AutoLoc: Toward Ubiquitous AoA-Based Indoor Localization Using Commodity WiFi

Abstract

The past decade's research in RF-based indoor localization has shown the possibility of decimeter-level accuracy based on the Angle of Arrival (AoA) of WiFi signals. However, existing systems require either recurring manual calibrations or the assistance of extra sensors to calibrate uncertain initial phases for accurate AoA estimation, hindering their large-scale deployment. In this paper, we propose AutoLoc, a ubiquitous localization system that overcomes the above limitation and can be seamlessly applied to the widely deployed WiFi Access Points. The key insight underlining AutoLoc is that the uncertain initial phase is constant once the RF oscillators are frequency locked. To this end, we propose three main innovations. First, we eliminate the uncertain initial phase while preserving the geometric information by combining Channel State Information (CSI) from different locations. Second, we introduce a cooperative confidence-aware localization algorithm that accurately localizes the target by quantifying the confidence of all possible AoA candidates. Third, we design a two-phase localization framework to fortify AutoLoc in multipath-rich scenarios. We implement AutoLoc on Commercial Off-The-Shelf WiFi devices. The experiment results in large-scale environments reveal that AutoLoc can achieve comparable median accuracy and 54 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> better 90 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> percentile accuracy compared with the state-of-the-art localization approach with manual calibration.