They might work, but if, as I do not rule out, the machine uses dead reckoning (i.e. it doesn't use its cameras to detect when a face is oriented at exactly 90 degrees, but just rotates a face by what it knows to be 90 or 180 degrees), I think it would be difficult to properly set up the machine.
You would have to place the suction cups exactly at the center of each face, cannot allow one cup to suck a bit harder than another, and have to make sure all faces are very close to being perfectly flat in the starting position.
Edit: also, if those suction cups slip, I guess you will be looking at more Rubik's cube parts than you thought existed.
I don't think you need all that. They use stepper motors so they do indeed turn exactly 90 degrees. What they would have to worry about is the suction cup not sticking perfectly. E.g. after a 90 degree turn if the momentum makes that part of the cube slide a little when they try to stop, it will introduce a small error. Maybe turning other parts of the cube would correct it.
I've seen other Rubik's cube solvers that handle this by heavily lubricating the cube, so that if the machine attempts to turn it without perfect alignment, it'll just slip into alignment rather than snapping.
You could, and you could also make it just as fast. Increasing the speed in the drilled approach or in the gripper approach ultimately comes down to how much power you can throw at the problem (the energy required to accelerate and then decelerate the moving parts would be the limiting factor).
Its fun to imagine a situation where the level of power had been taken to such an extreme that even a slight hitch of the cube's parts would result in an explosion of tiny bits of plastic showering the room :)
I think so. It's no different from, say, painting a line down the center of the pavement to help an "autonomous car" navigate. The only reason building a robot to solve a physical Rubik's cube is interesting at all is that manipulating a real cube with something like a human hand is hard. Turning something by connecting a motor directly to is is not interesting.
I remember seeing this a couple days ago and they said they had tried not to modify the cube but weren't able to get the right amount of leverage needed to rotate it and thus insisted on drilling into each center piece. Don't think it still counts, pretty cool set up none the less.
The creators themselves are apparently rather confident that it will indeed count, at least for the guiness record. The World Cube Association doesn't have a "robot category" as far as I am aware.
As someone who solves cubes and is a "get-shit-done" engineer, I think its not as neat as you make it sound. I'm horrified by the drilled holes, but amazed at the build and engineering.
What if the cube would be manufactured so that the middle squares were removable, and replaceable with alternative ones with holes? Would that still be horrific? Or would that cube be considered easier to solve? If so, would you solve it faster vs a regular cube?
You can actually buy speed cubes which the center squares can be pulled off so that you can adjust the tension on the cube with a screw driver. It'd be pretty easy to modify the machine turn the cube with the center caps pulled off without having to actually modify the cube itself. I personally still feel like that is cheating.
I think this comes down to this; either you think the act of solving the cube is impressive, or you think the ability to turn the cube quickly is impressive. As someone who can solve the cube in 40-45 seconds (and I'm admittedly pretty slow), the part that really bogs me down is I'm not good at finger tricks or turning the cube quickly. If I had a device (other than my fingers) which would let me turn the cube more quickly, I'm sure I could probably cut my solve times down dramatically. Solving the cube itself isn't all that interesting.
This doesn't seem to be a question of pragmatism -- they're going for a world record, not trying to solve a practical problem. I'm no expert on this, but it seems to me like the criteria for obtaining a world record title would include not modifying the object used. To me this is like trying to beat the record for furthest soccer ball kick by modifying the materials used in the soccer ball.
According to the Guinness World Record's article[1] the
cube has to follow the World Cube Association's
regulations[2] on competition legal cubes. The WCA allows
modified/custom made cubes as long as they meet a list of
guidelines, which this modified cube does.
once you can modify the cube you basically break the whole point of this excersize. I mean almost all the kids i knew who messed with these when i was a younging would pull the cube apart and then put it back together in the correct shape, or take off the stickers and stick them back...
If parts pop off during a competition, you're allowed to put them back together. I'm not sure if that includes a full explosion (when a majority of the pieces come off the cube). It probably doesn't matter because it would take too long anyway.
what would be interesting is if there was a category where the robot solver has to grab the cube from the table, analyze it, solve it, and put it back on the table in the same spot, would be cool to see the compromises that would have to be made for speed of transfer/analysis vs speed of solving
My point is that the intelligence required to pick up the cube and manipulate it, might require more processing power than solving it when attached to a frame.
Yeah but that's why it would be interesting. Solving a rubics cube quickly is interesting from a human standpoint, for a purpose built machine not so much.... I mean its doubtful that the majority of the competitors wrote the software algorithm to solve the cube... What they bring is the mechanics behind it... But really if the professional robotics orgs got involved they would probably just outclass the record holders... Thing is that they arent even interested because their is No money in it and its just not that interesting a robotics challenge
Why is it always solved in the same final orientation? (red to near left, blue to near right, ...)
It would seem to me that from a truly random start point to the truly fewest-moves solution, the colors would end up facing random orientations for each run.
Human's don't typically rotate the cube either when they're speed solving. Rotating the whole cube causes you to slow down. We usually do yellow on top, white on the bottom and then you just think of the other four faces as opposing sets relative to each other (blue and green, orange and red).
Because within the cube, the centerpieces are fixed to a core. And because these define the orientation of the cube and are fixed within the robot you end up having the same orientation every time.
This happens with every solving robot that only turns the outer layers of the cube, meaning the inner layers and therefore the centerpieces do not move.
It's mathematically interesting that the optimal solutions (I assume that the robot uses the fewest-moves-possible to get its high-speed results) never require rotating the middle layers.
On closer look it's actually trivial. Any move that requires rotating a middle layer is isomorphic to rotating the top and bottom layers in the other direction. :)
Does that imply that when they remove the cube (note that it is not permanently fixed to the machine), manually twist it (i.e. to shuffle to a random state, as in the video), and replace it, it has to be returned to a certain orientation?
Again, it still solved to the same final colors orientation even after it was removed, hand-shuffled, and replaced.
My apologies if I'm seeing something that isn't there.
More like a machine, not a robot. One of the traditionally viewed differences between industrial machine automation and robot automation is that robots are relatively general and can be repurposed.
Think of wiring vs software.
I would think a real robot solving a cube would perhaps at least have to pick it up and put it back down after solving.
> robots are relatively general and can be repurposed.
I see your point, but I'm not sure that's a good definition. It's a very fuzzy line. Is a Roomba a robot? It's not a general purpose device.
I've always found it difficult to define intelligence in any useful way, but when I think about it I usually come to the conclusion that it's almost indistinguishable from adaptation, which is easier to define. And adaptation is sort of an interesting line to use for the machine/robot distinction. Machines do work, but robots adapt?
By that definition the Roomba is a robot... it will adapt to foreign objects you place in front of it. This Rubiks machine can adapt to different configurations. But there is a small finite space of scenarios it can adapt to.
So maybe that's it. How many states can you adapt to? If not very many, you're a machine. If quite a lot, you're a robot!
After frustratedly fixing an unsolvable Rubik's cube over Christmas break (my dad thought it'd be funny to "solve" it by peeling off the labels and putting them back on, but got tired halfway through and left it in an unsolvable configuration), I'd love to see what their robot does if you give it an unsolvable cube.
44 comments
[ 3.0 ms ] story [ 94.4 ms ] threadI wonder if suction cups under high vacuum could work.
You would have to place the suction cups exactly at the center of each face, cannot allow one cup to suck a bit harder than another, and have to make sure all faces are very close to being perfectly flat in the starting position.
Edit: also, if those suction cups slip, I guess you will be looking at more Rubik's cube parts than you thought existed.
See https://www.youtube.com/watch?v=X0pFZG7j5cE for an example solver.
I guess it all depends on how much force it takes to turn such a stepper motor when it isn't powered.
Its fun to imagine a situation where the level of power had been taken to such an extreme that even a slight hitch of the cube's parts would result in an explosion of tiny bits of plastic showering the room :)
https://www.reddit.com/r/Cubers/comments/42ank6/worlds_faste...
People who say that drilling the wholes in the cube diminishes the achievement, the perfect-or-nothing crowd, belong to the first group.
People who see the drilling of holes as a valid tradeoff to achieve higher performance are more pragmatic and belong to the second group.
I think this comes down to this; either you think the act of solving the cube is impressive, or you think the ability to turn the cube quickly is impressive. As someone who can solve the cube in 40-45 seconds (and I'm admittedly pretty slow), the part that really bogs me down is I'm not good at finger tricks or turning the cube quickly. If I had a device (other than my fingers) which would let me turn the cube more quickly, I'm sure I could probably cut my solve times down dramatically. Solving the cube itself isn't all that interesting.
[2] https://www.worldcubeassociation.org/regulations/#article-3-...
doesnt really count...
https://www.worldcubeassociation.org/regulations/#article-5-...
http://www.npr.org/sections/money/2015/05/19/407736307/robot...
My point is that the intelligence required to pick up the cube and manipulate it, might require more processing power than solving it when attached to a frame.
It would seem to me that from a truly random start point to the truly fewest-moves solution, the colors would end up facing random orientations for each run.
The middle piece on each side of a Rubik's Cube doesn't move. The machine rotates each side via the drilled holes in each side's middle piece.
If you're interested in learning how to solve, here's a link on how to use the Fridrich method. http://ukcubestore.com/fridrich-tutorial.html
Again, it still solved to the same final colors orientation even after it was removed, hand-shuffled, and replaced.
My apologies if I'm seeing something that isn't there.
It might be interesting to see a competition based on number of moves.
https://en.wikipedia.org/wiki/Optimal_solutions_for_Rubik%27...
EDIT: Details, history, and math of proof here:
http://www.cube20.org/qtm/
Think of wiring vs software.
I would think a real robot solving a cube would perhaps at least have to pick it up and put it back down after solving.
I see your point, but I'm not sure that's a good definition. It's a very fuzzy line. Is a Roomba a robot? It's not a general purpose device.
I've always found it difficult to define intelligence in any useful way, but when I think about it I usually come to the conclusion that it's almost indistinguishable from adaptation, which is easier to define. And adaptation is sort of an interesting line to use for the machine/robot distinction. Machines do work, but robots adapt?
By that definition the Roomba is a robot... it will adapt to foreign objects you place in front of it. This Rubiks machine can adapt to different configurations. But there is a small finite space of scenarios it can adapt to.
So maybe that's it. How many states can you adapt to? If not very many, you're a machine. If quite a lot, you're a robot!