Edgesoil: Low-Cost Soil Sensor and Edge AI Fusion in Agricultural Robotics

Abstract

Robots are becoming increasingly utilized in the agricultural industry for a variety of tasks such as harvesting, planting, disease detection, and plant treatment. However, there is a considerable gap when it comes to using soil analysis, even though it is essential for farmers to have good crops. Existing laboratory soil analysis processes are laborious and time-consuming for farmers, and robotic systems are usually expensive due to the cost of sensors and other components. To address this, a low-cost, mobile robot for soil analysis was developed. In the first step, an edge device with a camera that runs an AI model was integrated. One of the primary advantages of this system is that the whole sensor and camera combination for measurement is easily adaptable to any other existing mobile robot, and the overall cost of the measuring system including the edge device is less than 100 euros. Our first results regarding soil classification are promising with an accuracy of up to 90% during real-time classification. The next step is the integration of an NPK and PH level analysis within the AI model comparing it to the real-time measurement of an NPK sensor that goes into the soil. The ultimate goal at the end is to create an edge computing device: EdgeSoil that performs soil analysis just with a camera, adaptable to any agricultural robotic system.