sentientorange/rubiks-cube

Reinforcement Learning program that looks to be able to quickly learn to solve a Rubik's Cube

Rubiks-Cube

This program was created for learning about Artificial Intelligence and Reinfocement Learning. Q Learning was the technique chosen for this implementation of a Rubik's Cube Solver.

Topics

• Reinforcement Learning with Q Learning
• Machine Learning
• Finding convergence for reward functions when evaluating huge state spaces (Linear Approximation)

Contributors

• Derek Wang
• Eric Eckert

rubiks-cube's Issues

Figure out how to make the Q scores converge on something useful

Lol the thing that makes the project work doesn't work

Add functionality for 2x2x2

Theoretically we can design this to scale to any number of cubes. Maybe we should start with this?

Add a GUI or something so that we can better see what each cube looks like

There should be a python library out there that lets us draw stuff. The CLI can be used for the set up and control unless we want to go more into UI stuff.