Design And Implementation of An Agricultural Robot for Weed Detection and Removal Using AI
N Jyothi,
Konda Rakesh Goud,
Kankala Akhila,
K. Abhishek,
K. Bhanu Prakash
Weed infestation significantly reduces crop yields by competing for essential resources, with losses
estimated at 20–40% in many Indian farms. Conventional weed control relies on manual labor or
indiscriminate herbicide spraying, which is labor-intensive, costly, and environmentally damaging. This
paper proposes an AI-powered autonomous agricultural robot for precise weed detection and targeted
removal. The system integrates computer vision using the YOLOv8 object detection model for real-time
weed identification, a Raspberry Pi 5 or NVIDIA Jetson Nano for edge processing, and a mechanical
end-effector (gripper or solenoid punch) for selective uprooting. A forward/downward-facing camera
captures field images, while navigation follows crop rows via simple vision-based guidance. Experimental
evaluation in simulated and small-scale field conditions (maize/vegetable plots) demonstrates detection
accuracy of ~92% mAP@0.5, successful removal rates of 85–93%, and up to 90% reduction in herbicide
usage. The prototype promotes sustainable precision agriculture by minimizing chemical input, labor,
and crop damage while enhancing traceability of operations.
