Competitive agglomeration based KNN in indoor WLAN localization environment

Abstract

In this paper, we proposed a novel localization algorithm in indoor Wireless Local Area Network (WLAN) environment. First of all, to conduct the Received Signal Strength (RSS) preprocessing, we eliminate the RSS outliers based on the density function of the difference of RSS. Second, to overcome the problem of the manual selection of the cluster number, as well as the number of the nearest neighbors in K nearest neighbor (KNN) algorithm, we propose to use the Competitive Agglomeration (CA) algorithm to achieve the localization. Third, the extensive experimental results conducted in an actual Nonline-of-sight (NLOS) indoor WLAN environment, as well as in a simulated Line-of-sight (LOS) environment prove that the proposed approach performs well in localization accuracy.