Machine Learning-Based Indoor Localization System to Support 5G Location-Based Services

Abstract

Indoor positioning is expected to play a vital role in future 5G networks, as it enables a wide range of location-based services such as indoor navigation, inventory monitoring, store locators, and anti-theft prevention. Accurate and reliable positioning information is crucial for many 5G applications, and it has been identified as a key enabler by the Third Generation Partnership Project (3GPP), which recently standardized advanced positioning methods in Release 15 and 16. This work focuses on the feasibility of using Machine Learning (ML) algorithms for Long Range Wide Area (LoRa)-based wireless indoor positioning systems, where LoRa-enabled sensor nodes can smoothly access 5G networks. The study trains different ML classifiers, such as Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Linear Discriminant Analysis (LDA), and Gaussian Naïve Basis, on Received Signal Strength Indicator (RSSI) values received from three different anchor nodes in an experimental setup. The experimental results demonstrate that KNN provides over 98% accuracy, 0.9831 precision, 0.9841 recall, and 0.9835 F1 score in estimating the location. This high level of accuracy and reliability makes ML algorithms a promising solution for indoor positioning systems in 5G networks, opening up new opportunities for location-based services such as roadside assistance in indoor scenarios, taxi-.hailing, and service locators.