[SOLVED] Speed up the brickuber

brickuber

#1

Is there anyway to speed up the rubik’s cube solving time using the brickuber?


#3

Hi @OBIDINZERIBE1,

Without investigating too much, I think you can improve it by having a better algorithm that does it in fewer steps. But then, if I look over its Github page, I can see it’s using the Kociemba algorithm, which is already pretty state of the art as far as I know.

My question is: how fast do you want to have it solved?

Thank you!


#4

Hello, thanks for your answer. I was thinking about solving the cube under a minute


#5

Hi @OBIDINZERIBE1,

At the moment, reading all faces of the cube takes roughly half a minute and for actually solving it, you need somewhere around 50-60 seconds, depending on the cube’s initial state.

All in all, I think the algorithm is already maxed out as I don’t think there’s another one that does it in fewer steps than the Kociemba.
As for the hardware, you’d be better off with having control over the rest of faces too as that helps you skip many unnecessary movements - but this implies a radical change in how the solver is built and that can take lots of time.

The most efficient thing to do would be to reduce the time needed to scan the cube - theoretically speaking, you could have 3 cameras that point to different faces in such a way that you only need to rotate the cube once to have all the faces read (or determined). That should cut the time down by quite a lot, but it’s never going to be more than those 30 seconds.

If you succeed in making the scanning process faster by using multiple cameras, then you could see the cube being solved in 55-65 seconds - that’s a rough number.

You could also put a single camera in an isometric view so that it can see 3 faces of the cube at once - but then you’d need a slightly more complex recognition algorithm.

Either way, I think solving the cube in under a minute is possible, but you need to put in lots of effort - personally, I don’t think it’s justifiable.

Thank you!

EDIT: I don’t mean to discourage you, it’s just that it’s important to note what a change of this magnitude implies.


#6

IIUC, Kociemba’ algorithm is sort of A* which calculates the path to the well-ordered cube faces during runtime (CMIIW).
OTOH, David Gilday’s Mindcuber uses huge look-up tables which already have been pre-calculated offline before on an independent PC, and then beeing just copied to the Lego PB: that makes it able to order several subsets of faces in just 1 single move - which finally is supposed to make things much quicker.
I personaly prefer Kociemba’s way because there is still some kind of AI behind it, even if it’s solving times are much slower compared to “stupid” pre-solved lookup tables.