AutoBot (Autonomous Utility for Trash Organization)
Course Instructor
Don Lee
Abstract
The global waste crisis is accelerating faster than current manual collection methods can manage, leading to significant environmental degradation and inefficient resource allocation. Traditional waste management systems often rely on fixed schedules and manual labor, which contribute to high carbon emissions and leave critical waste hotspots unaddressed. To bridge these operational and environmental gaps, our team has developed AUTO-BOT (Autonomous Utility for Trash Organization Robot). This system is designed to modernize waste collection through integrated robotics, computer vision, and real-time data tracking, providing a scalable solution for facility managers to identify and manage disposal trends effectively.
The core of the AUTO-BOT system is a mobile platform equipped with a high-precision robotic arm. While the project initially explored wide-scale autonomous navigation, the current focus has been refined to the critical tasks of object detection, classification, and physical collection. The robot utilizes a Raspberry Pi 5 as its primary processing unit, integrated with a high-definition camera and a potential 4-Degree of Freedom (4-DOF) robotic arm. To achieve high accuracy in cluttered environments, we implemented a custom-trained YOLOv8 (You Only Look Once) deep learning model. This model was trained using the TACO (Trash Annotations in Context) dataset, which we specifically refined to improve the detection of common litter items such as bottles, cans, and wrappers under varied lighting conditions.
The system's workflow begins with the computer vision module, which scans the immediate vicinity to locate and categorize trash. Once an item is identified, the software calculates the necessary coordinates for the robotic arm to execute a precise pick and place maneuver, depositing the waste into an onboard collection bin. A key innovation of AUTO-BOT is its data-driven approach; every item collected is logged and visualized on a Streamlit-based dashboard. This allows users to monitor the robot’s health and view real-time statistics on waste composition, transforming a simple cleaning tool into a sophisticated data-gathering asset.
Technically, the project addresses significant hardware challenges, including power management for multiple high-torque servos and the synchronization of dual Raspberry Pi units to balance processing loads. By isolating power supplies, the team prevented system reboots during intense motor actuation, ensuring continuous operation. AUTO-BOT represents a shift toward smart waste management, offering a reliable, data-centric alternative to manual labor. By automating the identification and retrieval of litter, this project demonstrates the potential for robotics to reduce the human footprint on the environment while providing the analytical tools necessary to foster long-term sustainability and recycling initiatives within large facilities and public spaces.
AutoBot (Autonomous Utility for Trash Organization)
The global waste crisis is accelerating faster than current manual collection methods can manage, leading to significant environmental degradation and inefficient resource allocation. Traditional waste management systems often rely on fixed schedules and manual labor, which contribute to high carbon emissions and leave critical waste hotspots unaddressed. To bridge these operational and environmental gaps, our team has developed AUTO-BOT (Autonomous Utility for Trash Organization Robot). This system is designed to modernize waste collection through integrated robotics, computer vision, and real-time data tracking, providing a scalable solution for facility managers to identify and manage disposal trends effectively.
The core of the AUTO-BOT system is a mobile platform equipped with a high-precision robotic arm. While the project initially explored wide-scale autonomous navigation, the current focus has been refined to the critical tasks of object detection, classification, and physical collection. The robot utilizes a Raspberry Pi 5 as its primary processing unit, integrated with a high-definition camera and a potential 4-Degree of Freedom (4-DOF) robotic arm. To achieve high accuracy in cluttered environments, we implemented a custom-trained YOLOv8 (You Only Look Once) deep learning model. This model was trained using the TACO (Trash Annotations in Context) dataset, which we specifically refined to improve the detection of common litter items such as bottles, cans, and wrappers under varied lighting conditions.
The system's workflow begins with the computer vision module, which scans the immediate vicinity to locate and categorize trash. Once an item is identified, the software calculates the necessary coordinates for the robotic arm to execute a precise pick and place maneuver, depositing the waste into an onboard collection bin. A key innovation of AUTO-BOT is its data-driven approach; every item collected is logged and visualized on a Streamlit-based dashboard. This allows users to monitor the robot’s health and view real-time statistics on waste composition, transforming a simple cleaning tool into a sophisticated data-gathering asset.
Technically, the project addresses significant hardware challenges, including power management for multiple high-torque servos and the synchronization of dual Raspberry Pi units to balance processing loads. By isolating power supplies, the team prevented system reboots during intense motor actuation, ensuring continuous operation. AUTO-BOT represents a shift toward smart waste management, offering a reliable, data-centric alternative to manual labor. By automating the identification and retrieval of litter, this project demonstrates the potential for robotics to reduce the human footprint on the environment while providing the analytical tools necessary to foster long-term sustainability and recycling initiatives within large facilities and public spaces.
