Comparative Analysis of Offline Databases for Wi-Fi Fingerprinting for Positioning Algorithms in Real Environment

Abstract

The demands for location-based services (LBS) are increasing day by day for indoor environments. Due to the inability of the Global Positioning System (GPS) signals to penetrate through surfaces like roofs, walls, and other objects in indoor environments, various alternative methods for user positioning have been proposed. Among them, the Wi-Fi fingerprinting approach has sparked significant interest in Indoor Positioning Systems (IPS) because it eliminates the need for line-of-sight measurements and achieves higher performance even in complex indoor environments. For indoor positioning, offline and online are the two phases of the fingerprinting method. Different authors have highlighted the problems in the offline phase as it deals with huge datasets and validation of the Fingerprints without pre-processing of the datasets become a concern. Efficient systems are important to minimize delays, complexity and the associated additional costs. Positioning accuracy of Wi-Fi indoor positioning systems highly depends upon offline databases. Therefore, development of robust Wi-Fi fingerprints is performed to improve the positioning results. Machine learning has been used for the model training in the offline phase whereas the locations are estimated in the online phase. Machine learning algorithms are a natural solution for winnowing through large datasets and determining the significant fragments of information for localization, creating precise models to predict an indoor location. Large training sets are important for improving results in machine learning problems. Therefore, an existing WLAN fingerprinting-based multi-story building location database has been used with 21049 samples, divided into 19938 training and 1111 testing samples. The proposed model uses mean and median filtering as pre-processing techniques applied to the database to improve accuracy by reducing the effect of environmental dispersion, as well as machine learning algorithms (kNN, WkNN, FSkNN, and SVM) for estimating the position. The proposed SVM with median filtering algorithm gives a reduced mean positioning error of 0.7959m and an improved efficiency of 92.84% as compared to all variants of the proposed method for 108703m2 area. On the basis of the results of this research, it can be concluded that it is possible to use Wi-Fi fingerprinting for indoor positioning to obtain a state-of-the-art accuracy.