Department of Computer Science and Engineering

Indoor location estimation, the process of reckoning the location of a device in an indoor environment, remains a technical challenge due to the poor performance of GPS in such settings. While a substantial amount of work has been done in this context, particularly employing the Wi-Fi fingerprinting technique, the very approach has certain shortcomings. A major limitation is the need of time-and-labor-intensive fingerprint acquisition process. What is more, the costly fingerprint soon gets outdated because of the dynamic environment. The alternative, triangulation-based systems are not only complex to build because of the increased multipath signal propagation indoors, but also require prior knowledge of the location of Wi-Fi access points which is not always possible, or requires dedicated beacons which is not costeffective. Here we present an indoor location estimation approach that employs an already deployed Wi-Fi network, without requiring any prior knowledge of the position of the access points, or the need for manually collecting the fingerprints, and with dynamic environmental adaptability. This is achieved by crowdsourcing the fingerprinting process using localized QRCodes and NFC tags as reference points for bootstrapping this process. We have developed the complete system including the location estimation algorithm and a mobile mapping application to demonstrate that our approach can achieve 10-meter accuracy for 64% of the location estimations, and 98% accuracy in estimating the floor, using a reference tag density of 1 tag per 400