IAssemble
Member
At LEGO World 2016, Quad-Cub3r solves the 4x4x4 cube significantly faster than the current official Guinness World Record time for a robot! ;-)
Enjoy!
Enjoy!
Last edited by a moderator:
-
Welcome to the Speedsolving.com, home of the web's largest puzzle community!
You are currently viewing our forum as a guest which gives you limited access to join discussions and access our other features.
Registration is fast, simple and absolutely free so please, join our community of 40,000+ people from around the world today!
If you are already a member, simply login to hide this message and begin participating in the community!
DGCubes
Member
Dang! That thing's insane! What method is used for solving?
Joel2274
Member
Idk but thats some insane FMC lol
IAssemble
Member
Thanks.
It uses an efficient reduction method that I developed myself for my robots.
- centres are solved in four phases including fixing edge parity
- edges are paired in groups of between 3 and 6 pairs at a time including fixing corner parity
- 3x3x3 stage uses the two-phase algorithm that I use on other robots such as CUBESTORMER 3
The method is further optimised by solving the cube multiple times from different starting orientations using different centre solutions for about 1 second on the smartphone and the shortest solution selected. The average final solution is about 51 moves.
Hssandwich
Member
What that's faster than me...
IAssemble
Member
Ollie
Member
Ok, this is amazing. Do you share your code at all? I'd like to see what your 444 solver does 
IAssemble
Member
Thanks. I suspect it outdoes humans in all stages of the solve as it solves more pieces at a time than humans can manage for both centres and edges and uses an efficient 20(ish) move solution for the final 3x3x3...
Ok, this is amazing. Do you share your code at all? I'd like to see what your 444 solver does
Thanks. I have shared my code for some of my simpler solvers such as MindCuber: http://mindcuber.com
At the moment I have no plans to share the code for this 4x4x4 algorithm.
Now you're just making us more curious (especially those who've tried 444 FMC). I'm guessing that choosing the best 'centres solution' from the 6 (or 24?) starting states is not just about the lowest move-count for that stage, but is heavily influenced by the number of ready-made dedges it creates for the subsequent stage. You could probably guarantee 3 dedges randomly made? That would open up the possibility of 20 -25 move edge-pairing.
IAssemble
Member
The algorithm finds many "good" centre solutions (with fixed edge parity) from each of 24 orientations. It does not specifically aim to pair a few dedges during this phase but does use this metric to influence whether to discard the solution attempt early in order to try more in the time available. The number of moves for the edge-pairing stage is also dependant on the combination of "cycles" of groups of un-paired dedges so this is also taken into account before starting and during the edge-pairing stage to prune "bad" solutions. It is a while since I analysed the relative stage lengths but I recall that typically centres take around 13 moves and edge-pairing around 18 in the good solutions although the number for edge-pairing probably includes some move cancellation when combined with the first stage.
Thanks for that, I was thinking the splits would be 10 moves for centres, 22 for edges, ..... sort of close.
How do you deal with parity? Adopt a specific strategy to create the wanted parity, or just generate all suitable solutions and disregard those that don't have the appropriate parity?
(I know I'm too curious, but someone else would likely ask, anyway...)
How do you deal with parity? Adopt a specific strategy to create the wanted parity, or just generate all suitable solutions and disregard those that don't have the appropriate parity?
(I know I'm too curious, but someone else would likely ask, anyway...)
IAssemble
Member
Edge parity is fixed while solving the centres by having two strategies for solving the final phase (of four) with either odd or even parity sequences. The corner parity is checked after pairing the edges and those solutions with odd parity disregarded. So both types of parity are resolved before the final 3x3x3 solve.
unsolved
Member
That is very impressive David, my congratulations.
I noticed the edges were solved with great alacrity. And the software was just using a handheld phone? Not too much processing power there.
IAssemble
Member
Thanks. I am glad you like the edge pairing. The algorithm chooses groups of edges to pair based on cycles of edges and leaves the resulting dedges wherever is the fewest moves. It attempts to solve the groups in different orders to find the combination with the shortest solution overall.
The "just a handheld phone" does have an ARM-based 2.5GHz quad-core processor with 2GB RAM and the algorithm runs on all cores for about 1 second to find a good solution so it uses a reasonable amount of processing power compared with say 50ms for the 3x3x3 in CUBESTORMER 3
Last edited:
G2013
Member
Wow, this is awesome David. This 4x4 solver is extremely efficient.
I'm sure that if you improve the turning system (to make it turn faster) and use a better cube (maybe a modified one) the times would be sub-30 or faster
I'm sure that if you improve the turning system (to make it turn faster) and use a better cube (maybe a modified one) the times would be sub-30 or faster
IAssemble
Member
G2013
Member
Watch this space!
It means you will improve on that? (Sorry, English is not my native language ;S)
IAssemble
Member
I mean that there are likely to be further developments but you will have to wait and see! In short, "hopefully, yes"
G2013
Member
I mean that there are likely to be further developments but you will have to wait and see! In short, "hopefully, yes"
Great, I'll wait and see new, faster WRs
Regards!
