BarFi: Barometer-Aided Wi-Fi Floor Localization Using Crowdsourcing

Abstract

As an important supporting technology, floor localization in multi-floor buildings plays significant roles in many indoor Location Based Service (LBS) applications such as the fire emergency response and the floor-based precise advertising. While the majority of Received-Signal-Strength (RSS)-fingerprint-based wireless indoor localization approaches suffer from the labor-intensive and time-consuming site-survey and the low localization accuracy, barometer-based floor localization is another promising direction due to the increasing availability of the barometer-sensor-equipped smartphones. This paper is the first indoor localization work that exploits the combination of Wi-Fi RSS and barometric pressure for accurate floor localization. Compared with an art-of-the-state algorithm, B-Loc, the highlight of the proposed Bar Fi approach is that it does not need all client smartphones but only low percentage of them equipped with barometer sensors. Using crowd sourcing, Bar Fi eliminates the need of war-driving of site-survey and prior knowledge about both the Wi-Fi infrastructure and the floor plans of buildings. The key novelty of Bar Fi is a two-phase clustering method proposed to train the RSS fingerprint floor map with the aid of barometer, which consists of a barometer-based hierarchical clustering phase and a Wi-Fi-based K-Means clustering phase. The real-world evaluation shows Bar Fi achieves satisfying performance that its accuracy reaches 96.3% when the proportion of smartphones equipped with barometer sensors is 12% out of the total.