I. Introduction
Accurate indoor localization enables a wide range of important applications including indoor navigation, mobile social networking, augmented reality, etc. [1], [2] Recently, significant effort has been devoted to this important enabling technology [3]–[10]. In addition, indoor localization is a promising global market for industry [11], which is estimated to exceed $4 billion by 2019 [12]. However, even with major effort from the research community and strong industrial demand, indoor localization remains very challenging and state-of-the-art performance remains unsatisfactory [7].