Enhancement of robustness and precision of indoor positioning by fusing wifi fingerprinting and Pdr techniques

Abstract

Fused approaches for enhancing robustness and precision of indoor positioning using pedestrian dead reckoning (PDR) and KNN (K-Nearest Neighbors) classifier based WiFi fingerprinting were proposed. The proposed machine learning approaches employed the rough position estimate by PDR as a pre-sorter of training vectors of KNN classifier and help improve precision by overcoming fluctuating radio signal and furthermore robustness in serious radio signal distortion by the undesired malfunction of WiFi signal sources. The experiment in real space showed significant improvement in both precision and robustness.