I have been playing the Asian board game GO since my youth. Since I am also very interested in robots, I already had the idea to build a robot for it in the 1990s. However, at that time I didn't have the resources to do so. Through the Raspberry PI and the inexpensive electronics, this has now become possible.
The robot should play GO authentically and behave as much as possible like a human gamer.
To do this, the robot can use a camera and OpenCV to recognize where the player has placed a stone on the playing field.
It then uses GnuGo artificial intelligence to calculate the next sensible move.
Using a pneumatic gripper, it can then pick up a tile and place it on the board.
He can also use the gripper to remove captured pieces from the board.
The most exciting part of this project for me was the construction of the gripper.
I made the first attempts with mechanical grippers, but since GO stones are very smooth, that didn't work out well.
I had no experience with pneumatics before and the first version of the pneumatic gripper was very noisy and still needed a big compressor. The construction of the separator for new tiles was also tricky because the tiles always have to be in the same position and they don't get lost. For this, I had to learn OpenSCAD intensively.
When I exhibited iGoBot at a Maker Faire, most visitors did not know the rules of GO.
But since these rules can be learned very quickly, the visitors were able to play a small game against the robot themselves after a few minutes.
I'm always happy when more people get to know the game and maybe want to come to a local GO game event in your city.