Indoor localization through wi-fi rf fingerprinting using machine learning
Loading...
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT.Lahore
Abstract
Global Positioning System (GPS) is the most reliable solution for outdoor localization. However, the GPS signal has less penetration power and does not provide accurate position estimation for indoor areas. In this research, Wi-Fi RF fingerprinting using RSSI and Machine Learning is implemented to predict the location of a person in an indoor environment. Radio Frequency (RF) signals provide long-range area coverage and Wi-Fi – which is the most widely used communication technology utilizes RF signals for communication. RSSI is accessible to all devices and does not require any additional hardware for its implementation. This means that no additional or special hardware is required for the implementation of Wi-Fi-based indoor location methods. RSSI signals are highly affected by obstacles and changes in environmental characteristics. Therefore, collecting large amounts of RSSI data at different time instances is crucial. Hence, to overcome these limitations and improve the accuracy of the model Machine Learning is introduced. Machine Learning model – Extreme Gradient Boost is implemented in this research for indoor localization and is trained on the RF mapping of RSSI fingerprints. Every indoor has its unique characteristics and to compare the performance of the desired approach with the models trained on open-source data sets is inappropriate. Therefore, traditional Machine Learning models, KNN, SVM, and Random Forest are also applied to evaluate the performance of XGBoost. From the results obtained, it is proved that the XGBoost achieves the maximum percentage of F1- score as 98 % , outperforming KNN, SVM, and Random Forest at 96.75 %,95.34 %,and 97.65 % respectively.
Index Terms: Indoor localization, machine learning, received strength indicator (RSSI), supervised machine learning, Wi-Fi fingerprinting.