LoRa-based Differential Fingerprint Localization in Outdoor Environments

Existing research on localization using LoRa-based fingerprinting have focused on machine learning algorithms and not on the form of location fingerprints. For instance, the fluctuations of RSS measurements from LoRa devices are overlooked. In this work, we propose a differential fingerprint which fuses Signal Strength Difference (SSD), gateway information, and Time Difference of Arrival (TDoA) to improve localization performance. A publicly available LoRaWAN dataset is preprocessed to acquire a database of the proposed differential fingerprint. The database is then used to train two algorithms: Random Forest (RF) and Multilayer Perceptron (MLP). The best performing model is the Multilayer Perceptron-Out of Range Differential (MLP-OoRD) regressor which achieved a mean error of 310 meters and a median error of 57 meters with chronological data split.