Simple Cycloidal Robot Knee
ฝัง
- เผยแพร่เมื่อ 29 ก.ย. 2024
- / gouldpa00
25:1 Triple Disk Cycloidal Gearbox
Open Source
It is 2Kg load @ 0.2m with 25:1 ratio. Peaks of 2A @ 12V.
www.thingivers...
It's my own ESC desgn and AS5147 magnetic sensor. The precursor to this hackaday.io/pr...
Music (TH-cam)
Clover - Vibe Mountain
Yummu - Text Me Records / Leviathe
Happy Sixth - Freedom Trail Studio
![ได้เวลา! พาน้องแฝดฉีดวัคซีนครบ 1 เดือน🫣 [cc] แดนแพทตี้ SS2 | EP.52 |](http://i.ytimg.com/vi/IzeLtaXBLNE/mqdefault.jpg)
This is a work of art. I've been following your public projects for a while now, and you've really outdone yourself with this one. Just, wow.
Thanks Leedan, I am trying to make TH-cam videos on more private projects. I just takes me many prototypes/revisions before I'm happy to share. I like this knee joint, so I will make the 2x hip joints next.
Love it! Love robotics! Jake from UK
Vamos trabalha Junto💞 podemos faze um Carro🚖🚔🚘 Voado Drone 🚀🛰📡 android🤖 Nunca Visto no Mundo um tranforme 💰💵
Do you have the bill of materials? I'd like to build one and convert it to an elbow joint for the inmoov.
Wonderful, thanks for sharing!
Great job as always ! Can you make a video that focus on the driver of brushless motor please ?
Thanks, I'm writing it up now and will put it on Sparkfun blog and Arduino forum. Maybe I can do a video too.
Did you put it on the forums? Would love to take a look.
can you send me the foc control code for BLDC motor? Thanks!
What is the advantage over a planetary gearbox?
wow , very well. Thanks
Thanks for showing up all the details of this amazing project! What ESC and positioning sensor do you use for it?
It's my own design ESC and AS5147 magnetic sensor. The precursor to this hackaday.io/project/165217-foc-brushless-actuator-arduino-compatible
@@PaulGouldRobotics What ESC would you recommend as a substitute for yours? Something from Aliexpress?
@@gedr7664 There is nothing that I could find, that is why I made my own. ODrive is best and only.
@@PaulGouldRobotics Thanks Paul, in that case I will wait for yours :). I don't have the expertise in electronics yet to do such a big project, I'm afraid. If you require something software/controls related, let me know though! Ie I could make a ROS port for your quadruped once its done
Thanks for all your hard work and looking forward to more of your wonderful ideas
George
Paul, I simply love this design!
Thanks, the rest of the leg is now printed and assembled. It just needs electronics installed and some control software.
awesome work man ! if only you could buy a good microphone to talk so we dont just watch you assemble things !
Thanks. I like making the robots more than making the videos. I'll try to put more effort into the videos in the future.
It may just be a matter of perspective, but when you were putting the pin in the shaft, it looked like you were trying very hard to impale your hand.
It wasn't as bad as it appears on the video. Filming while making is something that I'm yet to master.
Thx for sharing. I hope you keep evolving on it. Many projects will become viable because of your work, Paul. An affordable Cycloidal Gearbox is a gamechanger for us makers. I hope you´ll follow up with some infos about loadcapacity, durability and used materials. I had some succsses with nylon for the gears in the past. A nightmare to print and keeping the tolerances, but really good in the noise, friction and durability departement. Perhaps a combination of fiberinfused fillament, or PETG for the coreparts with Nylon on the contact surfaces would be an option. Keeps the hassle with nylon to a minimum.
I'm trying to get the design as simple and reliable as possible. I wonder if it is possible to replace the two large bearings with integrated 3D printed bearings. (They are US$10 each). I use ABS for this project, which took me a while to get right (Perforated, heated print bed and an enclosure). My old Chinese 3D printer can not get hot enough for PETG and Nylon. I think that Nylon would be the perfect material for this actuator.
Sintered bronze bushings would work ok for the large bearings, You dont need the balls for low speed.
@@PaulGouldRobotics check out this slew bearing if you like: th-cam.com/video/5o4Fj8OxkB8/w-d-xo.html I however believe regular bearings are more reliable.
@@PaulGouldRobotics If i guess right, the large bearings are roller bearings. This should be easy to integrate and work even printed in one go. But you need a good qualitiy printer , because of the clearences. Not easy to replicate on every machine. Because of your Thingiverse account, i had the impression, you got a Prusa MK2 or 3. If you got one of the chinese knock offs , it is easy and cheap to upgrade for Nylon and PETG. Just change the hotend to an original all metal e3d V6. Should be a drop in replacement if the power supply plays ball. Instead of the Perfboard you could use PEI Sheet. More reliable with every material, easier to work with, and you get a nice smooth finish. I'm building and working with 3D printers for about 10 years now. I used ABS almost exclusively because of the mechanical properties and the ability to smooth and weld the parts with aceton, and switched to PETG 6 months ago. The mechanical properties are better (layer bonding, threading, uv resistance), less fumes, and no warping or shrinking. My development process got much faster. I know how much time and iterrations are required to get from the first working prototype to an elegant, reliable "product". I would suggest to concider a nylon slipring design, instead of the large bearings. If you´re a willing to give me the CAD Files in a workable format, i could try to adapt the design for the nylon/slipring path and hand them back to you, if it works and is an improvement. Meanwhile you could follow the bearingpath without getting sidetracked. Hope there was something helpfull in this wall of text. And sorry for my remarks. I don´t want to criticize your work in any way, shape or form. I don´t know your designgoals and constraints. My brain just automatically switches in problemsolving mode, when i´m inspired by a good project. ;-)
@@Robonza Good Idea. Could you provide a link? I haven´t found one cheaper than Pauls $10.
Hi, I totally understand why you used 4 cyclo gears instead of 2 (for the same reason, why 2 is used instead of 1) this was my very first thought, when I learned about cyclo. Do you have any actual proof, that this actually works better irl?
Did the shielding on your bearing cage pop off ?
Yes. I have removed that bearing 100's of time with screwdriver. I must have deformed the seal one time.
I'm interested as to what level of positional accuracy this is capable of with 3D printed gears
It's not great. Good for quad, bad for robot arm. Unless you use a high res encoder for the joint feedback. Also backlash may be a problem, check out my other designs to fix that too.
@@PaulGouldRobotics Interesting. If I have the time I might work on creating a version with a high res feedback encoder, because this design has a lot of potential for a robot arm since it uses fairly inexpensive, compact parts and is capable of a fairly significant payload
Melting the plastic to stop the bearing from piling out is smart. Don’t know how I haven’t thought of that, definitely going to be using that method from now on.
Derived from putting nutserts into 3D printed parts.
@@PaulGouldRobotics use teh printer plate to give finished surface sheen? :)
Why did you not just taper the pockets for the bearings so they snap in? Could even just have some snap claws.
That is a good idea. I will try it.
@@PaulGouldRobotics I have successfully used three or four 'pimples' that stick into the bearing space (O/D or I/D). But they are placed just above the bearing so that the pimples hold the top edge of the bearing (and are not permanently squashed). The trick is that the pimple protrudes from a stem, created by a vertical cylinder cut behind it that extends a little above and below the pimple height and breaks out either side into the bore or O/D (ie: you cut half a cylinder behind the pimple). The space behind the stem allows the pimple and stem to deform into the space as the bearing passes by and then recover to hold the bearing in place. The cylinder cut could break out of the top surface which puts the pimple on a cantilever or make cut fully inside the part which makes the steam an encased beam (much better). Of course, this idea only works if you have enough space above the bearing (you only need an extra 1mm of height for the bearings you are using). The beauty is that this design can be integrated into the single 3D print. (BTW, I realise you are doing something similar for the large bearings but my idea incorporates some subtle but useful differences that scale down to sub mm features.)
My project is based on your project. Could you share cad files (not stl) thank you.
The Inventor Pack and GO files are on Thingiverse in the Knee_Simple.Zip file
@@PaulGouldRobotics thank you
Paul, AMAZING WORK! Well done. Do you share models? I love to work on this.
It's all open source. Links in the description.
@@PaulGouldRobotics I missed that sorry!
Are you willing to sell complete knee joint kits such as this?
NICE WORK
wonderful, Looks like perfect
I made a robotic arm using an aluminum profile. However, it was difficult to solve the stiffness problem and backlash problem of the aluminum profile.
do they use this type of joint in boston dynamics' robots?
I’m not sure. It is in Cassie.
can you put the motor inside of the gear to make it even more compact?
I am about a week ahead of your question. Check out my last two videos.
Value city
why have i fallen into the cycloidal robot joint rabbit hole
gorgeous
Extremely well done!
That is so nice! Congratulations.
Thanks
Noice
How much weight it can lift...?
Was just wondering if the torque on the output shaft is the same as the toque of the housing when you fix the output shaft in your design. Theoretically the input to output reduction is something like the difference between the number of ring pins and input lobes/number of ring pins. Is this the same reduction or is there a slight difference like between the output shaft of a planetary vs the ring gear of a planetary when the output shaft is fixed. Well done again.
Motor Torque x gear ratio (25:1) = Output torque
Efficiency also needs to be taken into account, which is quite bad ~80%. Most metal gearboxes are between 90% and 99%
reduction ratio = (outer lobes - ring pins)/(outer lobes) = (25-24)/25 = 1/25
Are those contact angle optimized cycloid disks? Was there a motivation for having non rotationally symmetric bearing placement in the cycloid disks aside from room for hole placement? Very cool work!
I'm not sure about being optimized mathematically, it just what looks about right. There was some limitations of bearing sizes, gear ratios and eccentric cam offset distance.
The cycloidal disks have 6 bearings, 3 bolts (to join the two outer disks) and one through bolt hole for final assembly. It was the best placement for all the things I needed balanced evenly as possible. I think that 6 pins are enough to transfer the required torque.
Very strong design and no blood or band-aids were used during the assembly.👌 Just need to engineer a good method to route all the necessary wiring for a clean and streamlined look.
There was blood, just edited out of the video. The motor shaft pin is a bit tight.
I've actually wondered if motors of that form factor existed and were usable for higher load strengths (like for the legs of a decent-sized robot). Thank you for sharing Paul! 2A @ 12V doesn't seem too bad at all.
Some people are using the hobbyking.com/en_us/9235-100kv-turnigy-multistar-brushless-multi-rotor-motor.html direct drive (no gearbox) for robotic arms and quads.
I really appreciate your hard work!
I need a way to contact you directly if you don’t mind
You can chat to me via hackaday.io or Odrive Discord (gouldpa 9015)
Very cool. Cant wait to see the thing walk
Making a leg (I can do, hopefully), making it walk will require someone more skills than me. I'll give it a go at some point.
@@PaulGouldRobotics might give they guy who did open dog a email. he might be able to help you. his dog looks more complex in the way he did his compared to yours. didnt i read some were you were a mechanical engineer?
th-cam.com/video/oJOV29fq_KY/w-d-xo.html
@@mastermoarman My Joints are more compact and lighter than James's Open dog. His would be more robust. I would love to have it walking like Nate's mini cheetah. Yes, I'm a Mechatronic Engineer. Gearboxes, FOC brushless controllers and robot mechanical design, I'm doing that, slowly and learning a lot. High level software for "real" walking and stable gaits, I'm not too sure. This is open source, so I do the the stuff I know, others do the stuff they know and it gets combined to a full walking robot.
Hi, I love this design and i will really try it out. Can y plz plz provide a BOM. All the Bearings are expensive (~150€). I have to be absolutely sure to buy the right types and the right count. i have also no clue what these shafts exactly named and there to buy.
I will update the Thinkiverse page tonight with more information. Also look at plaig.com.au for bearings
I have updated the Thingiverse page with quantities. Let me know if you need more information.
@Paul Gould so you don't try to incorporate anti-backlash anymore?
Or just because of space constraints.
btw. I really love to see your incremental changes in your designs its the little things ;-)
I only hope it didn't take to many iterations.
This knee joint has no backlash. It feels more like springy backlash (hard to explain). I don't think quadrupeds need high accuracy joints like robot arms do. As a few other people are also working on this type of gearbox, I can take their improvements and incorporate them into my design. That's why I love open source.
Once you have it all mechanically complete you can do some great things in software
Man where did you get those crazy big bearings with such large diameter OD and ID?
I get them from a local industrial bearing shop. They are not cheap. You can get them online for about US$3 search "bearing 6809RS"
@@PaulGouldRobotics Thanks! Also, awesome design. Very inspiring
good afternoon. what board do you use to control the motor?
It is my own design. hackaday.io/project/165217-brushless-actuator-arduino-compatible
I have a few versions based around different microcontrollers.
Have you done some torque/current control tests? 15:1 is quite a high ratio and I wonder if you will have enough torque sensitivity for the robot to "feel" the ground the way modern dynamic quadrupeds do.
It is 2Kg load @ 0.2m with 25:1 ratio. Peaks of 2A @ 12V. It is backdriveable so it should be able to sense the ground. I may also look at some sensors too
@@PaulGouldRobotics 2A @ 12V peak? According to the specs of the motor it could do a lot more amps, right? Or would those torques damage some other part?
@@eelcohoogendoorn8044 Baby Steps. Yes, it's a 850W motor and i'm putting 24W max into it. I will gradually increase the load and see where it breaks. It also moves quite fast with no load and it might get a bit dangerous to be close to. I will also have to change over to Li Batteries or my 30V 15A power supply (with no current limit)
EXCELENTE!!!!!!!!!!!!!
Why do you use three cycloidal disks?
It stops twist between the upper and lower housings. 3D printed material is quite flexible/springy and cannot take high loads. It also dynamically balances the system when running at high speed.
Hello. Are you able to link me to the motor you're using in this design? Thanks!
The motor is Multistar elite 5008/5010 from hobbyking but it is always out of stock so now I use Gartt ML5010 from gogo-rc.com/store/index.php?route=product/product&product_id=3448
Gartt ML5010 from gogo-rc.com/store/index.php?route=product/product&product_id=3448 with a modified shaft
Would you help me?
How did you work out the math for the cycloidal curves? I was designing my own cyclo gearbox for uni, but I was working off pre-made sets and I'm having trouble finding out anything
Also, isn't this a two stage design? Why do you call it Triple Disk?
hackaday.io/project/163766-low-cost-3d-printed-robot-tail/log/159482-gear-box-design
It is a single stage but uses three disks for balance and stopping twisting. For something "simpler" check out "cycloidal" on thingiverse
If you need more help, you can chat to me on the hackaday page.
Another great explanation on the disc geometry: www.tec-science.com/mechanical-power-transmission/cycloidal-drive-speed-reducer-gear/construction-of-the-cycloidal-disc/
@@PaulGouldRobotics I will, just as soon as they finally decide to send me verification mail lol
Which torque have this system and what is the operated weight this actuator ?
2kg @ 20cm @ 24W and the motor/controller is not even trying. I need to test the joint till breaking to find out it's mechanical limit. Hopefully 5kg @ 20cm is all need for walking. The Motor is 850W and the controller is 500W both with cooling. The joint in the video (no controller) weights 700g. One leg should weight about 1.5kg and the whole robot (with batteries) about 8Kg.
Great work
what motors are you using
Gartt ML5010 with the shaft removed and replaced with a piece of silver steel rod.
thank you for sharing.. good job!
i also work on a knee. i want to use a fishing line between the gear motor and the join...to have a spring for jumping and running..
what is the gear ratio and what is the name of that gear type? it seams to be very quiet. is it very effektive?
and what esc you use?
i hope you share more content ...of course i subscribed.
check out hackaday.io/paulgould
It is very quiet but has a bit of a "clunk" when it changes direction. I use coconut oil to reduce the noise and friction. Do you have a link of you work, I would love it see it? I have tried the rope joint with little success. I haven't tried the "twisted string" yet. The gear ratio is 25:1 and is called a cycloidal gearbox. There are some good animations on the web.
What bldc do you use?
ML5010 from gogo-rc.com
@@PaulGouldRobotics thanks, and how did you identified torque? Or you didn't analyze before assembling?
@@XkomitusKram I didn't measure the motor's torque. The 8318 brushless motor or similar is more common.
Hi Paul, What do you use to control brushless motors?
The precursor to this hackaday.io/project/165217-foc-brushless-actuator-arduino-compatible
I think I will write up this version though. It is quite simple only using A SAMD21 dev kit, 3x bts7960 half bridges and a AS5147 magnetic encoder.
@@PaulGouldRobotics will you post about it on this youtube channel?
Awesome Job!
Thanks
what is the name of the brushless motor?
ML5010
@@PaulGouldRobotics where did you get the magnetic sensor?
@@andressantos4054 AMS5147 from Digikey and I made my own board
@@PaulGouldRobotics Thanks for this info :) If I may, how did you attach/join the long shaft on it ?
@@rvg557 Removed the set screws. Used a bench vice a smaller shaft to press out the standard shaft. Got some 6mm silver steel rod. Made sure the 6mm bearings fit. Filed some flats for the set screw. Pressed the new longer shaft in with a vice. Tighten set screws.
Great job! Next step - I think you can smooth the movement by software.
Thanks, The positioning software is very simple. It only has P (no I or D) and no current monitoring. Once I get the new board and hip done I'll get back onto the software. I'm not good at High level control software, so I'll have to get help.
@@PaulGouldRobotics i worked in audio dsp but have thought a lot about robotics :) reviewed control methods used for quadcopters (last i remember kalman filters were trending, like zero feedback delay filters for audio last decade), never did much with it but my perspective is that audio dsp has extensive development of articulation in low dimensions other fields would benefit from.. if you feel like taking a detour, try a bit of audio coding :)
@@atomictraveller The maths have already been done on this. It's called Clarke and Park transformation. My problem with this is my CPU is a bit gutless (48MHz and no FPU). I will have to come up with a approximation.
🤦♂️
@@PaulGouldRobotics use a teensy 3.6. It has a FPU and a 180MHz. It's awesome and relatively inexpensive!
Hey Paul. What brand of motors are you using?
Gartt ML5010
24w for such motor size is really so few, you can give it triple more power for faster movements
It is a 500W motor and the controller can peak at 1KW. This is a prototype so I don't want to push it too hard too early. I'm pretty happy with the torque. I've upped the current for further tests. I may have to add a motor cooling fan at some point.
@@PaulGouldRobotics ok, and what power this motor can run continiously without any cooling system?
@@Tachi778 100W gets to 50deg C th-cam.com/video/EE5JBO_b6KY/w-d-xo.html
@@PaulGouldRobotics OK, checked it out. And one more question. What's the maximum safe temperature for this motor?
@@Tachi778 I don't know the quality of the magnets, so I would say 80degC max is safe to stop de-mag. Windings should be ok to 120degC.