| Definition | Solving Twisty Puzzles with Reinforcement Learning(Parallel Q-Learning) |
| No of environments | 7 |
This environment simulates a set of twisty puzzles, including the Rubik’s Cube, as part of the gym framework. Upon initialization, one of the three included twisty puzzles should be selected and scrambled in the reset function. Afterward, the RL model you design will be trained to solve these puzzles by learning efficient solution strategies. The 3x3x3 Rubik’s Cube has 43,252,003,274,489,856,000 different state spaces and only one of them is the solved state which makes it an interesting NP (nondeterministic polynomial time) problem.
During training, your model will be refined over multiple episodes, analyzing its performance at each checkpoint. The idea is to assess the model’s progress by comparing different checkpoints in order to identify the best-performing version based on key metrics like accuracy and solving time.
| Example observations | State of cube’s color-coded (eg. 54) fields |
| Example actions | Back, Left, Right and Up; |
| Example reward | Array of fields in order |