AutoCHESS System segments
AutoCHESS was my Master's thesis project. It combined several systems in order to enable a fully automated human versus robot chess match.
The system used a PyTorch trained neural network model inside a vision system called Ocellus. Ocellus sent the data to Node-RED, a visual scripting IoT tool that processed the data and created commands that would trigger specific routines on the robot controller. Stockfish was used to estimate the best possible move for the ABB's YuMi Cobot.
The system is described in more detail in the Master's thesis document which you can find here:
AutoCHESS was featured last year on the OTD Energy fair in Stavanger.
Ocellus, and many other parts of the system are confidential, so I'm not permitted to share that code. The thesis, images and videos give a good impression on how the system works.
https://www.youtube.com/watch?v=KIPpZ1LeXJQ&ab_channel=DomagojKudek
