RSSI-based Localization Zoning using K-Mean Clustering

Abstract

This document discusses the novel approach to localize human location based on the current zone via k-mean clustering. A pilot experimental analysis of k-mean clustering to group similar RSSI pattern is compared to user defined zones. The dataset collected has demonstrated the difficulties of deploying trilateration and fingerprinting methods in dynamic conditions, as it introduces fluctuating RSSI measurements. The k-mean clustering was proposed to validate antenna placement to divide the testbed into zones, and to create a baseline accuracy which can be compared with other algorithms in the future. It was found that, for Zones 2, 3 and 4, the k-mean clustering of antennas agree with the planned antenna placement grouping. However, for Zone 1, there are differences, which may be attributed to a large metal obstacle occupying the zone. The k-mean clustering also recorded a peak accuracy of 79% with k=4, which agrees with the number of planned zones.