Multi-Modal Indoor Positioning of Mobile Devices

In this paper we extend our previous results on WiFi and image localization to include magnetic sensing for multimodal indoor localization. A two-step process is proposed that performs an initial localization estimate, followed by particle filter based tracking. Initial localization is performed using WiFi and image observations. For tracking we fuse information from WiFi, magnetic, and inertial sensors. We demonstrate the feasibility of this system using fingerprint maps that are collected with a single walkthrough of the building at normal walking pace. Further we reduce our database generation method from previous works to require only a smartphone and a foot mounted inertial measurement unit (IMU). Only a smartphone is needed for positioning after database generation. We present results for two locations: the Stoneridge Mall in Pleasanton, California, and the Doe Library at the UC Berkeley campus. We achieve an average location error of 2.6m across both locations.