Enhanced Data-Driven LoRa LP-WAN Channel Model in Birmingham

Abstract

Innovative solutions providing better coverage and minimized power consumption by end nodes such as Low Power Wide Area Networks (LP-WAN) have facilitated the advances towards improved IoT connectivity. Long Range Wide Area Net-work (LoRaWAN) technology stands out as one leading platform of LP-WANs receiving vast attention from both industry and academia. Performance evaluation of LoRaWAN is promising, in particular in the field of outdoor localization and object tracking. Limitations of node ranging and tracking without the need of energy-draining solutions like GPS, however, has not been tackled thoroughly. In this work, we explore the performance of the LoRa LP-WAN technology using real-life measurements in Birmingham, UK, using commercially available equipment. We present a channel attenuation model that can be utilized to estimate the path loss in 868 MHz ISM band in urban-similar areas. The proposed channel model is then compared to previously well-identified empirical path loss models and enhanced by detecting and eliminating outlier data from the obtained real measurements for an optimal fitted model. We, further, propose a novel RSSI distribution-based and k-means clustering to enhance the power-to-distance prediction accuracy that improves absolute errors by 4% and 18%.