An efficient clustering with robust outlier mitigation for Wi-Fi fingerprint based indoor positioning

Wi-Fi fingerprint systems provide cost-effective and reliable solution for indoor positioning. However, such systems incur high calibration cost in the training phase and high searching overhead in the positioning phase. Moreover, huge storage requirement for the radio map of a large-scale fingerprint system is another major issue. Several solutions based on crowd-sourcing or machine learning technique have been proposed in literature to reduce the calibration overhead. On the other hand, various clustering methods have been proposed over the past decade to reduce the searching overhead. However, none of the existing systems has addressed the issue of high storage requirement for the fingerprint database constructed in the training phase. Moreover, presence of outlier in the received signal strength (RSS) measurements severely impacts the positioning accuracy of such systems. Thus, this paper proposes an efficient clustering strategy for fingerprint based positioning systems to reduce the storage overhead and searching overhead incurred by such systems and also proposes a robust outlier mitigation technique to improve their positioning accuracy. The performances of our proposed positioning system are evaluated and compared with five existing fingerprint techniques in both the simulation test bed as well as real indoor environment via extensive experimentation. The experimental results demonstrate that our proposed system can not only reduce the storage overhead and searching overhead but also improve the positioning accuracy compared to the other existing techniques.