How to improve the Robot Arm (MIT mini Cheetah style)

Only $7 for SMT Order: www.nextpcb.co...
Reliable Multilayer Boards Manufacturer:
www.nextpcb.com
PCB Boards 10pcs for Free:
www.nextpcb.co...
20% off - PCB Orders:
www.nextpcb.co...
Do you remember the 7 axis Robot Arm from my last video? In this video I've installed the MIT mini Cheetah controller on one of the actuators of this robot. Did it help with the vibrations? The answer is in this video.
Special thanks to my special Patrons: Wesley Corp, RobotDigg Shanghai, Pierre Freyssinet, M. Aali and Shounak Bhattacharya!
Please subscribe. This will help me to develop other projects like this, to bring the bright future closer!
One time donation:
www.paypal.me/...
If you want to help this channel, please support me on Patreon:
/ skyentific
Instagram: / skyentificinsta
Facebook: / skyentificface
Twitter: / skyentifictweet
#Robotics #RobotArm #Robot

[@Skyentific]

Do you remember the 7 axis Robot Arm from my last video? In this video I've installed the MIT mini Cheetah controller on one of the actuators of this robot. Did it help with the vibrations? The answer is in this video.

[@andrewgillies8]

Why did the new controller help so much with the vibrations?

[@pieak4066]

Do you have a project for ROS navigation mobile robot ?

[@TranscendentBen]

I'm guessing because it's lighter weight. The arm with the previous controller oscillated. The extra weight and flexing of the arm and joints makes a mass-sprint resonant system, and the feedback loop has enough gain (and lack of phase compensation) to make it oscillate. I wonder if it will oscillate again when there's an actuator carrying a payload weight.

[@gregd6022]

You need to take some time and understand control theory (yes it's not easy, but if ur doing these posts u owe it to ur viewers/patrons. If you have oscillations you just don't have your control system setup right. Every system and can be "critically damped", the difference is how fast it gets to final position, but again, all can do without oscillation.

[@l3d-3dmaker58]

that PCB looks absolutely GORGEOUS!😍

[@Skyentific]

Completely agree!

[@pieternagels8258]

Just the design of the robot arm is already really cool. amazing job

[@Skyentific]

Thank you!

[@klausnielsen1537]

You are an absolutely madman ( ref: 9:45 ) but I mean that in the best possible way! Stuff like " I contacted a handful of companies to see if anyone of them could do this and only one was able to... So I convinced them to sponsor this video! I love your work.

[@Scott_C]

From my experience with Chinese manufacturers. PCB redesign happens for lots of reasons. Commonly it's for ease of manufacturing, assembly of a motor with a round circuit board may just be logistically faster. Second is cost, producing the same PCB but only making it 2 or 3 layers may be much cheaper if they make it larger. 3rd, laziness/copying, if a Chinese manufacturer goes through the effort of designing one board, and all the documentation and diagrams are in Chinese, they just pick up the previous work and iterate on that. 4th, the country not the companies "own" the circular boards they like to make everything produced in China "theirs" the MIT board is (even though it's open source) wasn't created by them so they change everything in the board so they can "own" it and claim it's not open source.

[@diogosoaresmendes]

This above. Ease of manufacturing, if you have already sourced the motors it's much cheaper/easier to produce 1000's of PCBs that fit just right than to produce 1000's of PCBs + 1000's of mechanical adapters to make them fit, and I hadn't even thought about the number of layers.

[@ferosofranek8817]

So called Mini Cheetah BLDC motor controller in it's original form (from github) is something like functional prototype - it is not ready for commercial use. Hardware design is rough, there are many details that doesn't make sense (probably there was no revision of design), schematic documentation is mess and what is worse, design isn't foolproof (missing protection circuits, no analog filtering, no ESD protection, no temperature measurement, need for bulk capacitor for decreasing current ripple if power supply has significant impedance). Sensor MA700 is NRFND, there is plenty of new sensors (for example MA732 from same manufacturer). Also there are only two "phase current" measurements, which is enough for sinus vector control, but driver DRV8323 supports also third - for better precision and some redundancy I would use all 3 inverter shunt resistors measurements. It is a very basic design based on standard components connected in standard way - good startpoint for someone who want to make his own design. Decent firmware with sinus vector control is more important part of this project.

[@Scott_C]

Nice points both of you.

[@katerinarostova4312]

That's a great improvement in the arm performance, well fixed!

[@mattanimation]

Glad to see someone that had the board made and see it working, thanks!

[@AndyTeddTube]

Nice video... I can’t wait to see the final arm!!! The robot revolution is coming!

[@MrDRBlackBird]

Pro tip from an automotive engineer: The ground potentials of your CAN devices should be connected together, for reliability reasons. Depending on the CAN transceiver, they only tolerate a few dozen volt GND to GND offset (see datasheet). Hence, letting the GND potentials of CAN equipment fly wildly around in the breeze, is not a good idea.

[@Skyentific]

Thank you for this tip! I was not sure about ground. I will take care of this!

[@proceau]

@@Skyentific can add LTM2889 or equivalent for galvanic isolation, if need no common gnd

[@nomen_omen]

@@proceau ​ Could you tell me in robust enviroment is good to using ltm 2889 on every device and make something like ethernet network (star topology) to connect to main controler? (1 master and n-slave devices connected paralel with ltm2889) Maybe this solution is to expensive? What do you think?

[@zayanh2823]

Congrats! This is amazing work man. I aspire to be on your level eventually

[@brad7141]

I fear this channel... Skyentific

[@rayeaglenz]

Always interesting, the PCB looks great, well done, love your socks :)

[@mberoakoko24]

I literally started robotics cause of this channel, first semester begins soon. I also want to make cool robots

[@sculptaware4548]

Have you ever thought about making a SCARA 3D printer. I would love to see someone attempt it. I will eventually attempt it myself but I have a lot to learn about when it comes to robotic arms. Which is why I am subscribed to your channel. Thank you for all your hard work and time you invest to make these videos.

[@pranavsreedhar1402]

kzbin.info/www/bejne/Z4Kre56pideebtU kzbin.info/www/bejne/bqLGYWdqgb5pnsU the How to mechatronics channel has shown how to build a SCARA robot and has also demonstrated laser engraving! short answer: with 3d printed materials its not accurate enough all the time and requires recalibration for different homing points. kzbin.info/www/bejne/kHjKfpiQbZ2mjpo a SCARA robot I built years ago using mostly recycled materials and contains NO servos and NO microcontrollers. it ran directly from the PC using the parallel port. I was able to run and scara and cartesian robot to demonstrate pen plotting with the same software. I can tell you that the SCARA is not nearly as accurate as a Cartesian but is quite efficient in terms of space. The major problem is the requirement to move at least two motor to reach any point. unless you have end effector feedback, its not going to be accurate or precise for 3d printer like tasks. for everything else its fine! this may be a good idea to look into: kzbin.info/www/bejne/nXS3o6WkqdqikKs

[@wobblyjellyman69]

The difference is huge man great work

[@alexeymaslov3865]

One more great video, thank you!

[@Skyentific]

Hank you!

[@alexeymaslov3865]

@@Skyentific why you didn't use jlcpcb service for circuit assembly? The dont have this components?

[@PeachBologna]

I'm going to build up some of these boards for a rover project. Thanks for the inspiration.

[@viswadeepkopalli6160]

If you can make a video on how to order this pcbs it would be veryy helpfulll!! 🥺🥺

[@antonm4411]

Yes please!!! I promise I will subscribe all my family and close friends to your channel...

[@dekutree64]

Yeah, I clicked the NextPCB link, but I have no idea how to properly fill in the quote form, or what files I'm supposed to send from the 3phase_integrated repository.

[@alexanderh5392]

Please do IT for us. Пожалуйста. Bitte Bitte...... Waiting so much.

[@CyBoT_robot]

Hi, I think there are million of people waiting to order your pcb !! (ok maybe less :) ). Est il possible de commander cette carte? je ne comprends pas le lien en commentaire et j ai déjà deux moteurs qui attendent patiemment :)))

[@OGMann]

He doesn't make designs publicly available that I'm aware of.

[@Goodwill345]

Wow this is so cool, you are really amazing, you should sell the new circuit board mounted in the actuators, not everybody can do what you have done. It will make you good money for sure. Next you should sell the robot arm with a payload of 1.5 kg, 3 kg and 5kg, you will be really rich if the price is right and affordable, I will stand in line to buy

[@lecour]

So nice. I really could retrofit our old plasma with that.

[@redlines2535]

wow huge improvement, looks awesome. paka paka

[@jonwatte4293]

The vibrations are presumably because of PID tuning. If you didn't use the same values, it's not a good comparison. My guess is a larger board is easier to assemble and also can dissipate heat better.

[@Skyentific]

They have absolutely different firmware (with different control strategy). So I cannot put the same PID. But I spend 3 or 4 days tuning the PID on original controller, and it was still oscillating like crazy. But with MIT controller, I get no oscillation from the first try.

[@jonwatte4293]

@@Skyentific interesting! Did you try the MIT firmware on the original board?

[@BlueJeebs]

The robot revolutionm is already here! Lesson: Controller MATTERS!

[@pastek957]

With a good soldering iron tip (and a hot air blower or oven for the two QFN chips) you could totally do it at home though

[@nigelhungerford-symes5059]

well done on the build

[@Alexander-wb4dd]

Please Skyentific, upload needed Files, I wanted to order 10 of them for my robotic project, but there's a file left which you should upload. I waited all night for you xD already subscribed and your biggest fan

[@Skyentific]

Which files??? Gerber files on the github page of Ben Katz. I did not change them.

[@Alexander-wb4dd]

@@Skyentific Now I found it, thanks a lot. Learned how to use it

[@viswadeepkopalli6160]

@@Alexander-wb4dd can you help us too ?

[@mosth8ed]

@@Alexander-wb4dd I can't seem to locate these. Was the repo moved from the user to an organization or something?

[@jondhuse1549]

Love your socks!

[@chuongnguyen4980]

I tried ordering 12 or 15 pieces, and the quote I got is about 600$ for 12 boards, and 700$ for 15 boards. The price is about the same with ordering 6 ODrive boards. I am debating my self since ODrive has larger community to support and test. But using ODrive I will need to order 12 encoders, which about 15$/piece. So it is likely more expensive. I hope to get some recommendation, THanks

[@localhawk1]

Regarding the size of other Mini Cheetah Boards. It is much easyer, cheaper to produce. Four a bigger board you do not need four layers(!), or smaller and more expensive sizes of the components. Even with the original Mini Cheetah Board, the space around the box is lost.

[@Skyentific]

This is a really good point.

[@scienceofgame6270]

It is impressive work again. I am very curious about the performance of T-Motor quasi direct-drive motors which you made a video about. They already have Mini-Cheetah type controllers. When will we see T-Motor brand motors in one of your projects?

[@nomen_omen]

@Skyentific how many samples per second this solution provide ( connection between pc and ->arduino->cheetah device)? Could you tell whats is rate of sampling (position, motor currents or voltages) between chetah and motor? I;m very curious how many heat is dissipated during blocking posiotion of arm with bigger load. Do motors needs active cooling especialy when process often stops and blocks position? What happends during sudden power outage? Do motors are standing still or rotating freely during power outage? Many thanks for answers. I very enjoy watching Your adventure with robotics!

[@johnpollard8143]

Hi Gennady, great video on using the MIT controller. What file (or files) did you did you upload to the NextPCB site to get a quote from them for the MIT Mini-Cheetah Controller?

[@frankradermacher3469]

Hi, I liked it very much and tried my best on the links you posted from nextpcb. As I did not find out how to order some controllers there I sent them an email to order some but did not receive an answer. Can you tell us how to order then?

[@Lardzor]

I would think they designed the controller boards to use all available space so they can spread out the heat. Concentrating the components tightly together concentrates the heat that they produce.

[@tanjiro3285]

robo revolution.yaaay🔥

[@GrayStarInnovations]

Awesome!!!

[@AndyTeddTube]

Who designed the board? You or NextPCB? It’s awesome!

[@THESocialJusticeWarrior]

Now do a Kickstarter for the board.

[@ruifazenda]

Fantastic

[@atefbenhazem9073]

They design a circular pcb for ergonomic reasons !! Such as wires connections for some hollow actuators

[@ronvais5838]

Wouldnt stepper motors be higher torque and percision and efficiancy?

[@Skyentific]

The stepper motors are heavy. That is why most industrial robots are using brushless motors instead.

[@laiquocbao2565]

I do think that they redesigned just the form factor is to "deceive" the customer and DIY-er, into thinking that they are using their own secret 11 herbs and spices, lol. Maybe you can redesign or simplify the board for sale, or create control kit for sale to boost the revenue. I think that there are a lot of people interested in the subject (I believe that that number will only rise and rise, due to, you know..., the 4th industrial rev bla bla), but there are not many guide and pre-prepared kit outside for these people that interested but don't have enough knowledge to properly 'wet their feet'. And you know, the total cost for everything is painfully high, especially if you don't exactly know what to buy and what to do

[@laiquocbao2565]

And of course, you can make a STEM course on robotic/robotic arms, or online course, there are several platform out there

[@AWildNoodle]

Could you make a video showing how you do your 3D printed pulley design?

[@aniruddhkb2249]

Amazing job! What all did you take into consideration when designing the 3D printed parts for the arm? I may be doing something like this in the future ( fingers crossed!) and want to know a bit more about this. I am presuming the main factors are: a) Mounting the motors properly. b) Having structural strength to be able to take the shearing stress without deforming or fracturing. c) Not being so long or so heavy that the torques cause the motors to stall.

[@gedr7664]

it would be great if someone added temperature to the controller

[@michelegalliano6128]

The guy from MIT actually calculated temperature from the current consumption without the need of any temperature sensor but I am not sure if that part of the code is open source

[@Tetsujinfr]

@@michelegalliano6128 is temperature not a function of the surrounding environment too? Probably not the same if running the actuator on a hot day in spain or in winter in Novosibirsk...

[@michelegalliano6128]

Yes, I think so but maybe the ambient temperature could be measured once near the main microcontroller

[@fernandoangelrodriguezcort6080]

Ok but, I have a question. What happens when you need near-simultaneous motor control? The Mini Cheetah updates their motor driving values many many times per second. CAN bus is designed to transmit a message at time (according to priority) so, how do they do it? Does every motor controller have a specific priority/address?

[@guillermod.bouzon702]

nice stockings

[@antoniobentofilho2738]

Hi! II would like to other the pcb assembled with the components. The link you put above for NextPCB, opens the manufacturer's home page but I couldn't find the pcb there.

[@iangraham6730]

Excellent 👏🏻👏🏻

[@eeliasselias549]

Really cool video

[@ariello5077]

Is this whole arm works only on battery or it’s possible to connect this motor driver to regular DC power supply?

[@RyuuOujiXS]

Why do they make the board differently? If you're going to recreate something, you might as well make it to fit your needs. Their designs may also be simpler due to the increase in board space (e.g.: maybe they were able to eliminate 1/4 of the layers by using the excess space of the remaining 3 layers).

[@Skyentific]

This is a really good point about layers!!!

[@UnitSe7en]

I think the board re-design is purely for form factor, leading into mass assembly reasons.

[@THESocialJusticeWarrior]

Cudos for wearing eye protection!!!

[@ManWhoLostTooth]

Hey Skyentific have you thought about trying to build a Quadraped? Since you are so into the Cheetah actuators I think it would be the next step.

[@JohnLauerGplus]

Could you possibly post the schematic and BOM you used to create that PCB? I'd love to create my own. I can't find the MIT Mini Cheetah design online. I find their software repo on Github at github.com/mit-biomimetics/Cheetah-Software, but not the hardware.

[@chuongnguyen4980]

IT is here github.com/bgkatz/3phase_integrated and the documents docs.google.com/document/d/1dzNVzblz6mqB3eZVEMyi2MtSngALHdgpTaDJIW_BpS4/edit

[@Skyentific]

This is exactly the link which I used.

[@raafeh9601]

​@@chuongnguyen4980 Thank you so much for posting this @Chuong Nguyen.

[@siddhantsingh5266]

Great job man. Can u plz share schematic or board information. I can't find any hardware. I want to make my own. can u plz provide those details. And love from India!!

[@dennisb4922]

can someone upload the files to request the pcb, please

[@matt_nolan11]

Great video! If the original actuator also used a variation of the Mini Cheetah controller, why did the new one you installed perform so much better?

[@Skyentific]

Because original was not using MIT mini Cheetah controller at all.

[@matt_nolan11]

@@Skyentific Ah, I guess I got my actuators confused. Thanks for clearing that up!

[@Bit-while_going]

I have a pretty good controller in my shoulder too. Unfortunately it also gets shakey after about 20 push-ups.

[@keganholtzhausen7139]

Which design did you have nextPCB make? the new version or old version? which drivers? I am considering getting some made, but not sure what you have. There is a old version, new integrated driver version, and new integrated driver connector version.

[@gu4xinim]

EEEPC, now thats a name I dont hear in a long time.

[@cryptout]

I think keeping the esc (motor driver) on a separate board would be better since these components get hot and fail quite often.

[@Skyentific]

This is not ESC, this is completely different beast. And it should not fail, under the normal use.

[@lefilmigtario6789]

Can you talk about the prosthetic Arm made by the John Hopkins university in their Applied Physics Lab funded by Darpa it is a fascinating peace of technology probably the most advance robotic limb, having the same degrees of freedom as a human hand. I would love to see what you find on it.

[@ОлегСинопальников]

HI, Where I may take PCB of r this cheetah controller? And soft for stm

[@danielnikolov3757]

How much does each board cost assembled? Also, do I need eagle to send them the files for print and assembly?

[@jarekf3083]

Is is possible to order that board assembled from nextpcb?

[@jamesrye01]

I’d be keen for this also.

[@rbardellijr]

I want one too

[@chuongnguyen4980]

I want 12 of them

[@rbarnes0154]

I'm interested also.

[@justtestingonce]

That’s exactly what he done

[@Lichalisen]

I think the Chinese ones have the rotary encoder integrated.

[@Skyentific]

This one has the encoder too. They all have the same encoder integrated.

[@repair9204]

Great Work. I'd like to use the controller board on a BLDC motor that doesnt mount a shaft with magnet. Any suggestion ? Thanks, Paolo

[@bambumbambu]

what was the total cost for one board?

[@omartest309]

Thank you for the great video, I can see that you are using Batteries as a power supply. Is there any power supply which can be used instead of batteries? something with internal shunt resistance?

[@db-dp9ek]

How much would the pcb cost for them to make me?

[@fredericbenzaquine3510]

Hi, nice video, and love your enthusiasm !!! maybe the chinese board are larger for better heat dissipation ? or maybe my guess is just BS .....

[@Skyentific]

This is a very smart guess. Although, on one board they put mosfets one under another one, which I think really bad for heating. As, I suppose, these are components which heat the most.

[@akashsaini100]

skyentific is reaching the level of reprep

[@MAYERMAKES]

is there a reason for not doing a gnd Flood fill o nthe board?

[@samyguerra665]

Awesome video, i got a question tho. What kind of current ( torque) controller do you recommend for brushless motors?

[@DMonZ1988]

what happened to your other robot arm?? did it not work for some reason?

[@Skyentific]

I am still working on it. It is still active!

[@DMonZ1988]

@@Skyentific ah great, I'm really glad to hear that! I've been so excited about it from the beginning and it looks fantastic, so I'm really looking forward to seeing it come together! Thank you for the update! :-)

[@AlexKennbergProfile]

Did the other controllers reduce number of layers required which would reduce cost?

[@Build_the_Future]

When will you do KZbin full time, at a 100k subs?

[@Skyentific]

Great question. For the moment I cannot finance myself through the youtube and Patreon. I need to find a way to boost the revenue. Maybe to make more interesting videos, to increase views. Or maybe to create some product to sell. Or something else. If you have any ideas, I am really interested!

[@Build_the_Future]

@@Skyentific If you can build a high current and high voltage motor controller or manufacture a gearbox. (planetary or harmonic) Because the ones on the market are way too expensive. Maybe sell cad or custom control software I think that would help this hobby to be more accessible.

[@mihaim2]

@@Skyentific I agree with Eric, good gearboxes are still way to expensive for the hobby market. Another idea would be to produce a low cost, UR3 like robot arm

[@Scott_C]

Sell your CAD for the arms you've made.

[@onurmemis3618]

Skyentific Design a 3d printed robot arm and sell the electronics for it in a kit.

[@LongPigg]

A link to buy the board would be useful.

[@Eumldeuml1234]

Which Serial Monitor did you use? It looks like the one from the Arduino IDE but it doesn't autoscale, so it's probably another one? Or did you somehow configured it to not autoscale and if so, how do you do that? Thanks in advance! :)

[@alo1236546]

I love cobot ur5 with 6 motors

[@h3Xh3Xh3X]

When are you going to build an MIT Mini Cheetah?

[@yoghurtwow]

@Skyentific there is still some overshoot , is there room for fine PID tuning there?

[@Skyentific]

Yes, absolutely. The values which I used in video is kind of my first attempt. I did not tuned them.

[@hidalgoem]

Hi, amazing job... Do you have the link of the github repository so I can download the pcb design?

[@akutenshi7023]

Oooo wow

[@Skyentific]

Yes!

[@buck960]

Make it pour a beer for you! haha

[@Skyentific]

Yes, I should! :)

[@GiuseppeCaccavale]

Nice video, congratulations. I would like to replicate the same controller too, I produced the PCB but I'm having trouble compiling the firmware. I can't import it into mbed cloud compiler or compile it with mbed CLI and mbed studio. How did you do?

[@Skyentific]

I just imported the firmware from Ben Katz directory, and it worked strait forward. Just push the compile button and it makes .bin file. Which after you just drug and drop to you module.

[@hysoon6167]

Somehow I cant find the link to the MIT Mini Cheetah GitHub page, can someone kindly share the link?

[@naeemajilforoushan5784]

Thank you, all youe video are usfull for me. Can you introduce me a light Arm manipulator 3DOF (Roll-Pitch-Yaw) to make by myself. Its weight is about 1 kilogram. I want to design a new controller and evaluate my controller algorithm via a light Arm robot 3DOF? Thank you so much.

[@damedane913]

Love your content? Besides your channel, What do you make for job?

[@tommasorizzi6932]

Can you make more video about the coding??

[@sciencefpv7330]

You are not commenting on the GITHUB link when everyone is asking for it? Is there a reason? You mention the GitHub link numerous times but you are not providing it? Please help us out. @skyentific

[@dervideogucker8734]

Here's the comment for the algorithm

[@sebastianuhl]

how much did it cost to produce and assemble the pcb?

[@sokerisful]

This nextpcb shows around 300$ for 5 items.

[@dennisb4922]

can someone upload the files to request the pcbd, please

[@damiandanev9271]

What was a the cost of this PCB?

[@Skyentific]

It was around 60 us dollars then. Now it would cost 70 or even more (because of components shortage).

[@damiandanev9271]

@@Skyentific Thank you!

[@dennisb4922]

can someone upload the files to request the pcbd, please

[@jrohit1110]

Nice Goggles!

[@bjarnivalur6330]

I like you glasses

[@mohammadsam1604]

Hello, you are wonderful. Thank you. I can not order this board. Can you help me? I will send you a message on Instagram. I am waiting for your answer.

[@kristofpucejdl]

Hello Genady! Do you reckon it is possible to retrofit the Ben Katz (MIT Mini Cheetah) style controller into the smaller Gyems RMD-X6? I would like to use those for a small robot arm to fit on my bipedal wheeled robot (kzbin.info/www/bejne/ZmjdYnmYn82crq8)