ML Based Control in Precision Agriculture: LED Intensity and CO2 Emission Case Study

This paper presents a new approach in controlling fundamental parameters (Light Intensity, Temperature, humidity ... etc.) related to precision agriculture using Machine Learning. We propose and design a new autonomous control system that applies data-driven approach. Several AI prediction models are developed using local data sets. Leveraging the power of Long Short-Term Memory (LSTM) models, the system aims to dynamically adjust light output in response to varying levels of carbon dioxide emissions. The model is deployed in a local server using a hardware architecture based on Raspberry Pi and a ESP32 microcontroller. Pi Server facilitates model deployment and data storage, while ESP32 provides wireless communication and peripherals interface to ensure efficient real-time data sensing and light control. Our results show that our multivariate model, which uses temperature, humidity, and CO2 emissions, provides better accuracy in terms of RMSE. Also, the embedded developed architecture facilitates the real time data sensing, collection and control.