Novel RSSI-Based localization in LoRaWAN using probability density estimation similarity-based techniques

Abstract

In localization tasks of Internet of Things (IoT) End Nodes (ENs), the network lifetime and energy efficiency are critical. Due to power constraints, traditional systems like the Global Positioning System (GPS), Global Navigation Satellite System (GLONASS), and Galileo may be unsuitable for IoT applications. As a result, Long-Range Wide Area Network (LoRaWAN) has gained attention due to its large coverage and low power requirements. Traditional localization strategies typically estimate the distance between the EN and Anchor Nodes (ANs) using the Received Signal Strength Indicator (RSSI) combined with a path loss model. However, the accuracy of such an approach can be compromised by different undesirable transmission effects, such as interference, affecting the RSSI. This work introduces a novel distance estimation method that leverages the similarity between Probability Density Functions (PDFs) of RSSI from measurement campaigns and those from deployed ENs. By employing metrics including the enhanced versions of Euclidean and Minkowski distances, the proposed approach surpasses conventional channel-based techniques, achieving a Mean Absolute Percentage Error (MAPE) of 3.9% for wireless environments with a shadowing standard deviation up to 16 dB. Furthermore, when utilizing Kernel Density Estimation (KDE) for localization, the method demonstrated an 95.1% enhancement in accuracy compared to the localization strategy based on the loglinear path loss model.