Solving Rubik’s Cubes is a learnable skill. However, to compete at the top level, you’ll have to train hard. Speed cubers can solve a 3×3 cube in under ten seconds these days, after all. [aaedmusa] was a long way off that speed, but his robot is an absolute demon that solves at a rapid pace.
The robot relies on a Teensy 4.1 microcontroller to run the show, paired with its Ethernet kit for connectivity. It runs six stepper motors via TMC2208 drivers, enabling it to directly actuate each face of the cube. Purists will note, however, that the steppers are fitted with adapters that slot directly into modified center squares on the cube. A regulation Rubik’s, this is not.
The design doesn’t feature a machine vision system to capture the state of the cube. Instead, the cube’s status must be input to a web app on an attached computer. Once the cube’s state is loaded into the program, though, the mechanical job of solving the cube can be achieved in under five seconds. Even with six actuators, that’s not fast enough to beat the human world record of 3.47 seconds, but it’s still pretty darn good.
It’s funny to think that way back in 2011, both robots and humans were so much slower at solving cubes. If only all the world’s problems were as simple as a jumbled up toy from the 1970s.
https://hackaday.com/2016/01/27/robot-solves-rubiks-cube-in-just-one-second/
I think the world record has it down to 0.3s or so.
The human record is now 4.69 seconds:
https://www.youtube.com/watch?v=TODS4WQVIGY
It’s worth noting there are plenty of video-recorded/witnessed sub-4 human solves.
My son just got a 3.56 last week. Just no Guinness officials to verify them.
https://hackaday.com/2018/03/12/rubiks-robot-so-fast-it-looks-like-a-glitch-in-the-matrix/
I think the unofficial but pretty damn real one is that group of MIT students at their hackerspace MITERS a few years ago.
The video actually mentioned/reference that device
Someone must be reading my mind. Researching yesterday the construction of the rubix cube. Clever toy.
https://homes.luddy.indiana.edu/stsher/files/Rubiks_Cube.pdf
I always think of the inventor that came up with the ‘first’ Rubik cube. Amazing insight.
Neat project. Just getting it to twist, turn, and then solve is amazing (to me) . Regardless of the time.
Rebuild with servomotors and get the speed up by 10x!
Fantastic project, and I love the attention to detail on making a jam-free mechanism. I think with the better algorithm (monte-carlo tree search might work?) you can get the move count below god’s number for most solves without a massive compute time.
But because this is HaD, I guess it bears mention you could have just done this with a 555. ;P
Thistlethwaite’s algorithm
https://en.wikipedia.org/wiki/Optimal_solutions_for_Rubik%27s_Cube#Thistlethwaite's_algorithm
The GAN Robot, now priced at approximately $75, can typically solve the Rubik’s Cube in about 5 seconds.
However, it requires specific GAN cubes to accomplish this feat. The most practical ones are those with Bluetooth capabilities.
The Robot learns the orientation of the Bluetooth-enabled cube by twisting a couple of its faces. Afterward, it can solve the cube regardless of its orientation.
An app is paired with both the cube and the robot via Bluetooth, coordinating the communication between the cube and the robot.
This is a nice build video. Covers both the hardware and software, and the bumps in the road along the way to a successful build. Congrats!
Like the MIT solution, I question the fairness of modifying the cube to accept direct stepper motor drive of the six faces. As well as the method for inputting the scrambled cubes initial orientation. Does Guinness have a limit on just how many modifications can be made for machine-to-machine competitions?
Such detours echo IBM’s Watson on Jeopardy. Seemed pretty impressive, until you realize Watson was not in fair competition against the humans. Humans were “reading” or “hearing” the questions in real time, whereas the “text” of the questions were complied ahead of time and fed to Watson directly.
Awesome build!