Cube Prime

[Lchu613]

Now put cameras on the robot itself. Wonder why they didn't do that in the first place, can't be that hard.

 

[Brest]

I think it's very likely they they cut a lot of that out of the video -not due to dishonesty but due to wanting to have a short video.

The video shows play being clicked with a 19f solution in the window, however it doesn't show the result of the click.

Brest, did the scramble help or did you ignore it during reconstruction?

I did not base my reconstruction on the scramble. After I had finished I realized it was the same sequence.

18f is optimal for this cube state, although I'm not sure if there is more than one 18f sequence.
Before I started I presumed that an optimal solution was used by Cube Prime.
You can see in my video that I was going to save the 19f solution, however it refreshed to 18f just before I saved.

 

[Branflakeftw]

Robot Can Solve Rubik's Cube In One Second!

I believe this is the fastest robot solve I've seen. Maybe fastest ever. I'm not totally sure. But it's definitely worth watching nonetheless

 

[AvGalen]

Invalid for a WCA solve for lots of reasons like "too much inspection", not a human, use of technical device, missing stickers, manipulated centers, etc.
The cube-explorer part of this solve is really bad. The inputting of the faces is slow and a simple "get optimal solution" is very workable nowadays.
The actual turning of this machine is WOOOOOOOOOW though. And they should have shown that much more (and in slomo) in that video.

When someone asks about "the limits of cubing" in the future we can just point to this video right?

 

[sub20cuber]

hey robot what method do you use?
full gods number which includes over 14quintillion algorithms.

 

[rokicki]

I'd use a single camera that has an equal view of three faces. Then, a few quick moves to expose all the
necessary cubies for inspection, a fast (say, 50ms) worth of two-phase algorithm which would give you a
20-length or less almost all the time (ideally optimized to permit the robot to turn opposite faces
concurrently as often as possible).

Puzzle for the theorists out there: assuming one camera that can see three adjacent faces, and a bot
like this, what is the fewest number of moves required to give you full knowledge of the cube state?
Do you need to know the color scheme of the cube to achieve this? Assume every facelet on the
three faces is visible (even the center facelet).

 

[Lchu613]

Just give it a few cameras built-in. 6 if lazy, 2 if poor, 1 +cube rotations if really poor, small light sensor to scan each cubie if completely broke.

 

[rokicki]

Two cameras each seeing three faces is sufficient (even if the arms hide far corner cubies).

Six is so overkill; it's like putting six wheels on a wheelbarrow.

 

[AmazingCuber]

Two cameras each seeing three faces is sufficient (even if the arms hide far corner cubies).

Six is so overkill; it's like putting six wheels on a wheelbarrow.

what about three cameras?

 

[Lchu613]

Who wouldn't put six wheels on a wheelbarrow? XD