Home // International Journal On Advances in Systems and Measurements, volume 8, numbers 3 and 4, 2015 // View article

Multi-Scheme Smartphone Localization with Auto-Adaptive Dead Reckoning

Authors:
Michael Jäger
Sebastian Süß
Nils Becker

Keywords: Indoor Positioning; Pedestrian Activity Classification; Dead Reckoning; Wi-Fi Fingerprinting

Abstract:
Most indoor localization approaches for mobile devices depend on some building infrastructure to provide sufficient accuracy. A commonly used method is the fusion of absolute position measurements with relative motion information from sensor units. This paper examines the requirements for smartphone localization in areas consisting of several buildings and open space, where a single positioning method might deliver good results at one location but might also fail at another. It is shown that, for several disparate reasons, a localization system combining alternative positioning techniques and going beyond the scope of a single hybrid method, is desirable. The paper proposes such a multi-scheme system with a three-layer architecture consisting of base methods, hybrid methods, and scheme selection. Automatic selection of an appropriate scheme is described for heterogeneous infrastructure within multi-story buildings and for indoor-outdoor transitions. Support of several distinct hybrid methods can be based on the same generic fusion algorithm. The paper proposes a novel lightweight fusion algorithm, called "auto-adaptive dead reckoning". It can be used in indoor and outdoor environments to combine an absolute localization method, e.g., Wi-Fi-based signal strength fingerprinting, in an adaptive way with inertial pedestrian navigation. Based on an accuracy factor reflecting the current context conditions of a location measurement the influence of each of the involved estimates is weighted accordingly. In a case study using Wi-Fi fingerprinting, accuracy has been improved by 43% in an indoor environment. Hence, more genericity can be obtained without loss of accuracy.

Pages: 255 to 267

Copyright: Copyright (c) to authors, 2015. Used with permission.

Publication date: December 30, 2015

Published in: journal

ISSN: 1942-261x