Improvement of RSSI-Based LoRaWAN localization using Edge-AI

Abstract

Localization is an essential element of the Internet of Things (IoT) leading to meaningful data and more effective services. Long-Range Wide Area Network (LoRaWAN) is a low-power communications protocol specifically designed for the IoT ecosystem. In this protocol, the RF signals used to communicate between IoT end devices and a LoRaWAN gateway (GW) can be used for communication and localization simultaneously, using distinct approaches, such as Received Signal Strength Indicator (RSSI) or Time Difference of Arrival (TDoA). Typically, in a LoRaWAN network, different GWs are deployed in a wide area at distinct locations, contributing to different error sources as they experience a specific network geometry and particular environmental effects. Therefore, to improve the location estimation accuracy, the weather effect on each GW can be learned and evaluated separately to improve RSSI-based distance and location estimation. This work proposes an RSSI-based LoRaWAN location estimation method based on Edge-AI techniques, namely an Artificial Neural Network (ANN) that will be running at each GW to learn and reduce weather effects on estimated distance. Results have shown that the proposed method can effectively improve the RSSIbased distance estimation accuracy between 6% and 49%, and therefore reduce the impact of the environmental changes in different GWs. This leads to a location estimation improvement of approximately 101 m.