RubikGeneticSolver is a fully automatic 3×3×3 Rubik’s Cube solver based on a genetic algorithm (GA) approach.
Unlike most modern solvers, this project does not rely on any pre-defined move sequences (often called “algorithms” in the cubing community) nor on pre-computed optimal lookup tables. Instead, it follows a three-step, human-like solving method.
While it is not an optimal solver, it typically produces solutions in 25 to 40 moves, which is quite respectable for a knowledge-free approach.
The solver tackles the cube sequentially using the following steps:
- Solve a 2×2×3 block at the back of the cube, using a knowledge-free GA
- Orient the remaining edges and resolve corner parity , again using knowledge-free GA
- Solve the remaining pieces in the (U, F) 2-generator subgroup. The program will first try brute-forcing this phase up to 15 moves depth, then fall back to a GA approach.
I originally started this project in 2005.
Nearly 20 years later, I revisited the source code, and—with the help of a very nice AI assistant—translated it from Fortran 90 to C#.
If you’re interested in the background and evolution of the project, feel free to read more on my website: