A RSS-based fingerprinting method for positioning based on historical data

Estimating the position of people in an indoor WLAN environment poses a fundamental challenge in ubiquitous computing. By using Wi-Fi, it is possible to determine the position of people or assets with good accuracy. K Nearest Neighbors (KNN) is one of the most popular deterministic location fingerprinting algorithms generally used for WLAN-based indoor positioning. As KNN takes only the K nearest neighbors for estimating a position, in some cases it may not obtain satisfied accuracy because of the indoor environment factors such as reflections, diffraction, and scattering of the radio waves. In this paper, we propose a novel method named Predicted K Nearest Neighbors (PKNN) which estimates the current position of a mobile user not only by using K found neighbors but also by utilizing its previous positions and speed. With experiments, we found that PKNN does outperform KNN by 33% or at mean 1.3 meter improvement in error.