RiLoc: Representative guided subarea to exact localization from crowdsourced samples

With the emergence of the Internet of Things (IoT), indoor localization has become a critical infrastructure to be installed in large-scale environments such as airports, train stations, and shopping malls. Many systems have been proposed for accurate localization using the site survey, which is costly and tedious to implement in large-scale indoor environments. The high accuracy is also at the expense of computation time and memory. To deal with these challenges, we propose a representative guided subarea to exact localization from crowdsourced samples (RiLoc) for large-scale indoor environments. The basic idea of RiLoc is that online test points near each other in the signal space are also close to each other in the physical space. We cluster the crowdsourced samples in the offline phase to form subarea fingerprints and virtual anchors. We propose a novel virtual anchor-based dimension reduction algorithm using virtual anchors. In the online phase, the test points’ dimension is reduced and clustered to nominate a representative fingerprint. The representative identifies a more relevant and reduced search space and guides its members for exact localization. We conduct experiments on two field-measured datasets, and the results show significant improvements over the peer schemes in terms of lower localization errors.