Accurate indoor localization based on crowd sensing

Indoor localization is an important primitive that can enable many ubiquitous computing applications. In this paper, we choose the iBeacon as landmark and improve the localization scheme of iBeacon and inertial navigation through crowd sensing. Specifically, we use crowd sensing to design a parameter learning algorithm for device diversity. Furthermore, we take advantage of crowd sensing to collect the correction information of iBeacon about inertial navigation, which can optimize the step length estimation and direction inference. We demonstrate for the first time a meter-level indoor localization system that is self-improving, user adaptive, and inclusive to diversity. Extensive experiments on users with mobile devices, with over 37 subjects walking over an aggregate distance of over 10 kilometers were carried out. Evaluation results show that the accuracy is within 2m in a 39m×21m testing area.