Autonomous Farming Robots for Real-Time Weed Detection and Removal using YOLOv8

Résumé

The increasing demand for sustainable agriculture has driven the development of autonomous farming solutions for precise weed management. Traditional weed control methods, including manual removal and chemical herbicides, are labor-intensive, environmentally harmful, and economically inefficient. This study proposes an autonomous farming robot equipped with YOLOv8 (You Only Look Once, version 8) for real-time weed detection and removal. The system integrates high-resolution cameras, deep learning-based image processing, and a robotic arm with an adaptive end-effector to eliminate weeds efficiently. The YOLOv8 model, trained on a dataset of 50,000 images, achieved an mAP@50 of 92.4%, demonstrating superior performance compared to existing state-of-the-art detection models. The robot, tested across various crop fields, achieved an average weed removal accuracy of 89.7%, reducing herbicide usage by 67% while increasing yield potential by 15%. Compared to manual weeding, the system improved operational efficiency by 58%. These findings highlight the potential of AI-driven robotic systems in enhancing agricultural productivity, minimizing chemical dependency, and promoting eco-friendly farming practices. Future work will focus on multi-weed classification, real-time adaptation to diverse field conditions, and energy-efficient navigation to further optimize performance.