Mobile Crowdsensing Framework for a Large-Scale Wi-Fi Fingerprinting System

Abstract

Although Wi-Fi fingerprinting systems enable location-based services in indoor environments, the service coverage isn't scalable due to costly site surveys. Mobile crowdsensing (MCS), where casual smartphone users conduct the site survey, has recently emerged as a promising solution. Applying MCS to a system introduces new challenges: motivating active participation, inferring location information of unlabeled fingerprints, and managing a large amount of fingerprints. The proposed MCS framework obtains fingerprints from users by exploiting social network service applications and using the pedestrian dead reckoning technique. MCS accurately infers the location information of unlabeled fingerprints using physical-layout and signal-strength measurements. The framework also selects an optimal set of fingerprints, which introduces high accuracy with a slightly increased database size. This case study shows the feasibility of the proposed MCS framework.