Awesome, James! But if you use belts alone (with or without pulleys), then the holding current ought to be considerable. Excess heat might be a problem. Have you given thought (I'm sure you have) to using a worm gear in any way? They may introduce some backlash, but IMHO the torque and holding strength are just what you need in that arm. Looking forward to see the redesigned hand! Cheers from Brazil!
Why use springs? Now the force is always limited by the spring constant and how far the string is compressed or stretched. I would limit the torque on the motor dynamically.
You could mount both motors on the central arm stand. The first motor is then mounted on the stand and transfer power to the first elbow. Then use a wheel to transfer the power from the other motor to a pulley on the first elbow and then out too the outer elbow. That could be mounted on the upper part of the elbow and the other on the lower part of the elbow.
I would recommend using a couple Wheatstone Bridges epoxied to your aluminum extrusions. You wouldn't have large joint deflections when loads are applied. Just my 2 cents...
How about using a Magnet on the brushed motor's shaft and measuring the rotation of the magnetic field. There are Sensors out there, which can do exactly that. You just could glue on a magnet onto the small stump of the shaft. Thanks to the planetary gear you still should have plenty of resolution anyway!
Why not use a spiral springs for the joints? You know those springs in tape measures. Also, why not 3D print cross roller bearings instead of using those small bearings? It’s cool that you’re using springs in conjuncture to servos, I would have thought there were ways to measure the current being put into a servo as a way to measure the force being applied to them rather than using springs.
Using the mechanism with the black bits pulled together with springs, trapping the bit of plastic from the outer arm, I see a possible issue. You will not be able to measure the force applied until the force is large enough to stretch the springs. That is, small forces will not register at all.
Hello James, while I'm not suggesting that it's the best solution for this particular problem, have you explored compliant mechanisms and flexibility? I thought that if you have a way to use multiple filaments, you could try mixing flexible filament with rigid. I find your design process really interesting, and I'd love to see what you would do with it!
If you start to chain up multiple of these joints please stay far away from them ... Each moving joint will also apply a force on all neighbouring joints. If they are all controlled via PID controllers they will become unstable and start oscillating like crazy. I wonder if you get this to work as stable as your other robots. I think it is extremely difficult
A lot of posts have mentioned using a motor in place of the spring. Your response is always that it involves 'back driving' the motor. Can you explain why that is not something you want?
For the brushes motor it needs to encoder to drive the motor phases at the right time/position. For the brushed motor it will be better with the encoder on the motor because it'll be more precise - otherwise any backlash will cause the motor to oscillate.
@@jamesbruton I agree with you on both points. However, if the brushed motor is going through that 72:1 gearbox, there is a possibility that the resolution of the encoder will be coarse enough to damp the motor oscillations, especially if it is further geared down by a belt or gear after the gearbox. Perhaps I am wrong, but I would be surprised if the encoder has a resolution of greater than 140:1. But, in the case that it does, you could drill and tap a hole in the rotor shaft on your lathe (I think I remember you saying you have a lathe) and then locktite a standoff into it.
It's too bad you can't just sense the amount of current the motor driving the joint is using in order to calculate force. Certainly would simplify the mechanical part.
@@jamesbruton if you were sensing current couldnt you just have the motor drive however it needed to in order to maintain the same force? I can see in very high reduction gearboxes binding due to internal friction possibly being an issue but that's about it (but then again I'm not the robbit designer here:)
You do realise there's no second N in Turnigy, right? It's just TUR-ni-jee. Edit: I should say, I do like the videos. I thoroughly enjoyed your exosuit and BB-8 series. But yeah it's not Turningy :P
is it also possible to make it electronically force driven instead of mechanically force driven? By measuring and controlling the current or something?
hi james, you are on a good way -- following what robotics research already did and is still doing. the basic history goes like this: stiff joints, then all force controlled: soft joints, variable stiffness joints check these research robots and the publications around them: somewhat soft joints (~10kNm/rad): www.dlr.de/rm/en/desktopdefault.aspx/tabid-11427/#gallery/29202 variable stiffness joints: www.dlr.de/rm/en/desktopdefault.aspx/tabid-11666/#gallery/28234 soft joints (with mechanical springs): www.dlr.de/rm/en/desktopdefault.aspx/tabid-11826/#gallery/28795 keep up your "research"!
This seems like a rather complex way to measure something you already know? For an electric motor, phase current (note phase current, not supply current!!) is proportional to the torque the motor is generating (linearly proprotional in fact until saturation is reached). So if you operate your motor drives in order to control phase current, you have a "force controlled" actuator already. In software, you just wrap than inner torque control loop up in a velocity (and or position) control loop and bingo, you're done........
@@jamesbruton assuming your gearbox doesn't have such a high ratio as to make it mono-directional (ie like a worm drive gearbox that can only transmit torque one way) then if your motor control is in torque control mode, the motor drive can only apply the torque you ask for (and limit too) With positional control, as the actual departs from the target (ie the position error grows) then drive will apply more and more torque (phase current) to try to restore the position, up to the torque limit set in the drive. But, in torque control mode, the torque is limited by the torque control values and settings, and you just have to decide how "hard" you want your position control too be. And as that is a separate control loop outside the torque control loop, that can have its own unique calibration
The hardest part of making a force driven robot for James was mastering the force, since Anakin killed all the Padawans in the Jedi Academy -_- He had to watch The Empire Strikes back like 10 times to get enough tips from Yoda.
Depending on the format, you'd probably want to try Blender before you do anything else. Export it as an STL file. After that you need slicer software, but if you have a 3d printer around you should have that anyway. (and if you're using a printing service they'll probably do that bit for you and just ask you to send them the STL file...) Of course, this doesn't guarantee the model is designed correctly for 3d printing, but fixing that would depend on you having semi-decent 3d modelling skills, and that's not really a software problem.
@@KuraIthys I am used to use Autodesk inventor in high school last year. Now I just bought my first laptop and am looking to buy a 3d printer aswell. My older cousin has one maybe I can talk to him. Thanks for the info.
Oh james I know you're not welding without ventilation or a mask, that's going to tear you up inside, especially if you weld certain metals, like stainless steel, which gives off horrid fumes when welded
Of course, I know what a force driven robot is!
[Slowly hiding his toy light saber]
please give this guy a filament recycling machine, this guys prints so much more than any youtube ive seen
Awesome, James!
But if you use belts alone (with or without pulleys), then the holding current ought to be considerable. Excess heat might be a problem.
Have you given thought (I'm sure you have) to using a worm gear in any way? They may introduce some backlash, but IMHO the torque and holding strength are just what you need in that arm.
Looking forward to see the redesigned hand!
Cheers from Brazil!
Once ROS perfects self-awareness of balance and kinematics, you're going to have an army of robots!
" We could put 2 or 3 or 5" that's precision. Seriously you are awesome.
Prime numbers. :o)
Why use springs? Now the force is always limited by the spring constant and how far the string is compressed or stretched. I would limit the torque on the motor dynamically.
Great video. Ability to weld, opens up more alternatives. Can't wait to see you weld aluminum. : ) Cheers!
You could mount both motors on the central arm stand. The first motor is then mounted on the stand and transfer power to the first elbow. Then use a wheel to transfer the power from the other motor to a pulley on the first elbow and then out too the outer elbow. That could be mounted on the upper part of the elbow and the other on the lower part of the elbow.
That's going to be huge!
I would recommend using a couple Wheatstone Bridges epoxied to your aluminum extrusions. You wouldn't have large joint deflections when loads are applied. Just my 2 cents...
I concur. Also, use low-pass filters superglued to the 3D printed parts to attenuate any vibrations.
A force driven robot powered by the 40 foot tides in the Bay of Fundy would look awesome moving as smooth as openDog.
Woaa! What happened to your right arm?
Looks like a big scar!
Big nasty scratch!
How about using a Magnet on the brushed motor's shaft and measuring the rotation of the magnetic field. There are Sensors out there, which can do exactly that. You just could glue on a magnet onto the small stump of the shaft. Thanks to the planetary gear you still should have plenty of resolution anyway!
Drink every time James says 'force'.
I followed your instructions but ended up naked, unconscious and covered in hummus. Please advise.
Check out apetor, we drink whenever he says aaaah in his videos.
AMAZING AND AWSOME
James have you ever considered mentoring a FIRST robotics team?
you finally got new music!! thank you so much
Why not use a spiral springs for the joints? You know those springs in tape measures. Also, why not 3D print cross roller bearings instead of using those small bearings? It’s cool that you’re using springs in conjuncture to servos, I would have thought there were ways to measure the current being put into a servo as a way to measure the force being applied to them rather than using springs.
Please tell me you're finally working on the iron man suit up gantry, i've been waiting years for you to tackle it 😉
He needs to actually complete the suit first.
If you would use brushed DC motors the current would be liniair to the torque. So you could just drive your motor until a set limmit.
Using the mechanism with the black bits pulled together with springs, trapping the bit of plastic from the outer arm, I see a possible issue. You will not be able to measure the force applied until the force is large enough to stretch the springs. That is, small forces will not register at all.
nice stand looks great nice welds
this is looking good, your last few videos had some questionable designs and methods. Love seeing good CAD solutions.
Hello James, while I'm not suggesting that it's the best solution for this particular problem, have you explored compliant mechanisms and flexibility?
I thought that if you have a way to use multiple filaments, you could try mixing flexible filament with rigid. I find your design process really interesting, and I'd love to see what you would do with it!
That's interesting, I've not done much with that. openDog's new feet are flexible though
"We are Porg. Resistance is futile." -> who would have thought a green light is all it takes me to go there..
If you start to chain up multiple of these joints please stay far away from them ... Each moving joint will also apply a force on all neighbouring joints. If they are all controlled via PID controllers they will become unstable and start oscillating like crazy. I wonder if you get this to work as stable as your other robots. I think it is extremely difficult
Who else wants him to show us an update to his exoskeleton?
There isn't one, it needs rebuilding as a whole new project
Ooo, this is going to be awesome!
A lot of posts have mentioned using a motor in place of the spring. Your response is always that it involves 'back driving' the motor. Can you explain why that is not something you want?
It's back-driving the gears in between that's bad because it puts lots of load on the gearbox.
@@jamesbruton So, is that just because you are going to use plastic gears?
How come you didn’t mount the motor inside the hub. Less parts and more sturdy, no?
AMS encoders to the rescue, You should be able to glue the magnet to the shaft as there is very little force.
yep!
Is it necessary to attach the encoder to the motor itself? Could you not drive it with the belt or the gear you are going to use to drive the arm?
For the brushes motor it needs to encoder to drive the motor phases at the right time/position. For the brushed motor it will be better with the encoder on the motor because it'll be more precise - otherwise any backlash will cause the motor to oscillate.
@@jamesbruton Won't the additional precision be lost when considering the backlash of the gearbox?
@@IrishSkruffles I'll be measuring the absolute position of the actual arm as well... otherwise I can't work out the force on the springs.
@@jamesbruton Ahh I see, thanks for replying
@@jamesbruton I agree with you on both points. However, if the brushed motor is going through that 72:1 gearbox, there is a possibility that the resolution of the encoder will be coarse enough to damp the motor oscillations, especially if it is further geared down by a belt or gear after the gearbox. Perhaps I am wrong, but I would be surprised if the encoder has a resolution of greater than 140:1.
But, in the case that it does, you could drill and tap a hole in the rotor shaft on your lathe (I think I remember you saying you have a lathe) and then locktite a standoff into it.
Hello James, I am really interested in building this robotic arm myself, may I ask what is cost for getting all the building materials? Thank you
Wow, Nice design
Thanks for sharing 😀👍
How do you her the power thou to the motors? Sliprings? Or just wired flapping in the air?
It's too bad you can't just sense the amount of current the motor driving the joint is using in order to calculate force. Certainly would simplify the mechanical part.
Then you would always have to back-drive the gearbox.
@@jamesbruton if you were sensing current couldnt you just have the motor drive however it needed to in order to maintain the same force? I can see in very high reduction gearboxes binding due to internal friction possibly being an issue but that's about it (but then again I'm not the robbit designer here:)
What is the software you do 3D in there? Can't find it linked in the description and I don't recognise it.
your videos are GREAT
Look promising
Hey James it’s me riley I met you loads and I recently saw u at hacksoton
Noob question but why not use stepper motors? The precise control seems like a good fit for robots.
Yes that could work well also. I just happen to have the other two - with encoder driven motors I can get the same accuracy.
I see, thanks for clarifying. Keep up the good work! Your videos are both entertaining and educational at the same time. @@jamesbruton
No more OpenDog?
looks good
Good vid
You do realise there's no second N in Turnigy, right? It's just TUR-ni-jee.
Edit: I should say, I do like the videos. I thoroughly enjoyed your exosuit and BB-8 series. But yeah it's not Turningy :P
Which material are you printing?
is it also possible to make it electronically force driven instead of mechanically force driven? By measuring and controlling the current or something?
Yes but that always means driving the gear train backwards instead of pulling on some springs.
what that cut on your arm
hi james,
you are on a good way -- following what robotics research already did and is still doing.
the basic history goes like this: stiff joints, then all force controlled: soft joints, variable stiffness joints
check these research robots and the publications around them:
somewhat soft joints (~10kNm/rad): www.dlr.de/rm/en/desktopdefault.aspx/tabid-11427/#gallery/29202
variable stiffness joints: www.dlr.de/rm/en/desktopdefault.aspx/tabid-11666/#gallery/28234
soft joints (with mechanical springs): www.dlr.de/rm/en/desktopdefault.aspx/tabid-11826/#gallery/28795
keep up your "research"!
I know - Cog had series elastic actuators at MIT in the 90's (80's?)
Why aren't you just using the current sensing to measure the force on the motor, afaik Odrive makes it available ?
It would involve always back-driving the gear train instead of a spring that's designed to be stetched
Maybe I'm being stupid but why not use a worm gear?
Yep that could also work.
Why force control? Wouldn't it make more sense to use position control and set a force limit for any given movement?
How could you set a force limit without force control?
It will do position and force control.
This seems like a rather complex way to measure something you already know? For an electric motor, phase current (note phase current, not supply current!!) is proportional to the torque the motor is generating (linearly proprotional in fact until saturation is reached). So if you operate your motor drives in order to control phase current, you have a "force controlled" actuator already. In software, you just wrap than inner torque control loop up in a velocity (and or position) control loop and bingo, you're done........
I don't want to back-drive the gearbox, because it'll break.
Also what happens when I push the arm back and I want the motor to follow?
@@jamesbruton assuming your gearbox doesn't have such a high ratio as to make it mono-directional (ie like a worm drive gearbox that can only transmit torque one way) then if your motor control is in torque control mode, the motor drive can only apply the torque you ask for (and limit too) With positional control, as the actual departs from the target (ie the position error grows) then drive will apply more and more torque (phase current) to try to restore the position, up to the torque limit set in the drive. But, in torque control mode, the torque is limited by the torque control values and settings, and you just have to decide how "hard" you want your position control too be. And as that is a separate control loop outside the torque control loop, that can have its own unique calibration
Ok, this is a 'Series Elastic Actuator'. thanks
How do decide what color scheme to make your different projects?
I thought you were going to make a force hub kind of like the exo suit's
James what did you do to your elbow?
rose bushes while doing fencing.
The hardest part of making a force driven robot for James was mastering the force, since Anakin killed all the Padawans in the Jedi Academy -_-
He had to watch The Empire Strikes back like 10 times to get enough tips from Yoda.
The force may have been strong with Luke, but he didn't have access to gearboxes.
Hi! Would it be possible to measure the position using the springs as variable inductances in an LC oscillator?
that's quite clever, though it probably wouldn't be precise enough.
Why not use a load cell to measure the force?
James, what is that scratch on youre right arm in 6:45? It looks painful. Are you fine?
From a rose bush when I was doing some fencing.
James Bruton installing fences or long pointy metal sticks?
Can anybody tell me name of the design software here?
I think Odrive has a built in torque control option instead of position or velocity control?
that would make things a lot easier
Yes but that involves always back driving the gear box instead of some springs that are intended to stretch
whoa
How I wish I had subtitles in Portuguese :(
Me too!
I have a question. I have a 3d model of a fallout power armor. So what free or cheep 3d software can I use to make it readable for a 3d printer?
Look at the makers muse yt channel he did a fallout tutorial a number of years ago for fallout 4
Depending on the format, you'd probably want to try Blender before you do anything else.
Export it as an STL file.
After that you need slicer software, but if you have a 3d printer around you should have that anyway. (and if you're using a printing service they'll probably do that bit for you and just ask you to send them the STL file...)
Of course, this doesn't guarantee the model is designed correctly for 3d printing, but fixing that would depend on you having semi-decent 3d modelling skills, and that's not really a software problem.
@@KuraIthys I am used to use Autodesk inventor in high school last year. Now I just bought my first laptop and am looking to buy a 3d printer aswell. My older cousin has one maybe I can talk to him. Thanks for the info.
@@jokersredace Does Autodesk Inventor not have an STL export option?
@@clonkex I don't remember. 4 years of playing around in it and we never got to print anything.
6:35 just like a synchronous machine
Hi
You love robots. I love chocolate. Can you build a robot out of chocolate?
How about a chocolate made of robot?
If you didn’t put a gripper on it you could put a 3D printer, CNC or laser.
You could actually do interchangeable heads if you wanted
Woof
next week
@@jamesbruton Woof woof :-) lol
Make the canada arm
That's a nasty looking scratch on your arm James
Oh ! I had not seen it so I went looking for it and you can find it @6:40. James, tell us what happened !
Rose bushes when I did some fencing in my garden
@@jamesbruton Ouch ! You need a robot to take care of that ! Cheers from a Québécois now leaving in San Francisco !
Oh james I know you're not welding without ventilation or a mask, that's going to tear you up inside, especially if you weld certain metals, like stainless steel, which gives off horrid fumes when welded
There's a massive orange extractor in there...
@@jamesbruton good man 👍
That poor drill bit
soldering around so much wood looks spooky
I feel stupid.
Why don’t you have a head on that dog????
one day
☺😊😀😁😂5 Smile's
Likee