Calibration-Free Localization Based on Fingerprinting of Channel State Information

Indoor localization is playing a critical role in smart Internet of Things, which require high accuracy with effortless deployment and low cost. Despite decimeter-level accuracy, current localization systems require labor-intensive site survey or the deployment of multiple Access Points and antennas. This paper presents a calibration-free fingerprinting-based localization system implemented on less Commercial Off-the-Shelf (COTS) Wi-Fi devices achieving high accuracy. For this purpose, we first present a comprehensive analysis about how the hardware design (inter-antenna distance) influences the localization performance. Then we promote the localization performance via hybrid inter-antenna distance and context-aware fingerprinting recognition. Finally, we build a prototype on Intel 5300 NIC and conduct experiments in different real-world scenarios with diverse environment. The results show that we can realize the 80% errors within 0.2m, which is more accurate than the existing designs.