A new Wi-Fi/ GPS fusion method for robust positioning in urban environments

Abstract

This paper presents a tracking framework for enhancing the positioning accuracy of a mobile device by fusing the positions provided by a GPS navigation system and those obtained using Wi-Fi signal strength measurements, in urban environments. To achieve an efficient fusion, a structure based on two particle filters and a Multiple Model (MM) approach is proposed. It fuses the information coming from these two independent technologies, to overcome their own drawbacks. Indeed, the Wi-Fi and GPS are viewed as two models, whose probabilities are calculated using a Transition Probability Matrix (TPM) and a Mixing Likelihood Function (MLF). These probabilities are then used to combine the mobile state estimates, provided by the two particle filters. Matched to the two models, these filters interact by exchanging a part of their particles. The proposed architecture is experimentally evaluated and compared with the pure Wi-Fi and GPS positioning systems and other fusion methods. The results indicate that the positioning errors of the proposed scheme are the lowest.