Robust indoor localization in ultra dense networks: A fingerprint similarity approach

Abstract

The performance of fingerprint based indoor localization techniques is significantly degraded by environmental dynamics especially when devices such as Wi-Fi access points (APs), which are used to build fingerprint database, are densely deployed. The primary reason is that the plug-and-play feature of these devices would render the fingerprints in the database invalid. Worsestill, the time-varying wireless channels would make the fingerprints unreliable as well. To overcome the bottleneck, a robust fingerprint similarity based indoor localization (FSIL) method has been proposed in this work, targeting at alleviating the impact of environmental dynamics on indoor localization. The principle of FSIL is to use the relative values of received signal strength (RSS), instead of absolute values, to construct fingerprint database. Accordingly, FSIL is more robust to environmental dynamics. Aided by the reliable fingerprint database, the position could be steadily estimated by FSIL, especially when some APs disappear and new APs emerge. Moreover, in order to combat the time variances of wireless signals, a weight priority criterion has been tailored in FSIL. The concept is to make the APs with stronger signal strength have larger weights in localization estimation. Experimental results have confirmed the benefits of FSIL in improving localization performance. For instance, it is shown that FSIL could reduce the localization average error by 30.45% and 37.19%, respectively, compared to two state-of-the-art localization methods in 20-missing AP scenario.