I Built a Rubik's Cube Solving Robot

[@3DprintedLife]

This is sickkk. Awesome video! Hope you get the views this deserves :D

[@aaedmusa]

I love your videos dude! Thanks!

[@RWBHere]

@@aaedmusa I cannot believe that, at the time of typing this, you have 107,000 subscribers but this video has only 58,000 views and an even smaller 4,000 likes. Subbed, viewed and liked.

[@niil047]

insanely underrated channel

[@TheCharlatanWizard]

No cap

[@hajdaniel5778]

Fr

[@katunu6030]

Definitely. This channel needs more views.

[@Sgt_Church]

100%

[@mr.fluffypantz4150]

I bet you could modify the solver algorithm to allow opposite motors to turn at the same time. the right and left motors could turn in 4 different permutations at the same time, and every time they get used instead of a single motor turn, you chop off a move. Given that most algorithms allow for face turns and not just quarter turns, youd be adding a huge number of possible moves. I bet that would reduce the number of needed turns from 20 to ~16-18, but that's pure conjecture. If this doesn't work with the algorithm you used here, I'd love to see you use the original meet in the middle algorithm using these modified moves (even if just once since you'd have to wait on the solve for a long time) to show the fastest possible solve your machine could do.

[@aaedmusa]

Yep I've had my eye on an algorithm that someone made that does exactly that. I'll be exploring this in the future. With the standard kociemba algorithm there are few opportunities to pair moves in such a way that they can occur simultaneously.

[@Internetzspacezshipz]

This is the exact first thought I had when looking at this. I wasn’t sure if it was already implemented or not because the things spin so darn quick hahah.

[@mgancarzjr]

I came to this as well. Assuming you have an algorithm which scans your moves list and determines which adjacent moves can be "combined", the question is whether the controller can either a) send signals in parallel or b) get a second motor to start moving before the first one finishes. I don't know that the API for the motor controllers looks like or if they were programmed with raw high-low signaling. Ideally, you could just send both a message to turn to their respective rotations and simply ignore any kind of delay between the two signals.

[@davidtrott9469]

The other thought I had is that, all solves have Red on the top - since you don't have any middle squares you could have any color at the top - but it is kinda cheating because its only possible because you have removed the middle squares.

[@jiri-one]

LED strips inside could help with consistent lighting. What a great video though, good job! :)

[@TinyMaths]

Bro... you are a MADMAN! And I mean that as the highest form of compliment. Pretty impressive results. Honestly, I thought that, if someone knows how to solve the Rubik's cube, and they are halfway decent at maths, they could probably (after many headaches) come up with a very crude, really inefficient set of rules and logic for solving the cube, and pat themselves on the back; but then when you started talking about the colour recognition for the software, I was like, 'damn, one really does have to be a little bit crazy to take on a project like this' 😃

[@cheplays2482]

Absolutely love your channel. The way you challenge yourself with these builds is amazing to watch.

[@nirkarl]

Great video! I've made a similar project a few years ago, and i got to the same conclusions when it came to identifying the color of the facets. I did notice that switching to HSV as some of the other comments suggested helped in color identification.

[@billstrahan4791]

Yep, I've totally gotten hooked on your videos. If you ever need some parts machined for your projects (if 3d printing just won't do) let me know. I'd be happy to lend a hand. I'm in the Dallas area, but I can ship parts.

[@noahstockwell1408]

Awesome project and super well edited! Boiler up!

[@aaedmusa]

How fast can you solve a Rubik's Cube?

[@StormBurnX]

My current top time is less than one cube per year 😂

[@ljakeupl]

last time I tried and failed was probably when I was 10, so assuming the timer is still going, 24 years and still counting!

[@jorsanflo]

About 36 secs

[@damianomotta1809]

My record is 21 seconds i rarely go above 40. Anyway amazing project

[@darkgobelin4439]

my pb is 33 but I average on 36

[@Internetzspacezshipz]

Tip for solving the colour problem; instead of trying to find the same colours every time, instead look at the relative values between them. That way you can identify which colour is which based simply on the difference rather than having to essentially “hardcode” a value. Also as other people have mentioned, switching into HSV colourspace could be really helpful too! Especially if you were using the relative colouring, since you could really just look at H and V. Eg, the red is a lower V than orange.

[@xanderlehnardt3673]

This is awesome! I remember a rubiks cube solver project that I saw for the old lego robot kits, it was really cool to be able to build it and have it solve a cube, even if it was really slow. You should think about trying to tackle making a robotic solver for larger cubes, like 4x4 or 5x5, would probably be orders of magnitude more difficult but I'm sure you would find a way.

[@Mrbobinge]

OMG, not just remarkable ingenuity but the matching learned skills adapted to implement them in so many disciplines. Right down to screw lengths and friction reduction without WD40 or even the lady's moisturiser cream. Huge congratulations to a huge young brain.

[@RitooV_IITB]

Really appreciate your efforts in making this project (and solving the challenges). Controlling the BLDC so precisely is great and combining the robotics with one of the most fascinating puzzles of all time is really inspiring. Congrats!

[@timothysands5537]

Aaed, I am in love with your videos dude. FOC controllers look so much better than using stepper motors with stepper motor drivers. I'm definitely making the switch for my robot.

[@JuliusBurton]

This is great!

[@m41nd3]

Leaving a comment mainly to incentivise the YT algorithm into promoting your work, your videos are amazing and inspiring and more people should see them

[@arnemartt]

Amazing work, instant follow

[@joshdotson2190]

Wow, incredible job from start to finish! Have you considered using the HSV color space for your detection algorithm instead of RGB? It is far more common in computer vision and I think you would find it much easier to tune.

[@Leo-ir6ey]

Great video! The cost of 12 ODRIVE S1 is eye watering (am I seeing this right: $150 per controller??)! I am currently working on a project that uses FOC as well and have found the ST B-G431B-ESC1 ($20 per unit), which does not come with many features pre-programmed, but you can program the STM32G431CB, which is on the board, to do any controls you want. It unfortunately clocks the charge pump at the motor bridge switching frequency, which might cause some problems when operating at low speeds.

[@efivip93]

With FOC, you are using PWM, so the charge pump is not a problem at all. But you can't do 100% duty cycle though, which is only a problem if you want to have full torque at a very specific rotor position, and the rotor is stuck, but those cases are highly unlikely, like extremly unlikely to cause troubble...

[@Dect.]

This video deserves way more views, keep it up!!

[@MakeDataUseful]

Fantastic video, thanks for sharing!

[@martyschrader]

Nice to know that there are enginerds who can fully geek out on this kinda solution exploration. You've got an embedded systems gig waiting for you upon graduation, my man. Parsec can't afford you, but somebody can.

[@theoneandonly2025]

This is so awesome. I also built a Rubiks cube solver, that could solve cubes in under 0.7 seconds. I did it with cheap normal Nema 17 Stepper Motors, but I really like the idea of using drone motors. As others have suggested there is a different version of the kociemba algorithm that is optimized for solving machines wich turns 2 sides at once, which reduces the amount of steps needed. For me corner cutting worked great to a degree. I did not tension the cube that hard and was able to use the corner cutting during the acceleration/deceleration phase, which decreased the time by a lot. Maybe try just a bit of corner cutting and slowly step it up till you reach the limit. Especially the early phases of the acceleration is where you can get the biggest time benefit.

[@aaedmusa]

With cheap Nema 17’s? That’s almost world record time. How many moves were your under 0.7 second solves?

[@theoneandonly2025]

@@aaedmusa Yes, I did a lot of tuning and the motors/drivers where running right at their limits xD I burned through a LOT of the drivers. Because I used the specific algorithm, it was about 17-19 steps the machine had to do. (When counting the double moves as one)

[@aaedmusa]

@@theoneandonly2025 Have you posted this project anywhere that I can take a look?

[@theoneandonly2025]

@@aaedmusa I only have the earlier prototypes on my Instagram, but have never come around to post the final version. I really should do it xD. I wrote documentation for it, but it is completely in German.

[@aaedmusa]

@@theoneandonly2025 I would love to see it if you wouldn't mind. Also, if you could share the algorithm that you used that would be great help for my next iteration.

[@JaisonBuilds]

Love your stuff bro 🔥

[@jilliebean613]

Really impressive project! It’s sick you were able to get it to work so clean. Boiler up!

[@kaestar5]

Yeah this one definitely deserves a like.. very nice!

[@AgeofRobots]

The best creator on KZbin at the moment! Absolutely love your content, specially the latest stuff.

[@enricodesign619]

wonderful project, i love it, Im recently begun learning to code Arduino and seeing a world of options open for me, and being a mechanical engineer i know i can build things, and now make them move, like you are doing here, very well done, love the challenge to yourself, that's how we all learn and move forward

[@aaedmusa]

100% agree!

[@the.dataloop]

Another great video. Thank you for this.

[@GreenStarbird]

This video and project are amazing. Good job!

[@suomi35]

My favorite video of the week, cheers Aaed!

[@James68257]

This is awesome! Great work!

[@bishopdredd5349]

This is some amazing work and a great video!

[@810mb3Rg]

Dope video. 👌 If color/light is still a problem, might be worth trying a different color space than RGB, such as HSV or LAB for image processing

[@jeffkenny7060]

Very nice work! I'll be showing it to every kid I know to get them interested in engineering.

[@holzwurm_hd7029]

How fast of a drone could you build with 6 of these motors (on top of and below) a crbon fiber body if you were to keep the Weight wayyyy wayyy down low?

[@aaedmusa]

No idea. I'm not really into aviation. I never intend to use these drone motors to actually make a drone lol.

[@esser50k]

awesome results! use of the DC motors was a good choice

[@douglass.4487]

Hi! I have a similar project that i'm developing, and I came up with an idea that may help you solve problems with color identification. The thing is, you dont need to identify if a facelet is red or orange, but just what facelet has the same color as the other. This way, you'll be identifying color groups, and not the exact color. To give the cube's pattern to the kociemba's algorithm, pick a corner and assign it's color groups to white, green and red. Iterate over the other corners, checking if it has 2 color groups matching the first corner. If it does, then the third unmaching color groups are opposite colors. E.g: the first corner has groups 1,2 and 4, and I assigned 1=White, 2=Green, 4=Red. If the second corner has groups 1, 4 and 3, then color groups 2 and 3 are opposites. Since 2 is Green, then 3 is Blue. The kociemba's algorithm will work the same, since you'll be giving it the same cube, just with a different simmetry! Hope it helps you get the WR 😂 (who knows? haha)

[@thej3799]

Dude those speed controllers you use are so cool. i didn't know anything about them before i saw your videos. I'd only ever heard of stepping motors for these purposes. using brushless is amazing.

[@roundcheesewheel]

Top notch content my dude! More please!

[@StormBurnX]

Fantastic video, as always! I wonder, since you are using the o-drives, if you could use current sensing to determine when the cube has been jammed and immediately release the motors to reduce damage/strain on the cube.

[@aaedmusa]

For me, it's easier to shut everything off by hand than to program it to shut off. Once it's tuned properly jamming isn't an issue.

[@StormBurnX]

@@aaedmusa Fair! I wasn't sure at first but upon rewatching the video I think I misunderstood - I thought you were destroying multiple cubes and had to replace them when they jammed or loosened. Being able to shut it off by hand and fix the cube is definitely better!

[@sachasmart7139]

Incredible. Love this content

[@swegfesh]

Quality video, quality content, quality personality. Baited me with the cube and hooked me with my love for engineering. Subscribed!

[@jacksonvandenberghe9420]

amazing work as always

[@timothysands5537]

Your videos are more interesting and detailed than James Burton. I am definitely subscribed for life my man 🎉. I know making these projects and assembling these videos takes a massive amount of time, so let me support you by sponsoring you a bit of cash, add a Thank$ button or something!

[@chriskaprys]

Just found your channel today and I'm hooked. Incredible projects, and I love your delivery (you've a knack for choice moments of deadpan humour that make me literally laugh out loud). ps: next time for removing sharpie, you might try acetone (nail polish remover), especially in a scenario like this where you wouldn't be removing the paint/ink of the underlying material. And in certain situations you can use dry-erase marker over the permanent, then wipe away the dry-erase. Weird but true.👍

[@Roughmail-p5d]

this is so good i really like it. can you build another version of it but inner one like the motors and all are inside the cube

[@lucahermann3040]

14:32 You could put two colored rings on every shaft so there are always all six colors on camera for reference. That way, recognizing the colors doesn't depend on the lighting.

[@walkingpizza1796]

Vert well made video! thank you for sharing you process with us! really cool!

[@conorstewart2214]

Rather than increasing infill for things like the shafts you should try increasing the wall thickness, the further the material is out from the centre of rotation or bending the more strength it will add. You should also consider using a different material for printing parts, PLA is good and works well for a lot of things but there are much stronger plastics you can get.

[@oliversnow]

FACEs. FACEts. Click. Never made that connection before. This channel is a hidden gem.

[@KD-my6lo]

You are a Genius 👍

[@bborud]

Great work!

[@mr.stratholm4999]

Here is a pro tip. If you want to remove sharpie ink all you need is a dry erase marker. Just write over the sharpie ink with the dry erase then simply wipe it clean with a paper towel immediately after and it'll come right off.

[@aaedmusa]

interesting! never tried that

[@mr.stratholm4999]

@@aaedmusa Works every time. That's how I used to have a whiteboard for house chores on a permanent grid. Make grid and column labels in sharpie and then write the cells in dry erase. That way the grid is permanent but if you need to redraw the grid the write over the sharpie ink with dry erase and wipe it off the whiteboard. Enjoy!

[@docfarl]

You give us hope for the next generation, young man. Keep at it! Checking my subscription status right after this comment!

[@awnyx6799]

this is really cool. i was wondering if the black and white problem might be due to an overflow error with the RGB values. idk, still this is so inspiring

[@hypnolobster]

I've never subscribed to a channel so fast.

[@emanuelnolasco3745]

I can not describe how awesome is this project, INCREDIBLE! It is really amazing how a simple game for humans can be so difficult when we are looking for an automate robotic solution since de mechanical design process, best selection of hardware, implementation, programming the algorithm and finally tunning it is awemose. I wonder how much time does it take and what was the total amount on money spend on developing this project!

[@AboutTheBuild]

Awesome project good job.

[@MukweLaif]

Wow, excellent video as always. I wish more people knew about this channel

[@bartoszgaj7788]

Great project!

[@rohithmulampaka4769]

I subscribed your channel your effort is precious

[@Miner13-rk8qx]

this is insane! great video!

[@lavericklavericklave]

i just came across your channel, you should be famous! great videos and projects! keep it up :)

[@insanecomicdude]

If you programmed it to randomly scramble the cube, pause, and solve it over and over again, this would make for one hell of a cool art installation.

[@efivip93]

Just a quick tip. Use inrunner motors with long, but small diameter rotor to keep the inertia as low as possible. Lubricate the cube as well with some low viscosity grease or oil to keep the friction down, needing for even less power from the motor, further reducing the innertia. That way you can increase the acceleration/deceleration. Nice build though. ;)

[@aaedmusa]

I already had these motors from a previous project. At high speeds, high friction is actually preferred not to mention that the cube can be tensioned even if friction is a problem. Thanks though!

[@amogusenjoyer]

10:25 BAHAHAHA. The humour in this video was on point lmfao

[@IntenseGrid]

Very nice engineering. Got a link to those motors? I can't seem to find the 90KV version.

[@ohmundazolen7157]

You are the best, keep going my friend

[@BleughBleugh]

Gained a subscriber here! Algorithm’s being great tonight! Fantastic video, fantastic production quality, content and pacing… Thankyou.

[@ememeable]

The equation variables if I should guess would depend on the following: Motor speed Osulation speed or frequency speed of the microchip or refresh rate This unlike human variables is immuned to random complexities like a calculated. It takes a calculated almost the same amount of time to add 1+1 and 1*4+247=251 Sincerely Ejike

[@k4yd33yeah]

It would be cool to know how you scramble the cube each time. Great video dude!

[@abcqer555]

Your channel is awesome!

[@adicsbtw]

For the vision section, I highly recommend using a colorspace other than RGB I've used CIE Lab in the past to great success, and others have used YCrCb, HSV, and HSL to great effect as well these colorspaces are far less likely to be thrown off by lighting variations than RGB

[@donacento3291]

Thx for this awesome vid. I just have a technical question: Have you tried to convert the RGB values to HSV and to determent the color just by the hue value?

[@Natrlbrnkila]

Amazing work! I know nothing about Rubik's Cube solving; but could you rotate two opposite sides of the cube simultaneously to increase solve time?

[@aaedmusa]

Yes, but that can only happen under specific circumstances: when two sequential moves are on opposite faces. With this algorithm, it doesn't happen too often. Generally, sequential moves are on adjacent faces meaning that they have to occur independently rather than simultaneously.

[@khoda81]

Dude, so underrated. Your gonna blow up soon

[@jalenh371]

Awesome video dude. Keep it up :)

[@girrrrrrr2]

the way I like to set heat insert nuts is to heat it with my soldering iron (i have the special tip which does help inserting accurately) and then ill push it like 3/4 of the way in, and then press it in the rest of the way with something flat. Comes out almost perfectly every time.

[@mqb3gofjzkko7nzx38]

I really love the clicky sound it makes.

[@chillchilli2671]

How do you not have 1M subs dude, this is inspirational content!

[@Komeuppance]

Is facet "face-et" or "fass-et"?

[@aaedmusa]

Spelled facet, but I’ve been told that I mispronounced it.

[@bodeneinheit]

Great video!

[@JakeHaugen]

Such a great channel. Instant sub! Amazing project. Would love a deeper discussion on the algorithm!

[@janosadelsberger]

For detecting the color HSB makes much more sense because you’re independent of brightness and saturation. Also the stiffness of the brackets might be a bit low so they the motors move out of alignment quite a bit. Other than that cool project 😊

[@aaedmusa]

Thanks for the input. I have had a lot of people recommend HSB. Unfortunately the cameras that I use don’t support HSB. RGB is doable with consistent lighting. As for the brackets are you referring to the brackets that the motors are mounted onto? If so those are very stiff. There are 8 screws on each.

[@JulianMakes]

Awesome build! I’m a speedcuber (but slow!). You can buy your own custom stickers :)

[@willvolovich]

Get that bag! Sponsored already🤯

[@account-r3w]

I’m thinking of an algorithm which instead of solving cube in smallest amount of turns would give us best solution for using less motors. Meaning these turns would kinda connect together to be more smooth. + combining it with corner cutting and you could make it maybe even faster. Also you could turn opposite sides at the same time

[@jeepxj]

Can you gear up the motors so they use their torque for speed

[@aaedmusa]

Yes, this is possible. I'll be looking into this in a future video.

[@jeepxj]

keep going: put it in a vacuum chamber so there is no air resistance... @@aaedmusa

[@aamiddel8646]

Very interesting. Is there a need to test the position of the facets after every move? Just scan the initial positions and calculate the new positions. And maybe if you do the calculation during the move in a separate thread of the processor (if your processor has this capability) you gain some time.

[@aaedmusa]

It only scans the cube before a solve. When the motors start turning the cameras are not used.

[@julius6429]

Just for the algorithm. You definitely deserve it

[@tvctorj]

Well done!!!

[@seanhughes7167]

I've never heard facet pronounced that way but in the context of a sub component of a face it sounds right

[@JadenKirschner]

Great Videos. Keep it up man!

[@oM477o]

You probably already thought of it but one possible optimisation is making use of the fact that you can turn opposite motors at the same time without the cube jamming. To make best use of the optimisation the solver should consider rotating opposing sides at the same time as a single move.

[@michalsuchanek322]

The pixy confuses black with white because it's transforming the colors into HSV space and black and white are just varying degrees of saturation of the same (arbitrary) hue

[@carthagely122]

I think you have excellent skills good luck

[@trystanfrease]

How are you not more popular. Your content is pretty close to Stuff Made Here. Keep up the good work! I strive to have the skills you have!

[@castelmichel8670]

Agreed 100%

[@z0mb13h0rd3]

Always love your videos! This channel IS so underrated. How about aluminum or carbon fiber shafts to minimize the blocking of the camera view? Could also lighten up the part a well as reducing the rotational inertia. Btw, some alcohol would have taken that sharpie off quite quickly.

[@aaedmusa]

I assume you mean aluminum/carbon fiber shafts that are thinner than my current shafts? That could work, but the method that I created has a 100% success rate in finding the hidden facets whereas camera vision doesn’t. It’s also cheaper than buying custom shafts. I’ll be working on lighting and inertia in the next version. I don’t have any kind of alcohol at home so scratching it just seemed to be the easiest thing to do in the moment lol. Thanks for the input!

[@z0mb13h0rd3]

@@aaedmusa of course. I was actually thinking more towards the rotational inertia originally, then started thinking about the camera to be able to shave off some of the solve time. I’m sure processor speed must also be a consideration as well as any other bottlenecks you are encountering. At the end when you were talking about the looser cubes being better for humans because they can cut the corners on the next move. Are you sure you can’t use that to your advantage in your build? I realize your stated reason for the tight cube… By the way, thanks for interacting.

[@aaedmusa]

@@z0mb13h0rd3 Yep. Corner cutting would require a loose cube. A loose cube works very poorly at high speeds. Like I mentioned, the MIT guys came to this conclusion in their testing. I also tried it out and got a lot of jams as a result.

[@z0mb13h0rd3]

@@aaedmusa right, you did say that. Slipped my mind. So much of it comes down to the places you can cut down time, in data acquisition, processing, parallel moves, and move speed. You solved it in such few moves, there doesn’t seem any gain there. Fantastic work! You have my admiration per usual. Do you have a project for winter break, or are you taking a proper break?

[@aaedmusa]

@@z0mb13h0rd3 I think there’s a few more things I can do to reduce time in terms of the algorithm and motor settings. Anything beyond that will just have to be done by changing the motors or gearing them down. I’ll be working on a robotic actuator design this winter break.