Mini Cheetah at ICRA

Thanks for sharing this great post.This is so nice.

Hi Ben! Could you give us a little insight into what software is running on the Mini Cheetah? Are you using some kind of simplified version of Model-Predictive Control, or is this brand new algorithm?

See here for the locomotion control paper: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8594448 Relatedly, we're open-sourcing all our software in the next few days, so I'll post a link to that once it's up.

Hi, My name is Young Joon. I am making Mini Cheetah based upon your thesis. As I am looking for the Gates timing belt with aramid tensile cord I cannot find the exact belt you used on the Gates power transmission catalog. I guess your belt has 1.2mm width but aramid tensile members have a minimum 20 mm width. Can you share your belt product name and number instead? I am thinking of 15mm width instead because you commented this is performance bottle neck. Thank you for sharing all your work.

The belt is a Gates 5MGT-535-09 Poly Chain GT. I'm using 9mm width, but it wouldn't hurt to go up to 15mm for increased stiffness

Do you post design drawings for repeating mechanics? What is the date you put the source code?

Hi there, First off. AMAZING WORK!!! Can't believe I get to attempt something similar soon! I was looking through the parts i will need and came across this link to the motor https://www.aliexpress.com/item/32982236672.html?spm=2114.12057483.0.0.261e2cc0sA5QOO For some reason they posted an image with the exploded planetary setup? Do you know anything about this? Can we purchase the motor setup like that?

Ah I found this https://www.aliexpress.com/item/33003549263.html?spm=a2g0o.detail.1000014.7.7d684ce8qn4QtC&gps-id=pcDetailBottomMoreOtherSeller&scm=1007.13338.136569.000000000000000&scm_id=1007.13338.136569.000000000000000&scm-url=1007.13338.136569.000000000000000&pvid=8d5be42d-7893-4dfa-b537-c48a89bc91a4

I don't know about that one, but there is one knock-off on Aliexpress I know is real: https://www.aliexpress.com/item/32985671853.html?spm=a2g0s.12269583.0.0.109d67d4OreVgl I've been in touch with that company about manufacturing the motor modules, but they've just gone ahead and put up their own version for sale. They sent me a sample, and mechanically it is very good, but the motor drive that comes with it is useless.

https://ru.aliexpress.com/item/33009340574.html?spm=a2g0o.detail.1000014.2.4fed50f3Yp6LLe&gps-id=pcDetailBottomMoreOtherSeller&scm=1007.13338.112281.000000000000000&scm_id=1007.13338.112281.000000000000000&scm-url=1007.13338.112281.000000000000000&pvid=48b46301-8eca-421f-8fa0-b05188414cfa And what can you say about this seller and the robot? Can you recommend this product for purchase?

I have no idea. They are completely unaffiliated with me, and I have not seen their hardware in person.

Why is the driver useless? Low quality components/build, or not your newest firmware?

The driver is their own, not my design. It doesn't have phase current sensors, is full of bonus wires, as much worse fets, and so on: Here are some pictures: https://drive.google.com/open?id=1-Ko9CfRbY-AY-8eI0VaEVJIqv3RkCwFi

Yeah, the drivers don't look like anything special, I wonder why do they need the extra two wires going into the motor? On a side note, do you see a possibility of connecting these actuators in parallel for increased joint torque in a safe manner e.g. with fully synchronized movement?

The two extra wires are for a thermistor in the windings. You could connect them in parallel as long as you were only using them for torque control, rather than position/velocity control. With position or velocity control, if there were any discrepancy between position or speed the motors will fight each other.

Hello after reading your thesis which is very interesting, i find the 3phase_integrated PCBs and SPIne on github, but i would like to ask you if it is possible to have also PCB power supply out 12v and 5v with kobalt 24v battery, thank you

Hey Ben! Love the blog posts; you've been a huge inspiration over the years to finally get into robotics and control theory. Two quick questions: 1) Any chance you could share your ICRA 2019 paper? I'm having little success finding a copy of it anywhere. 2) I've been following your excellent thesis and posts, bumbling my way through EE and CAD design, but I'm finally starting to make a bit of progress. I just wanted to check with you to see if it was okay to begin a blog/writeup of the process behind reproducing your results, and what I'm toying with that's slightly different than what you've accomplished (e.g. using a resin casted planetary set instead of steel gears). I have yet to post about it publicly - wanted to get your permission first - but you can find some unorganized photos here: https://photos.app.goo.gl/ZQk9NPxY2Py8Hzps5 Thanks again for sharing your knowledge; super appreciated to have such a great learning resource!

Awesome! Feel free to post publicly, of course. I think IEEE hasn't publicly posted the final versions yet, but here's the preprint: https://drive.google.com/file/d/1vn5ZH6VeRR7EZAmNfsooBHeNsIp2QG_O/view?usp=sharing

Good day! When will you publish the source code for the CAD files of the robot parts and the source code that you use on the UP Board?

Hi Ben, what do you think about the new live drive motors from genesis robotics?

Ok so I'm looking to get into robotics and have a very low budget, where should I start. Probably not the place to ask though.

Your're getting famous: https://www.youtube.com/watch?v=hBKNxoj02KM Here, Sanbae Kim is advertising mini cheetah in Germany...

Dear Ben, it seems most robot dogs have all 4 logs designed with back-facing knees, maybe all of them except the ANYmal robot, which seems to have the back legs designed with front-facing knees. I wonder what are the reasons to choose any of these two options. It seems biological quadruped animals have their back legs designed with front knees, maybe that would bring any advantage? have you given a thought to this? Thank you, David.