GeoSeg: Path Loss Estimation Method for Outdoor LoRa LPWAN

Accurate path loss modeling in low-power wide-area net-works (LPWAN) benefits a variety of processes such as network design and analysis, and supports applications such as localization and tracking. Land cover information is indispensable in accurate path loss modeling; most of the path loss models employed in LoRa LPWAN studies utilize land cover information to create a regional description of the propagation environment (e.g., urban, semi urban, etc.). However, this regional approach to path loss modeling is incongruent with the anisotropic characteristic of propagation environments with long radio links and diverse land covers which is the case in outdoor LPWAN deployments. Recent LPWAN studies show that using per-link modeling with detailed land cover data from satellite images gives more accurate path loss estimates than regional modeling. Similarly, geographical clustering is another technique that addresses the issue related to regional modeling by identifying subregions within the propagation environment and developing per-subregion path loss model. This study introduces GeoSeg, an enhanced log-distance path loss model that integrates per-link modeling with geographical clustering. Developed and evaluated using an open-access dataset for LoRa, GeoSeg demonstrates 23% improved path loss estimation accuracy compared to the standard log-distance model.