Multiuser Smartphone-Based Localization by Composite Fingerprint Descriptors in Large Indoor Environments

The accuracy of existing Wi-Fi fingerprint-based indoor localization methods with smartphones is generally limited by the perceptual aliasing of Wi-Fi signals and device heterogeneity. In this article, we present a high-accuracy composite fingerprint-based localization method for these problems by combining Wi-Fi, space, and pedestrian dead reckoning (PDR) information in a multiuser scenario. Unlike common Wi-Fi fingerprints, we adopt fast appearance-based mapping (FAB-MAP) schemes for Wi-Fi and space information to improve fingerprint-matching accuracy with negligible computation. To achieve this, we propose a binarization procedure for composite fingerprints combining Wi-Fi mean differential fingerprint (MDF) and turning landmarks. The turning landmarks are marked on a floorplan in the offline stage and subsequently identified using inertial measurement unit (IMU) sensor data in the online stage. We then use the Bayesian-based FAB-MAP to obtain the matched location. Finally, a global pose graph is constructed to estimate the users’ trajectories given the composite fingerprint, PDR, and loop closing constraints. The experiments on real and public datasets demonstrate that our method achieves an average localization error of 1.75 and 1.77 m, respectively, which surmounts the classic Wi-Fi fingerprint-based methods, especially in multiuser scenarios.