This repository contains the technical documentation for the AFGCR project. It serves as a comprehensive record of the research methodologies and algorithmic developments of the project.
The Autonomous Floating Garbage Collecting Robot (AFGCR) is a research-oriented autonomous surface vehicle (ASV) designed to detect and retrieve floating plastic waste in unstructured aquatic environments (lakes, canals, and rivers) while monitoring water quality.
Unlike commercial solutions that rely on expensive industrial sensors or manual remote control, this project focused on developing a low-cost, fully autonomous system using computer vision and edge computing.
- Autonomous Navigation (Visual SLAM): Capable of mapping unknown aquatic environments without GPS reliance, utilizing a custom-tuned GraphSLAM approach (RTAB-Map) optimized for dynamic water surfaces.
- Deep Learning Object Detection: Real-time detection of 7 classes of floating plastic waste using a custom-trained YOLOv3 model, achieving 75.87% mAP.
- Environment-Aware Path Planning: Integrated ROS Navigation Stack with a Dynamic Window Approach (DWA) local planner, modified to handle the drift and inertia of hydrodynamic movement.
- Water Quality Monitoring: Real-time IoT logging of pH levels to correlate plastic pollution with water acidity.
- Containerized Architecture: Fully modular software stack deployed via Docker on Nvidia Jetson Xavier NX.
Team Members:
- Asmaa Hassan
- Mahmoud Fathy
- Marwa Abd-El-Ghany
- Mohammed Hussein
- Mostafa Ashraf
Supervisor:
- Dr. May Ahmed Salama (Benha University)
Funding & Support:
- Funded by the Academy of Scientific Research and Technology (ASRT), Egypt.
- Valeo National Undergraduate Competition.