A Robot-Ready AI Decision Support System for Real-Time Detection and Weather-Based Management of Potato Late Blight

Abstract Late blight is one of the most destructive plant diseases and a particular problem for Algerian potato farmers. Currently, most Algerian farmers still use traditional methods, such as weekly fungicide spraying, to prevent the disease from appearing and spreading. Others prefer to take the risk and avoid spraying due to the high cost and harm of these fungicides. The developed system addresses this issue with an intelligent Decision Support System (DSS) based on real-time blight detection by a YOLOv8 (You Only Look Once) model trained on a customized dataset and on online weather data. It provides the best-timed decisions and schedules for spraying. The system is built on the Robot Operating System 2 (ROS 2) framework for easy integration into a real robotic platform. The critical information is presented in a web interface, which allows the farmer to monitor and control the system’s actions (like accepting or refusing spray missions or turning the camera on or off). The experiments demonstrate that the YOLOv8n model scored an overall mean average precision ([email protected]) of 0.920, with an overall precision of 0.885 and recall of 0.832. Furthermore, the DSS successfully integrates disease detection alerts and weather-based risk prediction to suggest accurately when to trigger spray missions. This combination of weather forecasting (as a proactive defense layer) and real-time artificial intelligence detection (as a reactive defense layer) aims to guarantee crop safety and minimize fungicide utilization, which could lead to lower costs and improved crop quality.