Overview ROBOGOAT is an autonomous weeding robot developed as a final year mechatronics project at Queensland University of Technology. The robot is designed to navigate public footpaths while detecting and eliminating weeds using computer vision and targeted pesticide application. Key Features
Custom trained YOLOv8 object detection model for weed and path detection IR-filtered computer vision system using Raspberry Pi Global Shutter camera Targeted pesticide delivery system to minimize chemical usage Autonomous navigation using AI path detection Custom PCB design for power management and camera control Solar-powered operation for extended runtime
Technical Stack
Vision System: Python, OpenCV, YOLOv8 Hardware: Raspberry Pi 4, Arduino Mega, Custom PCB Computer Vision: NDVI (Normalized Difference Vegetation Index), IR filtering Control: C++ for motor and sprayer control Design: Fusion360 for chassis and components
Results
Path detection accuracy: >90% Weed detection accuracy: 49% overall, higher for weeds directly on paths Successfully implemented safety features including watchdog states Demonstrated reliable autonomous operation during field testing
Project Purpose The project aims to address the challenges of footpath maintenance in urban areas, specifically:
Reducing manual labor costs in weed management Minimizing human exposure to pesticides Preventing footpath damage from weed growth Reducing overall pesticide usage through targeted application