Indigenous Community Connectivity: Enhancing LoRaWAN Performance Through Machine Learning in Palm Oil Plantations

The Internet of Things (IoT) has gained traction in agriculture. In Malaysia, palm oil plantation is the largest agricultural sector, which also offers jobs for indigenous communities in remote areas. This study explores the potential of LoRaWAnnetworks to enhance connectivity for IoT applications in palm oil plantations by developing a machine learning-based signal attenuation model at the 923 MHz LoRa frequency band. This model addresses the limitations of existing empirical models, providing more accurate predictions for LoRaWAnnetwork planning and optimization. Measurements were conducted under various conditions, including line-of-sight (LOS) and nonline-of-sight (NLOS) scenarios, with different gateway heights of 2.1m and 4.5m. The predictive machine learning (ML) model incorporated features such as 3D distance, environment type, gateway height, average branch width, and tree density. The results demonstrated a higher signal attenuation at 4.5m gateway height compared to 2.1m. The developed ML model achieved an accuracy of 87%, with an average attenuation error of 2.5 dB. This study offers a robust methodology for optimizing gateway placement, thereby enhancing the efficiency and cost-effectiveness in IoT-based palm oil plantations.