Eliminating Arm Wobble

[DinerKid]

This year team 1768 uses an arm driven off of an Andy Mark work-gearbox. The issue we are running into is that when the motor is stopped the arm has quite a bit of wobble back and forth, some due to imperfect gear meshing, some due to imperfect coupling of the arm to the shaft. We have added dampening to the arm in the form of a gas spring, however we still see a decent around of this wobble.

How do other teams deal with this for smooth operations of larger arms? Watching some of these other robots operate their arms so smoothly has me thinking there must be a handful of tricks that all need to be completed.

~DK

[kevin.li.rit]

This year team 1768 uses an arm driven off of an Andy Mark work-gearbox. The issue we are running into is that when the motor is stopped the arm has quite a bit of wobble back and forth, some due to imperfect gear meshing, some due to imperfect coupling of the arm to the shaft. We have added dampening to the arm in the form of a gas spring, however we still see a decent around of this wobble.

How do other teams deal with this for smooth operations of larger arms? Watching some of these other robots operate their arms so smoothly has me thinking there must be a handful of tricks that all need to be completed.

~DK

What kind of sensors are you using on your arm? If you can provide feed back to the controller about the position of the arm and the desired set point, you can then take that feedback and use the motor to stop the wobble.

Could you also describe what is wobbling? Is it flex in the arm or is it the worm gear as well?

[lucas.alvarez96]

What's the current setting on your motor controller? Have you got it on the brake position?

[DinerKid]

The drive-shaft to the arm is monitored by a pot, and we are using this data to bring the arm to the desired position. however, with the slop in the worm-gear box and through the attachment points between the actual arm and the drive-shaft.
The motor is set to brake mode.

I am partially looking for solutions to our current design as well as ways to design future arms from the beginning. To present the extreme end of this, I was watching chronicle today (maybe a repeat) and they showed the da vinci medical robot. It has massively long arms with absolutely zero slop in them. Clearly this is a high budget extremely precise machine, but what are the baseline ideas they they are working by in their designs to make this possible?

~DK

[MrForbes]

Pictures of what you have, would help us help you.

[kevin.li.rit]

What kind of resolution do you get on the pot? What kind of control loop are you using? What exactly is causing the slop?

[apalrd]

In the past, when we have been worried about lash on arms, we have intentionally increased the moving friction of the arm to keep it stable, and spring loaded the arm one way to force it to stay on the same side of the lash whenever possible.

For example, last year we pressed the arm axle into a block of delrin with an on-size hole to make our bearing. The delrin was slick enough for the axle to slide, but also had quite a bit of friction. We also sprung the arm up with surgical tubing to keep it on the same side of the lash in most cases.

[rsisk]

We are having a similar problem with our arm. We are adding a disc brake from a bicycle. When the arm should be stopped, we engage the brake. When the arm should be moving, we release the brake.

Obviously there are a lot of implementation details that have to be worked out.

We are using an avid bb5 brake, about $45 on amazon
The disc is fixed to our dead axle using a keyed hub bolted to disc
The calipers are mounted on the arm so they turn around the disc
A cylinder is mounted on the arm to close/open the calipers
And the magic happens in the programming
When the motor is moving the calipers a re released
When the motor is stopped the calers are engaged

[Chris Fultz]

For our 2011 robot, with a long arm to extend out to get the tubes, we used a gas shock to balance the arm. It is hard to see, but you can see it in the linked picture below.

I think there are two keys to stability. The gas shock on our arm was selected to balance the load, so the motor was just moving the arm up and down, and not really lifting it. This made it more stable in both directions. The other thing to look at is the "deadband" on the pot. You want it to have a range of "ok" readings for the arm location, otherwise the control will continue to hunt for the perfect pot value and the arm will go up and down while searches.


http://www.chiefdelphi.com/media/photos/36530

[DinerKid]

In the past, when we have been worried about lash on arms, we have intentionally increased the moving friction of the arm to keep it stable, and spring loaded the arm one way to force it to stay on the same side of the lash whenever possible.

For example, last year we pressed the arm axle into a block of delrin with an on-size hole to make our bearing. The delrin was slick enough for the axle to slide, but also had quite a bit of friction. We also sprung the arm up with surgical tubing to keep it on the same side of the lash in most cases.

Did the delrin hole wear over time or did it remain fairly constant for you?

https://www.youtube.com/watch?v=RqkAqpr-sHk

you can see the wobble at 3-5 seconds

I am working on pictures of the gearbox to the arm but it is just that, a drive-shaft (the output of the worm gear box) press-fit into a block of aluminum bolted to the arm.

The arm is currently back-heavy, however we have a mot of mass there and quite a bit of mass to stop rotating when we stop the tilt motor.

Edit: the talk of adding latex tubing to stay on one side of the lash, we are very unevenly weighted, the arm wants to fall back, I think just having so much moving mass is causing the arm to counteract this, the amount of latex we would need to add in order to completely stop this would likely just make the arm impossible to drive.

~DK

[nixiebunny]

I work on radio telescopes. They solve this problem by using two motors pushing against each other and on the arm. The control is done simply by issuing a command of torque X to one motor and torque X-K to the other motor, K being the amount of bias needed to keep the motors pushing each other to remove the backlash.

[Rob Stehlik]

This is actually a very good question, one that I would like to see some answers for as well. Dealing with backlash on a heavy arm is inherently a mechanical problem, so tuning your position control code won't get you very far. Obviously higher precision gear meshing and shaft coupling would help, but within the fabrication capabilities of most teams, this isn't a realistic solution.
I tend to agree with Andrew that adding friction to the system is a good approach. Generally speaking, once you gear an arm joint down to a reasonable speed you usually have torque to spare, so a little extra drag won't be a problem. If you can add enough drag to prevent the arm from wobbling on it's own without straining your motor too much, you're in business. In 2012 we had a lot of issues with overshoot and oscillation on our shooter turret. I ended up making a brake band out of plastic to add some drag to the turntable bearing, and this helped a lot. It's best to have some method of adjusting the amount of drag, so you can set it to the right amount, and compensate for wear over time. A simple way to check that you aren't adding too much drag is to measure the motor current while it is running.

[MrForbes]

This is the worm driven gear

http://cdn3.volusion.com/vyfsn.knvgw/v/vspfiles/photos/am-0933-2.jpg

It is connected to the shaft with a pin.

There's no way you're going to get that massive arm to be solidly attached to the worm gear using this stuff....it is designed to turn a very small chain sprocket, in it's original garage door application.

[gurellia53]

In 07, we used a motor driven lead screw to move a linkage connected to our arm. The screw did a good job of holding the arm stable. We also had some gas springs to offset the weight of the arm. The arm behaved like it was weightless before we attached the linkage.

[Travis Schuh]

We are also very concerned about backlash on our arms. This year we achieved some pretty good backlash performance. We used a large chain reduction for our first reduction, and then on the first shaft we made a custom shaft where we sized the hex shaft as a press for the sprocket on it and the gear for the next reduction. The Vex Pro hex in their hex broaches gives a lot of wobble, I think we sized it at .507 for the light press, a big difference from the .500 undersized hex they sell. We use #35 chain because not only would we be on the edge of #25's strength, but #35 chain is much stiffer than #25. This isn't backlash, but does impact the dynamic stiffness of the arm system. A classic example is looking at 968's arm in 2005 compared to 254's that year. 968 ran #25 and was in spec, but it added a low natural frequency to the system that was apparent when the arm stopped. 254 ran #35 and didn't have this problem. Another thing we do is our encoder is mounted via MXL timing belt off of the chain reduction shaft so that what little backlash we have is practically all behind our encoder.

[fr05ty27355]

A good solution I have seen is to add a system with a piston-actuated bolt that slides into a notch to physically lock the arm in a certain position. Good examples are 254's Shockwave and 118's 2014 robot.

http://www.team254.com/photos/72157634878291642/9410678832/

https://ccisdrobonauts.org/?p=pictures

You can see it in action at about 9 seconds in this video:
https://www.youtube.com/watch?v=PtRewwr59d8

[cadandcookies]

A good solution I have seen is to add a system with a piston-actuated bolt that slides into a notch to physically lock the arm in a certain position.

Main thing with this method is to MAKE SURE that you aren't stressing the piston's shaft anyway other than linearly. This means don't use the piston's shaft as the bolt, isolate it.

This was the main mechanism that caused my team copious amounts of pain at Northern Lights. I'm not really sure if there's a better way to do it and genuinely lock it in place, though.