openDog Dog Robot #2 | Leg & Joint Prototype | James Bruton
Вставка
- Опубліковано 8 вер 2024
- Part 2 of my openDog open source hardware & software Dog Robot. This time I'm prototyping one leg using Odrive and a ball screw to check joint angles and control using the remote and code from part 1. This robot dog project is open source! - get part 2's CAD and code from: github.com/XRo...
My open source robot dog project. It includes everything from 3D printing to coding, including a lot of CNC cutting and installing all the electronics myself.
If you want your very own four-legged friend to play fetch with and go on long walks then this is the perfect project for you. This way a dog can literally just be for Christmas. The full CAD is available at the link below for anyone that’s keen to build their own.
github.com/XRo...
odriverobotics...
github.com/mad...
github.com/mad...
① GENIUS
No MacArthur grant yet. But while I’m still waiting by the phone for that call, maybe you might patronise me with your Patreon-age. Or, if you’re all out of digital cash just comment on one of my hilarious Instagram pics, pop me a Tweet or generally yell in my direction.
Patreon: / xrobots
Instagram: / xrobotsuk
Facebook: / xrobotsuk
Twitter: / xrobotsuk
Huge thanks to my patrons, without whom my standard of living would drastically decline. Like, inside out Farm Foods bag decline. Plus a very special shoutout to Aleph Objects, Inc who keep me in LulzBot 3D printers and support me via Patreon. www.lulzbot.com/
Why not join my community, who are mostly made up of actual geniuses. There’s a Facebook group and everything: / community
② XROBOTS
Former toy designer, current UA-cam maker and general robotics, electrical and mechanical engineer, I’m a fan of doing it yourself and innovation by trial and error. My channel is where I share some of my useful and not-so-useful inventions, designs and maker advice. Iron Man is my go-to cosplay, and 3D printing can solve most issues - broken bolts, missing parts, world hunger, you name it.
XRobots is the community around my content where you can get in touch, share tips and advice, and more build FAQs, schematics and designs are also available.
③ MUSIC AND SFX
The majority of the music and SFX used in my videos is from Epidemic Sound share.epidemics...
"I'm not as good as Boston Dynamics"... Oh James, so humble. Don't sell yourself short man. You've got the tinkering mind and that makes you every bit as good in my book.
Bruton Dynamics.. nice
good grief, i freaking love this series already. Great work James!
Thank you for being a great inspiration in my life!
Securing your ball joints in the M10 rod ends could be done with LOCTITE 638 RETAINING COMPOUND (available in the U.K.) Inexpensive, super easy to apply and rock solid. Great work you're doing !!
I’ve not even started to watch this yet but I already know this will be a great video! It’s right up my street in terms of both topic and the approach James is taking.
Look forward to after work when I can watch it!
I think this motors have enought power to run the join directly with a belt reduction. You would get a lot more control response and less mechanicals parts.
Sorry to be off topic but does anyone know of a way to log back into an instagram account?
I stupidly forgot my login password. I appreciate any help you can give me.
@Boone Samuel instablaster ;)
@Cesar Michael i really appreciate your reply. I found the site through google and I'm in the hacking process atm.
I see it takes quite some time so I will reply here later with my results.
@Cesar Michael it did the trick and I actually got access to my account again. I'm so happy!
Thank you so much you really help me out !
@Boone Samuel No problem =)
Amazing, this project inspire me for my school project ! Thank you, you'r very good as always !
Thank you for helping and supporting me to keep on 😄😄😃😃
bro i love you man!! you are so good in telling the idea basically behind the technics in this project its very easy to follow and listening to you!!
UA-cam safety hero here: I whinced when you moved the leg with your fingers near the ball screw.. There's enough advantage there to completely crush an appendage!
congratulations, im writing from Venezuela, i wish one day have the money to build this robot like you.
We have to make this thing dance on its back legs!
You can always increase the speed (and travel, for that matter) by moving the connection point of the lower leg to the actuated linkage closer to the knee joint.
For your bolt clearance you should user countersunk machined screws :) You'll be able to CNC these in metal with a countersunk drill bit too
Super excited to see this project come to life!
Love your channel. So interesting even tho I don't do anything like this myself. Keep inspiring the masses. ThankS. TkEZ
Hi James, I was the guy that was on the train with you going to Portsmouth. Keep going buddy love the vids.
I always watch and comment for your videos . I am 12 years old from Lebanon .my name is Malek daroub
The deadspot/sensitivity issue with analog input is typically solved in software by adding a software deadspot if applicable, and adjusting the curve or smoothing the value before intent is interpreted.
openDog could also be a dog taxidermy service
13:24 - Slow motion footage of the Boston Dynamics evil robots reveal that they appear to use feathery/flickery movements and dissipate the weight of the robot proportionately so I agree with you regarding the suitability and agility of the motors for the project.
They're not Boston Dynamics agile but definitely suitable for an open source project.
Loving the whole development. Still think stability would benefit from a DoF on the spine, though.
I might try to simulate the body, to see if I can teach a neural net to use it to move around. Beats handcrafting locomotion logic, because it can learn while being unattended, and adapts well going from sim to RL, changing surfaces, limb length etc.
Ingenious drivetrain design, looking forwards to new episodes :)
You might want to use inrunner motors instead of outrunners, they may have less torque but are far more responsive for positioning tasks...
the robot which finally made james go outside to build a shed to get away from robots. :P
It is going to be the most Amazing robot you have created !
You should name it Dogmatic! That's an actual word. Also, you would be as good as Boston Dynamics if you had the same resources they have (tons of funding, teams of people to help, etc.). And another also, you should build a mechanical arm like from Iron Man 1 that has a camera and, for example, tracks a colored soldering iron and moves itself like your Ultron robot to help you for holding solder up to where you need it or something.
Even though you need the triangle for trigonometry, it doesn't mean you can't bend the studding with ball joints to avoid hitting those other parts.
Nice progress! The series could use a test rig. For example, measure force by having the joint push against a weight scale. Some plastic parts might give but easy to reprint.
I paused the vid on the pulse math to write a comment about the belt reduction. Thankfully I decided to watch another 10 seconds :D
I'm trying to drive a mechanical car speedometer using a cheap hall effect sensor board as a rotary encoder. I'll machine the trigger wheel into the adapter that fits in the speedo cable hole. Motor is brushed DC. The input is coming from a GPS unit, allthough I could possibly read the sensor on the gearbox as well. I might need to run some sort of input filtering as well if using the GPS..
It seems there is more than enough toque for you to directly connect the motor to the ball screw and you'd get the additional speed
But it's fantastic as it is and I'm so dam excited to see the result. Also excited to hopefully contribute to the code if an opportunity arises.
Now that you have more toys to play with (CNC, etc), just looking with the motor + belt + torque.. have you looked at how to mount the motor to be more like an electric skateboard hub or the carvon direct drive system instead of a belt?
Just a thought towards the end of the video.
I really like this project. I'm trying to roughly model the behavior of such a 4-leg dog with 3 joints per leg in Python with Pybullet. I have something 'walking' but the dimensions and masses are not yet similar to this project. On the other hand the modeling with Pybullet is easy and seems flexible and realistic enough to be valuable for a project like this. Once it works the challenge will be rough terrain (stairs, steps, stepping stones etc).
Also I worked out the inverse kinematics formula's for the two pitch-directed leg angles and I noticed that it is not really necessary to be so strict on equaly of lengths of leg parts or on putting joint locations exactly on the ball screw shuttle axes. It may require a few extra cosines and sines but I think computational overhead should still be minimal.
Thirdly, apart from mentioning Spot Mini from Boston Dynamics in your videos, there are also others you may give a look/comparison. The Swiss have ANYmal, MIT has Cheetah and I also saw some nice work from e.g. Japanese, Russian and South Korean universities and hobbyists.
As a last suggestion you might want to add a small project to your project list where you make a very small robot with cheap parts (e.g. sg90 servos) that people can easily build along at home without a huge budget. Making it walk nicely will still be a rewarding puzzle.
Thanks again for your superb videos!!
Rather than committing several directories, you could just add tags to certain commits detonating a video part. This is typically the method used for release versions. of software.
James, have you seriously uploaded a video at 4:20 AM, You madman.
Just gonna say the same thing.. :P Still gonna watch it tho..
I think you can choose a time and date to make a video available after uploading, even making it only viewable to select people for a while before making it public. If he isn't doing that then he probably set it to update then went to bed.
James Bruton What program ddo you use to make the legs and other parts move, simultaing them working as they do in real life?
That's Fusion 360 ... it is free for hobbyists
Florian Müller thank you (I feel stupid now). I didn’t know you could move parts in the program itself. Thanks again
3:47 : please keep the balljoints at the end of your rods.
It’s actually releasing stress, not causing it. It also helps with vibrations and assembly.
If the ball moves without friction then there is no stress. If the ball doesn’t move but should (rigid links) then you have a lot of stress going on.
Source : I am a mechanical engineer
James i wish i knew you IRL !!! you are awesome!
What I don't understand is why you are mounting the screw sideways? If you turned it so the flat side of the nut was facing the extrusion and mounted the bearings on the extrusion, the extrusion and some grease would keep the part fairly stable. If you add a flat steel stock to the flat side of the nut and to the bearings, you could adjust the height so it is perfect and use the flat stock to attach the extensions to the distal appendage and stabilize the nut from rotating.
Maybe you could (change the ball screw pitch) drive the ball screw directly with the electric motor. No belts and no motors hanging out of the legs. You have lots of extra power in the motors.
That size ball screw with that size motor could present some inertia problems in direct drive. Also, ball screws do have a rated max RPM, which would limit the top end RPM even with that short a screw in a fixed-free format.
I can see that could be a problem. Just hope the belts wil hold under stress. A motor with reduction box 2.5:1 would be nice.
Hey James, you should use those balled eyelet bolt ends for the hip joint, or perhaps something similiar to allow for an eventual push over test lol
Awesome work man! I definitely need to learn this kind of thing!
I am living in HK, really hope to have a working place like you did to build the things :)
I love your projects, but... I think you should have made the leg joints as a 80/20 Fulcrum pivot joint, with the motor connecting to the end of the 20 percent side. I would have used a high speed linear actuator step motor. This would cut the weight down drastically, increase speed and strength, and reduce cost as you would not need an encoder as that can be done directly with the stepper motor. It would have also simplified the design. :-)
There's zero chance a stepper motor would be strong enough, these motors are rated at 2KW
They use stepper motors in high speed industrial robots and more. As you know, DC, AC and Step motors are not created equal, so whilst you may think you need a 2KW brute force rated DC motor here, you would not require anything close to that for a stepper motor, with a nice design of course. :-)
These motors as faster, more powerful, and more accurate than steppers - the encoders have 8192 counts per motor revolution which is far more than any stepper, even with microstepping. AND CHEAPER!
... and these never lose steps even if you jam them because they are encoder driven.
What if you take those side mounts for the ballscrew and wrap them around underneath the aluminum, bolt them underneath and then use .25, .5, 1 and 2mm shims to dial the ballscrew wheels into place? That way you wont have to be exactly exact. Plus I'm sure the extrusion isn't all the same, so that'll give you the slop you need.
I was just about the write how this system seems clumsy and slow and something like a pneumatic/hydraulic piston would be much much better, but it seems to work far better than I expected!
I was worried that it would've ended up like with your Hulkbuster, those motors were very damn slow
You could use a branch for each episode, and then have a production and a development branch for the final product in your GitHub repository.
Not sure why you didnt go with placing the motor at the end of the drive shaft instead of having it belt driven. Moving the motor at the end would reduce brackets and width of the leg while possibly increasing the length as well as a slight reduction in weight.
Let's make a pull request to youtube for a "Really Like Button"!
This Series needs one...
You could use excentric nuts to tighten the wheels into the slots
11:30 Can't you mount the bearing block horizontally? Then the sides would be free of obstructions. If the spacing of the holes doesn't fit the slots, then the bearing blocks have holes in line with the screw that you could use.
...what will the "insert part here" be made of?
METAL!
On github you could use branches instead of folders.
Love the videos James. In addition to the controller I think having a visual tracking feature would be cool so you could not have to use it and just have it follow you around, I'm sure your already planning this but if not :) Thumbs up as usual, my favorite channel on youtube!
I can't wait for an openDog race day where everyone brings along their robot and they're all raced to see who's robot is the most evil behind the original and most evil Boston Dynamics creations!
On second thoughts, maybe I can't deal with the psychological torture such #nightmarefuel might deliver!
James, you keep saying you are not as good as Boston Dynamics. I have been a fan of Boston Dynamics since early 2000's. Here is something to put you in perspective: Boston Dynamics was a team that split from MIT in early 1990's, they got the wind put in their sales by DARPA, where they had access to the best technologies and data in the world. They are always on the leading edge of robotics tech (They are working on 3d printed hydraulic internal actuation that is intrinsic to the framework of the new ATLAS, for instance), and they have the most advanced team with the greatest education and over twenty years of experiance.
You are working out of your house with basic materials, no history of government funding and support, and completely by yourself. And yet you are able to design and build a comparable product, if only on the edge of entry level. BY YOURSELF. I think you would be right at home with Boston Dynamics.
You seem like the man who could answer this as I'm curious about the economy. What was the difference in cost that you'd worked out between prototyping in aluminium compared to 3D printing?
James Bruton thanks for the response.
Why not use a cam or eccentric spacer in the middle of the block that supports the V wheels. Use an internal screw to adjust it.
then your able to adjust tension but wont need to make all the frames movable and thus stronger.
Is you search for "JH-D300X" on eBay, you'll find the joysticks. I wondered what they were actually made for, it seems they're almost exclusively used for monitoring CCTV.
A laser cutter would be veryyyy helpfully to prototype those side panels and the the height correct faster....
again some impressive work there James . Top stuff
@James, hey man that is what I was waiting for, make my own Boston Dynamics Dog robot. Man I am going to build it. Normalky I build CNC milling machines for the hobbyist.
I was always wondering if the legs and Arms (Atlas) are electric motor driven or hydraulic driven, cause those robots can also run which mine also will have to be able to do.
Mine will be called Tars and will have intelligence and learn like a real human, not a dog, so a dog with learnings of humans, thats bloody stellar.
So cool and well done project, well done
This is such an exciting project!!
Why dont you just bolt the ball screw end holders straight to the 2060 aluminium rod with correct spacers? Would save on weight and possible look better.
James Bruton Thanks for replying! But what I meant is to turn the holders 90°, so the holders monting screws are facing the 2060 and not to the side. Then you could bolt it straight to the 2060 and there wouldn't be need for the big side panels.
So instead of using endstop why not dectect when the encoder stops turning despite the motor wanting to. That way it hits slowly and then you know. Kinda like how the trinamic drivers do with steppers just with a brushless motor. Plus if you can change the odrives max amps on the fly you could lower it super low for the homing process.
So you're just counting the encoder rotations? But if someone rotates the screw a little bit the robot won't even notice? I guess you can always use the end switch to reset... Wouldn't it be better to have a sensor for the leadscrew that measures the position of the joint..?
Do you think an even stress optimised bar design would be awesome for a light metalprinted exobone like kuka robots structural hull? "There are no stupid questions?" "Answers disguised as Questions?", "Who axe's?
That's gonna be a great project🤖
I think the motors can be cut in half size and driven bei an FOC inverter. The ball screw can be a 12mm. I use those in CNC Z Axis as well. Any experiments going in that direction?....
Your video are just so awesome all the time (very average comment but you deserve it) :)
Regarding prototyping material waste, for bits that don't make it to the "Bruton Museum of Engineering Design" (aka the workshop wall), do you recycle your plastic e.g. through 3D filament recyclers, local recycling programs, or any other way?
im liking this video before iv even watched it # qualitycontent
Great video! But as always I find the music a bit distracting. I liked the old fashion just you talking. Other than that keep up the great work!
Prof.Porpoise, i totally agree with you. Please james, stop making music videos just add som working sounds.
I didn't even notice the music. I think it's fine as is, and adds to his production value.
Thanks for the video.
I still hear a high pitch sound from when the motor is set to holding power. Anyone else hear it?
Nope...
....so very old :(
I would leave it on as an audio indicator that the joint is in hold mode, since this is development.
@@JacobianMatrix9 The only way to get rid of the noise, at least our ablity to hear it, is to use a pulse frequency that is above human hearing in the mid kilohertz range. This requires higher end micro controllers and motor drivers. Currently, being able to hear the motors is a benefit while developing the gait and standing software since you can hear the motors engage.
Awesome project! )
But what about more dynamic movements, like running or jumping? Will it be able to work faster if to change motor->ball screw gear ratio (the motor seem to be really capable)?
I like this video good job very inspiring
You're bloody amazing
Why do you need spherical head rods for something which only has one degree of freedom?...
possibly the easiest item he could find to do the job and would fit on the rods.
Yes, sure, but by giving more degree of freedom with those spheres, now he needs the little wheels to block the nut on the leading screw...
He mentioned this explicitly.
What made you select the outrunner motors? Is it purely cost-based or do they offer better performance over inrunners (I'm under the impression that inrunners have higher low-RPM torque than outrunners)?
Regarding the planned end-stops and using the speed mode of the ODrive to then find the endpoint of the encoder, this video might me interesting: ua-cam.com/video/q39G8FfuTjU/v-deo.html
On the shown delta-machine encoders with absolute position sensing are used which probably would work as a direct replacement for the current solution of openDog.
More details on these encoders can be found here: www.monolithicpower.com/en/ma730.html
I think this would facilitate the whole initialization procedure of the openDog.
very good
It will be perfect if you could make more compact and light weight drive system.
I really love this project, especially since it's open source. I have a couple of concerns with it though, like the fact that brushless motors don't have any holding torque, so when the robot gets up and is supposed to stay up, it might weigh enough to get the ball screw turning, resulting in it lowering. an easy fix would be to use leadscrews instead, but they also have more friction in moving, not only when stationary. also, for those v-wheels, why not put some cam/eccentric nuts on them? openbuilds does this too, and it allows you to adjust the tightness against the extrusion without having to move the entire leadscrew. it would be able to still take a similar amount of load. also-also, please don't use threaded rod as an axle! i have to see Ivan Miranda do this already too many times, just use some steel rod with clamp collars(3D printed or not) instead. they will be able to take more load, are way more accurate, are straight and can be used just as easily. this is again heavily dependant on how much it will weigh, but while getting up, when the nut is reaching the far end of the ball screw, it has a serious mechanical disadvantage, and idk if that motor is torqy enough at relatively low speeds to lift the robot's weight tat way. i think i'll try and build one pretty soon! great vid as always, but keep this in mind before progressing in the project.
oh like that, with the rotary encoder. the thing is, it doesn't have actual holding torque when not moving, it just compensates when it does move
Amazing things that you bild,and amazing sponsors that you have ,can you send me some of your trash....:)
Ever heard of counter sunk bolts?
James, nice video as always. Are you prototyping with pla or abs?
Thanks :)
why not use electro-hydraulic actuators like Boston Dynamics?
your viedos are so great. i am wondering what costs all the hardware. Just some gess, because i woud like to build it by myselfe.
Thanks for sharing
i feel like half of the "i'll fix that later" piles up and doesn't actually get done unless something breaks
how many parts on the exosuit were eventually supposed to be CNC'd?
'technical debt'?
Lol. That's the reason why projects like FreeBSD insist that you get it right the first time before you try and put it in the code.
All well and good, but as I'm sure you're away, nobody uses FreeBSD, and everyone uses Linux.
What's the main difference?
Linux is basically just made by organically adding things as and when people feel like it.
A lot of dubious, sort of working code gets put in with the idea that it'll 'get fixed later', but of course it never does. XD
Well, if it works... Does it really matter?
What's wrong with that? Proper British engineering. Made our car industry what it is today.
Thats the joke. British dont have car industry anymore.
It might be easier to just run the v-wheels through an eccentric nut so you always have adjustability and aren’t relying on the absolute tolerances of the printed ends.
Edit: further to that I have seen full grown men sitting on lengths of 2040 without it bending. I don’t think putting a hole through it will cause too much of an issue.
nice work mind your fingers it looked like you nearly lost one in the vid (i would use 1 lever arm from the top of the sliding block to the leg) but what do i know sat on my arse back seat advising
Hello James, which encoders are you using?
If im not wrong, he mentioned in the previous video that he used the encoders from the same brand with the controllers.
20:57 Isn't that what branches and tags are for???
One branch per episode would be a perfect solution for the situation James describes. It sounds like James is not fully aware of all of GITs possibilities.
Hi James!
Do you know the max force this actuator can exert? You can learn this just with a weighbridge.
If you give me this information, you'il be very helpful to me.
Thank you.
More than I can hold back with my hand, I haven't measured it.
Why not use branches or releases to manage code versions? It's easy enough to link to those and make it clear.
Is this mechanism same as the one used in most of dog robots
Like the MIT's Cheetah dog robots
or if anything is different please make a video about that
Why don’t you bolt the bewaring blocks for the spindle straight to the aluminium extrusion? It seems like quite A work around to maak A plate and mount them to that
Joren Snauwaert I guess the holes are not spaced out right for t-nuts. Therefore he would have to drill the aluminium and weaken the structure. Also the ballscrew would be awfully close to the extrusions. The wheels wouldn't fit.
Oh, i thought you made the bewaring blocks yourself. My mistake! Thanks for the answer anyway!