Towards Dense and Scalable Soil Sensing Through Low-Cost WiFi Sensing Networks

Abstract

Precision agriculture uses precise sensor data collected throughout farmland to give farmers better insight into their land, allowing for greater crop yields and reduced resource usage. However, existing solutions require high hardware costs thus limiting large scale deployments. To address that, we propose a low-cost and scalable solution for sensing physical attributes of soil using IoT based WiFi sensing devices. By understanding variations in WiFi radio signals with channel state information (CSI) and machine learning models, we evaluate the proposed soil sensing system through experiments on physical soil traits such as soil moisture content, soil texture and position. Moreover, we also demonstrate how a mesh network of WiFi sensing devices allows us to predict the physical traits of the soil in the area between each pair of sensors, allowing for an increase in sensing area coverage as nodes are added.