Enhancing LoRa-based Outdoor Localization Accuracy Using Machine Learning

Autorzy Kelesoglu N.; Halama M.; Strzoda A.
Tytuł Enhancing LoRa-based Outdoor Localization Accuracy Using Machine Learning
Czasopismo IEEE Access
Rok 2025
Status Published
Tom 13
DOI 10.1109/ACCESS.2025.3589032
URL https://ieeexplore.ieee.org/document/11080010
Abstrakt <p>The Internet of Things is gaining significant relevance, driving increasing interest in location-<br />
based services using wireless signals, particularly Low Power Wide Area Network (LPWAN) technology.<br />
LoRa (Long Range), together with LoRaWAN, is a prominent LPWAN standard that provides long-range<br />
connectivity and low energy consumption, making it viable for IoT-based positioning systems in smart<br />
cities. For localization systems leveraging LoRa signals, Machine Learning (ML) approaches are being<br />
increasingly explored, as ML-based solutions offer a powerful way to enhance the accuracy of positioning.<br />
In this study, we propose various ML approaches for LoRa-based positioning in outdoor environments. We<br />
evaluate six different ML models: k-NN, CNN, SVR, ANN, XG-Boost, and LightGBM-using an open-<br />
source urban LoRaWAN dataset. We further propose a Hybrid Model that combines convolutional feature<br />
extraction with gradient-boosted regression. This architecture integrates the strengths of deep learning and<br />
tree-based models, aiming to capture both temporal signal patterns and structured input correlations for<br />
improved localization accuracy. The models are trained offline and tested for performance in terms of<br />
localization accuracy, mean square error, and computational efficiency. Additionally, we investigate the<br />
impact of different Feature Vector (FV) subsets on localization performance by analyzing the significance of<br />
LoRaWAN signal attributes. Our results highlight the effectiveness of ML models in enhancing localization<br />
accuracy for LoRa-based outdoor positioning systems, demonstrating performance improvements ranging<br />
from 10% to 73% compared to previous ML studies in outdoor localization.</p>
ISSN 2169-3536