MCMC-based indoor localization with a smart phone and sparse WiFi access points

We present a novel indoor localization approach using a pocket-placed smart phone and sparsely deployed WiFi access points (APs). Position information of a person carrying a smart phone can typically be obtained by two means: dead-reckoning(DR) based on local inertial sensors measurements and WiFi fingerprinting (FP). While a number of methods have been developed to fuse these two types of position information, due to extended training requirements in FP, their approaches suffer from limited practicality in most applications. To overcome some of the limitations of previous approaches, we propose to apply Markov Chain Monte Carlo (MCMC) sampling to fuse the local inertial measurements and the received signal strengths from nearby WiFi APs. MCMC technique can provide high accurate probability distribution inference. The performance of our proposed method is benchmarked using previous state-of-the-art through simulations as well as a real experiment in a typical office environment.