Gojo Satoru

chartFactor Studio Visualization of Litter in New York City
PCB Design
PCB Design
One of our LitterBots without it's main components
What we envisioned for our LitterBots
LitterBot Prototype programming
Azure AI training data
the azure ai in action

Inspiration

Many may remember the Cosmo and Vector robots from many years ago, they were small robots filled with personality that could sit on your desk and interact with you, but these robots served no real productive purpose. Although they were neat to play with, we wanted to create something that could help the environment around us while providing a small amount of personality to themselves as well. This led us to create the LitterBot, a small robot that is meant to detect, retrieve, and recycle trash.

What it does

LitterBot would use AI detection to recognize and locate trash in its retrieval area, and receive signals from its Base Station, the point at which the LitterBot automatically homes to. This Base Station would serve two purposes - communication with the LitterBots in its range, and as the designated trash area. When the LitterBot retrieves a piece of trash and homes back to its Base Station, it drops the trash underneath the Base Station, where a hidden trash bin will be waiting for the LitterBot to deposit anything it finds. By using Base Stations, we can make each LitterBot inexpensive, as most of the heavy computing can be done in the cloud. However, picking up trash isn't the only path to stopping the rampant polluting of our planet. We will stop it at the source. Our LitterBots also come with unique, shocking, so to say, capabilities. Each LitterBot has a built-in high voltage boost converter, to correct anyone who may think littering is ok.

How we built it

The LitterBot is broken up into three parts, mechanical, electrical, and software. Mechanically, the LitterBot was constructed using TPU for its wheels, and PLA for the housing and shell. The electrical components are a custom milled PCB, ultra affordable SMG chips, DC motors, and 9 gram servos. The software for LitterBot was written in embedded C, and we used Microsoft's Azure for trash identification.

Challenges we ran into

The challenges we ran into were more related to manufacturing as they were to design. Milling PCBs gives great flexibility, but it also comes with its many challenges, including unreliable traces. On the software side of things, we ran into problems with creating a user-interface to use our robots, and thanks to Microsoft's Azure we were able to do object detection much easier than if we had done it from scratch. We also had a team member accidentally encrypt their entire laptop with many files we needed, but luckily we were able to recover everything after about an hour.

Accomplishments that we're proud of

We are proud of what we have managed to build in the past 36 hours, although not anything we would want to put out into the market currently, with a team around LitterBot and more time, we believe it could help clean city litter greatly. We are also very proud of how we managed to work as a team, channeling our strengths into where they were needed, and always staying busy making sure we were contributing until our skills were no longer needed.

What we learned

We learned about how ai uses pipelines, and how the pipelines are how the data flows for the AI. We also learned how to train the AI using data labeling, and how to properly label the data so that the AI has a solid foundation for how to recognize the tags.

What's next for Gojo Satoru

We believe that should we return to this project, we would like to focus on the litter recognition and retrieval most of all. After we have a robot that can detect and retrieve litter consistently, we would like to theme it in a similar manner as to the Cosmo and Vector robots, with personalities and minor interactions with those who come into contact with our LitterBots to help bring our anti-littering community closer together