[Unofficial] CubeStormer II beats own record live at ARM TechCon

[IAssemble]

While Mike and I were demonstrating CubeStormer II at ARM TechCon 2011 it achieved a new "personal best" of 4.762s

This time is the total solve time including scan, software solve and mechanical solve.

The times and solutions for all solves at the show were recorded. For this solve, CubeStormer II found the following 21 move solution:

U2 D L' B R F B2 L' U D2 L' D' U' R' U2 R D U2 B U' B2

 

[qqwref]

Plenty of people have already beaten this time in unofficial practice (which this counts as, not that human vs robot are really comparable). Stop being noobs.

IAssemble: Very nice stuff, your robot is really fast compared to previous ones. I'm still wondering what kind of average times you get - even if you don't want to film it, would you mind writing down the times of 12 consecutive runs on typical scrambles?

 

[IAssemble]

Plenty of people have already beaten this time in unofficial practice (which this counts as, not that human vs robot are really comparable). Stop being noobs.

IAssemble: Very nice stuff, your robot is really fast compared to previous ones. I'm still wondering what kind of average times you get - even if you don't want to film it, would you mind writing down the times of 12 consecutive runs on typical scrambles?

Thanks.

I mentioned in a post on another thread that Mike and I recently demonstrated CubeStormer II live at ARM TechCon 2011 in Santa Clara.

The software recorded all the solve times and solution sequences during this two day event where the public scrambled most of the cubes. The software excluded any solutions fewer than 18 moves that would suggest that the cube had not been scrambled well.

For the 200 or so solves of 18 or more moves achieved at the event, the average time was just under 5.9s with an average solve length of just under 21 moves.

 

[RoboticSolution]

Cubestormer II's average is less than 21 moves, but not by much, great vid of a sub 5 second human solve though, totally awesome, its a shame that it took so long to inspect it first, more like a 15 second solve if you ask me !

 

[aaronb]

Wow, a sub-5 BLD solve is almost unheard of! And to have all those fast BLD solves without a DNF!

This is incredible. I would have loved to been there to see it.

 

[IAssemble]

Plenty of people have already beaten this time in unofficial practice (which this counts as, not that human vs robot are really comparable). Stop being noobs.

IAssemble: Very nice stuff, your robot is really fast compared to previous ones. I'm still wondering what kind of average times you get - even if you don't want to film it, would you mind writing down the times of 12 consecutive runs on typical scrambles?

I just extracted a series of 12 consecutive solves for you from a random position in the log recorded at TechCon (after eliminating any solutions fewer than 18 moves). I think the format should be compatible with the way CubeExplorer reads files.

U2 F' U' F B' U F B' U2 F' R2 L D L2 D' B2 U B2 R' F2 (20f,5374ms)
U R D' F B2 L D' R B2 U' L' F' L' U' F U2 F' U2 F' U2 (20f,5694ms)
U2 F2 L' F B U F2 B' D' F' U2 R2 U' R U F' D' F D F' R2 F2 (22f,5602ms)
B2 L2 B U2 F' U D L2 F2 R2 F R' D' R' D2 R2 D' R' D (19f,5051ms)
U F2 U R' L2 B2 L2 F2 L D' L2 U' D B U' B' L B L' U2 R D (22f,5594ms)
L U2 F B2 U2 R' U D' B' U2 R' F L D' L D F2 U F R2 B2 (21f,6030ms)
F2 U R' D' R' L F2 D2 R U D' B D' B' D' F D' F' U2 F' (20f,5729ms)
F2 L2 U B2 R' F D' L F2 L' B R2 F U' R U R' F' R2 F R' F' (22f,6937ms)
D L2 B2 R B' U D2 B L B2 D2 R2 D' R2 B' R2 B D R B2 (20f,6409ms)
R2 L' U2 F2 D2 B' R' D L' U D2 L D2 L' D F' D' F D' L2 D (21f,5392ms)
U' R B' D R2 D' L U2 R F' U F U' R' U2 R U' R' (18f,5073ms)
U2 L2 D' R' F2 B' D2 B D2 R B' L F' L2 U' L F2 U2 F' U2 F2 U' (22f,6004ms)

Hopefully it is obvious that the times are in milliseconds after the number of face turns.

 

[RoboticSolution]

Cubestormer II is capable of producing a sub 5 BLD on a daily basis, maybe next time we are in The USA you could come along to see it?

 

[Mike Hughey]

I just extracted a series of 12 consecutive solves for you from a random position in the log recorded at TechCon (after eliminating any solutions fewer than 18 moves). I think the format should be compatible with the way CubeExplorer reads files.

U2 F' U' F B' U F B' U2 F' R2 L D L2 D' B2 U B2 R' F2 (20f,5374ms)
U R D' F B2 L D' R B2 U' L' F' L' U' F U2 F' U2 F' U2 (20f,5694ms)
U2 F2 L' F B U F2 B' D' F' U2 R2 U' R U F' D' F D F' R2 F2 (22f,5602ms)
B2 L2 B U2 F' U D L2 F2 R2 F R' D' R' D2 R2 D' R' D (19f,5051ms)
U F2 U R' L2 B2 L2 F2 L D' L2 U' D B U' B' L B L' U2 R D (22f,5594ms)
L U2 F B2 U2 R' U D' B' U2 R' F L D' L D F2 U F R2 B2 (21f,6030ms)
F2 U R' D' R' L F2 D2 R U D' B D' B' D' F D' F' U2 F' (20f,5729ms)
F2 L2 U B2 R' F D' L F2 L' B R2 F U' R U R' F' R2 F R' F' (22f,6937ms)
D L2 B2 R B' U D2 B L B2 D2 R2 D' R2 B' R2 B D R B2 (20f,6409ms)
R2 L' U2 F2 D2 B' R' D L' U D2 L D2 L' D F' D' F D' L2 D (21f,5392ms)
U' R B' D R2 D' L U2 R F' U F U' R' U2 R U' R' (18f,5073ms)
U2 L2 D' R' F2 B' D2 B D2 R B' L F' L2 U' L F2 U2 F' U2 F2 U' (22f,6004ms)

Hopefully it is obvious that the times are in milliseconds after the number of face turns.

So that's a 5.69 average 10/12. And note that these are probably slightly harder than typical scrambles, since he threw out solutions fewer than 18 moves. In my experience, three or four scrambles out of twelve are fewer than 18 moves optimally. (I optimize a bunch of scrambles every week for the weekly competitions.)

 

[IAssemble]

So that's a 5.69 average 10/12. And note that these are probably slightly harder than typical scrambles, since he threw out solutions fewer than 18 moves. In my experience, three or four scrambles out of twelve are fewer than 18 moves optimally. (I optimize a bunch of scrambles every week for the weekly competitions.)

To clarify, the software removed any solutions that CubeStormer II found that were shorter than 18 moves. This doesn't necessarily mean that the optimal solution was less than 18.

 

[qqwref]

Well, it does mean the optimal solution was less than 18, but the other way around isn't true (optimal solution of <18 doesn't mean the software will necessarily find a solution of <18).

 

[IAssemble]

Well, it does mean the optimal solution was less than 18, but the other way around isn't true (optimal solution of <18 doesn't mean the software will necessarily find a solution of <18).

I was actually trying to say that the solutions that were eliminated were those which CubeStormer II solved fewer than 18 moves rather than all those which had optimal solutions less than 18. As a result, there may be some solutions remaining for which CubeStormer II found solutions of 18 or more moves but in fact may have optimal solutions less than 18 moves. Does that make sense now?

Perhaps someone would like to find the optimal solutions for the solves recorded above?

 

[Stefan]

Cube Explorer found 17 moves for scrambles 4 and 5, I stopped the others at depth 18.

 

[IAssemble]

Cube Explorer found 17 moves for scrambles 4 and 5, I stopped the others at depth 18.

Thanks Stefan. It sounds like these were scrambled ok then!

 

[Tim Major]

Um... what? The advance from the last WR robot solver is crazy. That TPS O_O
It's a shame we can't see much, or the very start. How long did it take to inspect? And was it counted? O_O

 

[ben1996123]

Um... what? The advance from the last WR robot solver is crazy. That TPS O_O
It's a shame we can't see much, or the very start. How long did it take to inspect? And was it counted? O_O

Inspection is just over 1 second, so the solving was probably like 3.4ish seconds. Did you see the other video?

 

[Florian]

They should make a roboter which solves the cube with WCA-rules