Status : Verified
Personal Name Arriesgado, Alimuddin S.
Resource Title GeoSeg: Path Loss Estimation Method for Outdoor LoRa LPWAN
Date Issued 29 August 2025
Abstract Accurate path loss modeling in low-power wide-area networks (LPWAN) is critical for network design, analysis, and applications such as localization. While conventional models use regional land cover classifications (e.g., urban, semi-urban) to characterize propagation environments, these approximations fail to capture the anisotropic nature of long-range LPWAN links traversing diverse land covers. Recent studies show that per-link modeling achieves higher accuracy than regional methods. Similarly, geographical clustering is another technique that mitigates regional modeling limitations. This study introduces GeoSeg, an enhanced log-distance path loss model. GeoSeg incorporates the following methodological innovations: (1) combined per-link modeling with geographical clustering to develop a per-link variant of the log-distance path loss model; (2) developed a context aware embedding process that characterizes the 2D propagation environment between transceivers, incorporating land cover types, their spatial sequence, and positions relative to line-of-sight; (3) modified a Hidden Markov Model (HMM)-based technique for variable-length sequence clustering, to handle large-scale dataset processing; and (4) introduced a calibration process for the log-distance path loss model that is easily integrable to preexisting LPWAN deployments and/or datasets. Developed using an open-access LoRaWAN dataset, preliminary model evaluation shows improved path loss estimation accuracy compared to the standard log distance path loss model.
Degree Course Master of Science in Electrical Engineering
Language English
Keyword Low Power Wide Area Network, Path loss model, LoRa
Material Type Thesis/Dissertation
Preliminary Pages
645.91 Kb
Category : F - Regular work, i.e., it has no patentable invention or creation, the author does not wish for personal publication, there is no confidential information.
 
Access Permission : Open Access