performance of a positioning algorithm, which implements the message-passing paradigm, in an indoor environment where a mobile device measures fingerprints of received signals. The key innovation in our approach is a stochastic parametric model for the fingerprint map that is adaptively tuned using on-line position estimates. The framework naturally extends to enable cooperative positioning in a network of mobile devices and we study the case of vehicle positioning as an illustration.