Comparison of RSSI-Based Fingerprinting Methods for Indoor Localization
Dominik Csík, Ákos Odry, Peter Šarčević
Abstract
The widespread of Internet of Things (IoT) has increased the need of accurate indoor localization methods. The indoor localization problem aims to determine the object position in technology-deficiated environments., where the solution requires the application of both alternative sensors and efficient algorithms. This paper addresses three fingerprinting techniques and provides a comparative analysis based on real measurements. Namely, the performance of Weighted K-Nearest Neighbor (WKNN), Random Forest (RF) and Artificial Neural Network (ANN) fingerprinting approaches are evaluated. First, a database is generated of Received Signal Strength Indication (RSSI) values of five access points (APs) in laboratory environment. Then, heatmap-based fingerprinting is elaborated., and a comprehensive analysis is conducted in two important cases. In the first case all points are line of sight (LOS), while in the second case, the modules are covered by a column. The obtained results show that the ANN-based approach outperforms the WKNN and RF methods, thereby proving its efficient applicability in indoor localization problems.