Sky Chess
A physical chessboard you play against a machine. A camera reads the board, and a repurposed laser engraver moves the pieces.
- Year
- 2025
- Context
- Independent study, spring of junior year
- Timeline
- One quarter
- Discipline
- Robotics, Computer vision
A chessboard system that uses machine learning, OpenCV and an XY-gantry to let you play physical chess against a machine. I built it as an independent study in the spring of my junior year.
After one quarter it worked around 65% of the time, and I plan to make it more reliable.
The build
01
Custom grippers
To pick up the pieces, I designed a custom end-effector. A micro servo drives geared clamps that grab a piece, and a stepper motor on a lead screw raises and lowers it.
02
Computer vision + machine learning
To read the state of the board, I taught myself computer vision. The system averages the color values of each square and feeds them into a k-nearest-neighbors classifier I trained with scikit-learn.
03
Gantry and communication
I repurposed an inexpensive laser engraver as the XY-gantry. A Raspberry Pi talks to it over serial, pulling G-code positions from a Python dictionary driven by the game logic.
