Analysis of Distance Measures for WiFi-based Indoor Positioning in Different Settings

Abstract

Recently, indoor positioning systems based on wireless technologies such as WiFi fingerprinting become more popular. The nearest neighbor-based algorithms using Euclidean distance are very common and used in many fingerprinting systems. Thus, the distance measure is very important and it affects much to the tracking result. In this paper, we present an analytical study of using different distance measures for the weighted K-nearest neighbor algorithm to determine the position of a user. We implement five distance measures and compare the positioning results of each measure to find out the best one. To check the robustness of the measures, we change some settings when creating the radio map in the offline phase such as the number of access points or the distance between two reference points. From the experiments, it is shown that the Chi-Squared distance outperforms other distance measures since it achieves the mean error of 1.13 meters in a simple test case and 1.20 meters in a more complicated test case. Even when we change the settings, Chi-Squared distance remains the best positioning result.