Similar to all other 6-axis robots, SquidCuber solves a cube with slight modifications (in this case a small Lego plate glued onto each center). However, recently I have also made Cuboth, a super fast Lego robot that solves a cube without any such modifications: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Kjb-MmwueEQ.html and it is in fact currently the world's fastest robot to do so (even counting non-Lego machines)!
Thank you! Yes, the idea of this project (as well as my previous robots) was to see how far Lego can pushed. While you of course cannot compete with the most recent overall record holders that use top of the line professional industry motors (which are orders of magnitude more powerful than the Lego ones), I am myself still a bit surprised how far one can actually get with "toy" hardware just by meticulous optimization.
@@bigsmall246 yeah Lego really makes the mark there as opposed to other ripoffs who don't go for the same amount of durability with individual bricks so major shout-out to Lego for doing that mad making our childhood infinitely better
@@EliasFrantar At first I was like "sub1 did that like years ago wdym fastest Rubik's Cube solving robot" And then I read the title again That's awesome
Thank you! Yes, given the speed of modern solvers (now even of the ones made with just Lego) I think the average is really what should matter (thus I also only put this solver's lucky PB at the very end of the video). Unfortunately, there are only videos of a single solve for most other really fast solvers (Sub 1 Reloaded, MIT robot, Master P) making it quite hard to judge their actual speed (they are certainly faster than SquidCuber, and they better be using much more powerful motors, but I suspect that their average is also quite a bit higher than the solve we see). Because of this I think SquidCuber might actually have the fastest (inofficial) average on camera by any robot in general?
Thanks! Yes, the video is getting a lot of views recently (and views are also slowly but steadily increasing on my newest robot). I am very happy that so many people are now (after almost a year) watching my projects that I have put so much time into :)
Wow. I make a lot of Lego Technic designs but this is by far the most amazing feat of Lego engineering and programming I have ever seen. This is a truly inspiring piece of work and achievement that can't be over rated. You are a genius.
Thank you! I am also a bit surprised that I managed to make it that fast, I just never stopped optimizing, always tried out new tricks and eventually (over the course of 3 robot iterations) reached the time you see in this video. If you had asked me a year ago I would have probably told you that a 1 second average with Lego is essentially impossible (at that time the best Lego solver was sitting at 2.2 average and already looked pretty much like the limit), yet here we are. :)
I can see a way to do a 5x5x5. However, while all odd numbers have an easy anchor non moving point, the middle of the face, all even numbers do not! Don't say it's not feasible for even edges, just the approach has to be different!
@@olivierleynaud7841 how would you do a 5x5? Assuming the anchor points are in the center, you could only rotate the edges. The center 3x3 of the 5x5 would be untouched.
@@wolflordy3193 Two ways, the first and fastest is two concentric axles (an axle inside a pipe), the first one digging through the first layer to grab the second. However not all cubes are the same build, not sur it will work for all. Second, far slower, grabbing second layer with a retractable and rotatable two or four digits grabber. It will be slower as it would nees to open fully to let the next move happen. It will also be mechanically more fragile.
The fact it's made from legos makes this 100× cooler! Have you managed to get monetised and recieve ad revenue for all those views?! Cuz you 100% DESERVE it.
Would it be possible to make a 4x4 solver with legos, I know middle layers will be pain but you can be the first one, maybe you can make fast popping sticks that can turn these.
A (fast) Lego 4x4x4 solver is actually my next project, which is planned for this year :) The 6-axis mechanism of SquidCuber is going to be very difficult to adapt for handling a 4x4x4 cube. I think extending the mechanism used by my newest robot Cuboth (that solves 3x3x3 cubes without any center modifications) is a lot more promising, but you will have to stay tuned to see what exactly I am planning! ;)
Question: does the fact that every starting scenario has the correct color in the middle of the face influence the difficulty of the solving algorithm?
Not at all. The solving algorithm does not use the colors red, blue, green, etc. but only treats them as right, left, up, and so on. This means, whatever color the center on the right face is, it will always be called R. Note also that the 6 centers of a cube can never move (you can check this on your own cube), which is why this 6 axis design actually works in the first place.
Hi Elias, I am from Science Centre Singapore and I would like to reach out to you with regards to SquidCuber. We are developing an exhibition on Robotics and while we were researching on what to showcase, we came across this video and would really love to showcase your SquidCuber. Is there any way we can discuss further?
It uses "rob-twophase" github.com/efrantar/rob-twophase, which is my own extremely efficient implementation of Herbert Kociemba's two-phase algorithm further adapted to take the robot mechanics (i.e. turns on opposite faces can be performed in parallel, a 180 degree turn takes about twice as long as a 90 degree one) directly into account while searching for a solution. This gives about 20% faster to execute solutions on average. Additionally, my solver returns more than 1 solution so that I can then also select the fastest one based on data about the timing of turn transition that I collected from prior solves. This solving algorithm is one of the keys to the exceptional performance of the robot.
@@ChrisLee-yr7tz well if we're arguing semantics then everything is real life. I'm asking what use it has outside of what it is currently demonstrated here, that is speedcubing, that we see out there? Fancy looking gears beyond cog wheels find itself in car engines' Reuleaux Triangle, so what other applications is there for this contraption that requires high speed calculation and quick, precise, physical adjustments?
I've seen real robots do better, but I can understand why I shouldn't compare. This is a fantastic machine! I'm a LEGO fan and a cuber myself and amazed to see these both working together! Hope you can do more in the future!
That's true, I also think this is a problem, especially since most other really fast robots (on RU-vid) show you exactly one solve (presumably the luckiest one they have ever gotten). This is why I always show multiple random solves in my videos and also consider the speed of my machine in terms of average solving time (i.e. ~1 second here, instead of the ~0.75 best solve; similarly ~1.9 seconds average for my newest robot that solves cubes without center modifications).
@@EliasFrantar BTW, dont take my comment as a way of hitting you, I think what you're doing is great and big brain, I was just talking about actually people world records.
From an engineering standpoint, friction and inertia would be the most difficult obstacles to overcome. I do see that the cube is modified for this reason.
My 8 year old learned to follow a technique to solve it. He is pretty quick now just from memory. He is developing his own variations I'm sure. This will blow his mind