How still can a robotic arm hold itself?

John Welden · Dec 1, 2016

How still can a robotic arm hold itself?

Let's say it has a reach radius of 850mm and it's supporting an object that weighs 3kg. Can robotic arms hold a position with very little vibration?

Sorry if this is a dumb question. I don't know anything else about this stuff.

Milland · Dec 1, 2016

Depends on the weight class, the servo control and gearing mechanism, and what other forces are involved (is the thing it's holding moving? Is there floor vibration?). But yes, if the sizing and floor mounting is right you can have a robot arm hold something (functionally) still.

Halcohead · Dec 2, 2016

Seconding MIlland but if you want more detail and/or real numbers:

Unless the robot is doing active damping/control (unlikely for environmental vibrations), you can assume the best case scenario is it can lock every joint when stopped (due to no-backdrive mechanisms, powerful servos and/or actual brakes).

Then you can model the arm as a series of cantilevered beams bonded* end-to-end and from this calculate the stiffness of the arm and natural frequencies of the arm+load.

*This essentially assumes every joint is infinite stiffness which is obviously optimistic, but it'll get you in the ballpark and without active damping the arm definitely won't outperform this.

Depending on what sort of vibration/frequencies you're worried about, you could add a tuned mass damper (think devibe boring bar) to the end of the arm. But this seems exotic. Probably better to just buy a bigger arm.

hanermo · Dec 2, 2016

I worked on this with ABB robots.

The (midsize, 8-16 kg capacity )robot will position within about 0.1 - 0.3 mm total error.

The mechanism is steel via many joints, and if you push on the end effector, it will bend, of course.
I am not sure how fast they lock, but when not commanded to move, I believe they stay perfectly still, ie less than 0.01 mm movement.

I am pretty sure they have no servo jitter.

My ac servos have no jitter, none.
After == 0.5 secs, the servo locks, at zero count error.

They don´t hunt between positions like some older brushes servos did.
Looking at raw encoder counts from the output signals, a flicker of 1 count is observable.
The servo drive ignores this, until it gets input signals to move again.

Same behaviour on 400W cheap ac servos, and bigger 750W and 2.5 kW 220V ac servos.
3 different manufacturers.

Hope this helps.

Greg White · Dec 2, 2016

would how much money one had to spent enter into this?
Gw

Tonytn36 · Dec 2, 2016

Greg White said:
would how much money one had to spent enter into this?
Gw

3-80 kg robots will run in price comparable to a new vehicle. Smaller robots in the sub-compact pricing class, larger robots in the mid-sized pricing class, depending on manufacturer. Robots in the 100 kg and up are in the luxury car pricing class. You can get very significant discounts for volume purchases.

Greg White · Dec 2, 2016

So, will more money
Buy a stiller bot?
Gw

motion guru · Dec 2, 2016

Greg White said:
So, will more money
Buy a stiller bot?
Gw

There are robots with secondary encoders and you pay a little more for these robots . . . but the real money comes in when you add a 3-axis accelerometer and then write code to accomplish active damping.

I wouldn't even try this with Fanuc or ABB. Kuka is the brand for this kind of "stillness" with environmental disturbances and variable process loads.

It all depends on the mounting / foundation / environment. I had a customer ask me to use a robot to control ground face width of +/- 0.0005 inches on a 30 degree inclined angle. There is not a commercially available robot that can hit that spec or anywhere near to it.

The only chance we had of hitting that spec was linear motors with stiff mechanisms and close to 4 million counts per inch feedback encoders.

Jashley73 · Dec 2, 2016

John, my previous job we had a Fanuc R2000-iB 165F - 165kg with a close to a 2 meter radius? This was a bruiser, with a near 400lb payload... With little outside influence, I'd say it would hold very still itself. When it would complete a move, the brakes would activate, which means the joints were locked up stiff. Granted, the brakes lock the servos, so if there's any slop in the gears and/or joint mechanics, then the servo-brakes can't fix that. Even still, the joints locked up stiff...

What I will say however, is that even though the joints were stiff, I wouldn't call the entire robot "rigid" like a machine-tool is. With it parked in a fairly "crouched" position (as opposed to fully extended) you could smack the end-of-arm-tooling with your hand, and feel the thing vibrate for a second or so... I'm guessing if you recorded this with an indicator, you might have .010" peak-deflection, and maybe average +/-.005" deflection for 1-1.2 seconds ...?

That's why I said "absent of outside forces..." If you isolate the robot itself from outside forces, then I think it will hold itself very still. Maybe not the scientific answer you were looking for, but that my "from the floor" experience.

Tonytn36 · Dec 2, 2016

I think we need some clarification on how much movement is allowed. EVERYTHING bends....and vibrates...... scale is the key here.

CarbideBob · Dec 2, 2016

Jashley73 said:
......
That's why I said "absent of outside forces..." If you isolate the robot itself from outside forces, then I think it will hold itself very still. Maybe not the scientific answer you were looking for, but that my "from the floor" experience.

Just siting there holding a weight I'd expect zero movement. No motion, not a sub-micron once the sag is out of the joints and beams. It should be dead silent.
Such a system should not be dithering, hunting, or vibrating at all. If it does the control loop is just plain puked up but yes, seen this done even on expensive machine tools.
Now make it do something or have a big forklft drive by, different.
33 inches out of a base rotary joint calls for very insane high res if you need a tenth once things start moving so you tend to go with Cartesian robots there and it begins to look like a machine tool not a robot.
A 40 inch HMC or VMC, how much should it oscillate sitting and waiting? It is a robot. They come in many flavors.
Arms most think of are just A and B rotaries. Same deal, same guts and servos. How tight can you hold the tolerance on a A axis part?
It's a joint length to base resolution problem along with beam deflection under load when actually working. Even just mean ole mister gravity screws things up accuracy wise.

Free of outside influence it should hold within a level you most likely can not measure, it will not be accurate or where you want it to be.

Of very little help to OP and maybe makes no sense so feel free with the dope slap here John.
Bob

tygerdawg · Dec 2, 2016

You will only know for certain with a test. If really important, then get the Applications Engineering Department of your robot-of-choice's manufacturer to perform a test.

Else, the easy answer is "hold a position within the repeatability spec of the robot." Which will be listed as +/- <some quantity> your robot's data sheet.

SteveBausch · Dec 3, 2016

It's steadier than a human being, since it doesn't have a pulse.

It doesn't have a heart either, a trait it shares with many corporations.

:-)

IntegralCirc · Dec 5, 2016

We don't use any bots in my warehouse but a buddy of mine works in an assembly warehouse where they use them. They pretty much pay top dollar to make sure they are at the pinnacle of precision and accuracy.

IronReb · Dec 5, 2016

What is active dampening on a robot?
A process that acts against forces on the arm?
If something is acting against the part the robot is handling, like say intermittent air blast from a tool the robot is holding, it compensates
for it with "active dampening"?

Tonytn36 · Dec 5, 2016

IronReb said:
What is active dampening on a robot?
A process that acts against forces on the arm?
If something is acting against the part the robot is handling, like say intermittent air blast from a tool the robot is holding, it compensates
for it with "active dampening"?

In a nutshell, yes. They also have available "alignment compensation" or "clamping compensation", where the robot is allowed to deviate from the programmed path / position such as for inserting a pin in a hole. You don't have to align it perfectly, the robot can "adjust" to the forces the pin is exerting on the arm to self-align. Same goes for a part clamping situation. The robot can move while the part it is holding is clamped.

CarbideBob · Dec 5, 2016

SteveBausch said:
It's steadier than a human being, since it doesn't have a pulse.

It doesn't have a heart either, a trait it shares with many corporations.

:-)

They don't have a heart but for sure they have a pulse which is a combination of all the time constants and hard to nail.
Bob

Spyderedge · Dec 5, 2016

John Welden said:
How still can a robotic arm hold itself?

When welded together, pretty solid

Jashley73 · Dec 5, 2016

Tonytn36 said:
In a nutshell, yes. They also have available "alignment compensation" or "clamping compensation", where the robot is allowed to deviate from the programmed path / position such as for inserting a pin in a hole. You don't have to align it perfectly, the robot can "adjust" to the forces the pin is exerting on the arm to self-align. Same goes for a part clamping situation. The robot can move while the part it is holding is clamped.

This was one of the coolest demos at Fanuc NA. They have a big robot insert a ground-shaft with a keyway, into a super-tight fit hole. I believe it's coupled with a 3D force sensor (accelerometer?) and adjusts to align the shaft inside the hole. Outside the demo cell, they have the same shaft & hole sitting there for the observer to try. There can't be a 1/2-thou clearance between the shaft and the hole, yet the massive robot does it with absolute finesse & grace. So impressive...

(Re-reading that, I'm waiting for the jokes to pour in...)

Tonytn36 · Dec 6, 2016

Jashley73 said:
This was one of the coolest demos at Fanuc NA. They have a big robot insert a ground-shaft with a keyway, into a super-tight fit hole. I believe it's coupled with a 3D force sensor (accelerometer?) and adjusts to align the shaft inside the hole. Outside the demo cell, they have the same shaft & hole sitting there for the observer to try. There can't be a 1/2-thou clearance between the shaft and the hole, yet the massive robot does it with absolute finesse & grace. So impressive...

(Re-reading that, I'm waiting for the jokes to pour in...)

I'm sure Fanuc is similar, but on Kawasaki, you can actually specify what direction(s) the robot is allowed to deviate and of course the limit it is allowed to also.

How still can a robotic arm hold itself?

Diamond

Diamond

Stainless

Titanium

Titanium

Diamond

Titanium

Diamond

Titanium

Diamond

Diamond

Plastic

Stainless

Plastic

Stainless

Diamond

Diamond

Titanium

Titanium

Diamond

Similar threads