Indoor Wi-Fi RSS-fingerprint location algorithm based on sample points clustering and AP reduction

Abstract

The accuracy of RSS fingerprint based indoor location algorithms in Wi-Fi environment depends on the density of sample points and the quality of AP radios. It has been observed that in a given area the accuracy can be improved by just using the RSS data from a sub set of whole APs. So the location algorithm based on AP reduction is studied in this paper, and 3 kinds of sample points clustering methods, which are spatial clustering, K-means clustering and Affinity Propagation Clustering, are tested to generate the appropriate area for each AP sub set. The results of experiments shows that the AP reduction algorithm can obviously reduce location error. At the same time, the algorithm's complexity gets reduced.