MA-NO

Inspiration

Recycling is still heavily dependent on manual sorting, which is repetitive and inefficient. We wanted to explore how robotics and AI could automate this process in a simple, understandable way. MA-NO was inspired by the idea of removing humans from the most repetitive part of recycling. The name is a play on the Italian word mano, meaning hand, and also “man no,” meaning no human required. Our goal was to build a clear demonstration of how computer vision and a robotic arm can work together to identify and sort recyclable materials. We also created a website with a live demo hosted on Jade so people can see how the system works in real time.

What it does

MA-NO is an AI powered robotic recycling sorter. A camera observes items placed in front of the system and a computer vision model identifies whether the object is plastic, aluminum, or paper. Once the item is classified, the robotic arm picks it up and places it into the correct recycling bin. The process is fully automated and requires no human intervention after the item is placed in the sorting area. The goal is to demonstrate how robotics can handle repetitive sorting tasks. We also built a website with a live demo hosted on Jade so users can see the system working and understand the pipeline behind the project.

How we built it

MA-NO combines robotics, computer vision, and simulation tools. The system uses an SO101 robotic arm connected through the Cyberwave platform to perform the physical sorting. A camera captures images of objects placed in the sorting area, and a computer vision model identifies the material category. Because our access to the robot was limited, we developed and tested most of the workflow using Cyberwave simulation tools. The pipeline connects detection, classification, and robotic motion into a simple process. We also built a website that explains the project and includes a live demo hosted on Jade so viewers can interact with the concept and see how the system works.

Challenges we ran into

One of the biggest challenges was limited access to the robotic arm, which meant most development had to happen in simulation before testing on real hardware. Integrating the computer vision model with robot control also required careful coordination between different tools and libraries. Another challenge was making the robot reliably pick up different types of objects, since materials like paper or bottles behave differently when grasped. We simplified the setup to make the demo reliable and easy to understand. We also spent time building the website and live demo hosted on Jade so people could explore the project even without seeing the physical robot.

Accomplishments that we're proud of

We are proud of building a complete end to end system that connects perception, decision making, and robotic action. The system can identify objects and physically sort them into bins with minimal human involvement. Another accomplishment is keeping the project simple and focused on a real problem that people immediately understand. We also successfully used simulation tools to design the workflow before running it on real hardware. In addition to the robotics system, we built a website that presents the project clearly and includes a live demo hosted on Jade so anyone can explore how MA-NO works and see the concept in action.

What we learned

This project taught us how challenging and rewarding it is to integrate multiple parts of a robotics system. Computer vision, simulation tools, and robotic control all need to work together reliably for the system to function. We also learned how valuable simulation is when hardware access is limited. Designing the workflow in a simulated environment helped us move faster and avoid mistakes. Another lesson was the importance of simplifying robotics problems to create a reliable demo. Finally, building a website and live demo on Jade showed us how important it is to present technical projects clearly so others can understand and interact with them.

What's next for MA-NO

The current version of MA-NO is a proof of concept, but there are many ways it could be expanded. One improvement would be training the vision model to recognize more types of recyclable materials and objects. Another direction is improving the robot’s grasping ability so it can handle a wider variety of shapes and materials. We would also like to improve the user experience of the website and expand the live demo hosted on Jade so visitors can interact with the system more directly. In the future, MA-NO could grow into a larger automated recycling platform or integrate with mobile robots for collection and sorting tasks.