PDA

View Full Version : 3DOF palettizing robot running on Arduino



aggrav8d
03-22-2014, 10:20 AM
5449
50cm reach. 0.05mm precision at full extension. Max load untested.

I used an UNO and two Adafruit motor shield v2. Limit switches so the software can home on demand. Understands GCODE thanks to https://github.com/MarginallyClever/Gcodecncdemo.

The IK code is written and being tested. I'm at the Seattle Mini Maker fair this weekend and I brought it along to keep me busy while my other machines show off.

I'm looking for the right encoders to give this machine some force feedback safeties and training abilities.

What you'd like to see this arm do?

CasperH
03-22-2014, 01:56 PM
How big is it, it's kind of heard to get an idea for the size. It has a nice weathered factory look, did you tweak the picture lighting? :wink:

It reminds me of the uArm:

http://forums.trossenrobotics.com/showthread.php?6557-uArm%E2%85%A0%EF%BC%9AAn-open-source-robot-arm-Project
https://www.kickstarter.com/projects/ufactory/uarm-put-a-miniature-industrial-robot-arm-on-your

jwatte
03-22-2014, 04:19 PM
This is surprising!
I would expect the backlash, and the flex in the wood, to be on the order of millimeters out at 50 cm (assuming this is laser cut plywood, and the length between joints for each segment is 25 cm) 0.05mm is 50 micron, or two thousandths of an inch. Many all-metal milling machines don't get that kind of precision!
Could you show how you calculated (or measured) this value?

aggrav8d
03-22-2014, 08:02 PM
This is surprising!
I would expect the backlash, and the flex in the wood, to be on the order of millimeters out at 50 cm (assuming this is laser cut plywood, and the length between joints for each segment is 25 cm) 0.05mm is 50 micron, or two thousandths of an inch. Many all-metal milling machines don't get that kind of precision!
Could you show how you calculated (or measured) this value?

Yes, backlash is a serious problem. I guess I should say that the motors and the gearing and the electronics can - in theory - achieve as much as 1/20th mm precision. I'm working up to it. I'd be happy with repeatability <1mm for v1.

Yes, it's very similar to the uArm. I don't like uFactory. I don't like their methods. I don't like their claims. The only thing they've done right is given me the kick in the pants to finally stop talking and get building. This arm is just the start.

aggrav8d
03-24-2014, 05:27 PM
Part files: http://www.thingiverse.com/thing:280263

Firmware: https://github.com/MarginallyClever/arm3

More info: http://www.marginallyclever.com/blog/2014/03/building-an-open-source-3dof-palletizing-robot/

aggrav8d
03-25-2014, 03:11 PM
Writing "HELLO WORLD!" had to create the gcode by hand. :P


https://www.youtube.com/watch?v=y1-ZIk8nOSQ

jwatte
03-25-2014, 07:53 PM
You need to display the drawing in full frame when it's done :-)

Th232
03-25-2014, 09:08 PM
So what's the actual precision like, and could you improve on the pen holder (saw it wobbling a lot)?

I'd like to see a full shot of the drawing as well.

aggrav8d
03-26-2014, 12:35 AM
jwatte - the original post on my site includes the result picture close up.

th232 - I just switched to a arduino+RAMPS controller with 1/16th microstepping and the precision is great, plus all the crunching noise is gone. A better pen holder would greatly increase the print quality. The pen holder was intended for a Sharpie, but the grip was so tight that it was smashing the pen into the table and damaging the nib. I used the next pen I could find that didn't drop through the holder, which is what you see here.

I printed that pen holder just to show an application. The only tool I'm interested in making is an automatic tool changer. I see the potential for ANY tool, so focussing on just one seems like an poor use of my time.

Th232
03-26-2014, 03:07 AM
Good to hear the control is improving, I look forward to seeing future videos of it.

Maybe I'm missing something though, but looking at your pen holder I'm not sure how grip tightness would cause the pen issues. Shouldn't that issue stem more from control and resolution of the Z axis since all the holder is doing is applying a radial force on the pen?

aggrav8d
03-26-2014, 12:40 PM
Maybe I'm missing something though, but looking at your pen holder I'm not sure how grip tightness would cause the pen issues. Shouldn't that issue stem more from control and resolution of the Z axis since all the holder is doing is applying a radial force on the pen?

The grip on the pen is very tight. If the arm comes down too far it smashes the pen into the table, potentially damaging the table, the pen, or the arm. Ideally the pen holder keeps the pen from wobbling off the desired axis, stops it from falling out the bottom of the holder, and lets it "rise" if the arm gets too close to the paper. That way as long as the arm doesn't hit the table nobody gets hurt.

I just added G28 auto-homing to the source code. Today's goal is to film the improved movement, then fix some terrible design decisions that make assembly and maintenance really hairy.

jwatte
03-26-2014, 03:50 PM
For a good pen holder, you want a spring-loaded tube that holds the pen. This allows you to be off by a little bit without it breaking the pen.
However, with 0.05mm resolution, clearly you can get a good pen/paper interface? ;-)

Th232
03-26-2014, 06:48 PM
The grip on the pen is very tight. If the arm comes down too far it smashes the pen into the table, potentially damaging the table, the pen, or the arm. Ideally the pen holder keeps the pen from wobbling off the desired axis, stops it from falling out the bottom of the holder, and lets it "rise" if the arm gets too close to the paper. That way as long as the arm doesn't hit the table nobody gets hurt.

I just added G28 auto-homing to the source code. Today's goal is to film the improved movement, then fix some terrible design decisions that make assembly and maintenance really hairy.

It's your project, but I still think it comes down to a z axis control issue since, as you note, the whole problem starts "if the arm comes down too far". How much do you estimate your actual precision is right now with the 1/16 microstepping?

In the meantime, if you haven't made the next video then perhaps you can wrap the pen with some blu-tack or even just tape to help it fit in the pen holder better?

aggrav8d
03-27-2014, 01:56 AM
Looks like I've calculated the gearing wrong. 45 teeth on the big gear, 9 teeth on the small gear, ratio is 5:1, not 9:1. at 50cm reach the circumference is 2*r*pi or 314. the stepper motors are 400 steps per turn at 1/16 microstepping. that should give me 314/(400*5*16) cm per step at full extension, or 0.098125mm per step. 1/10th instead of 1/20th.

I just posted an update on my blog about what I've achieved today and how. in the morning when the light is good I'll film another test.

Xevel
03-27-2014, 06:17 AM
That's the resolution, not the precision.
To paraphrase a smarter man than I (on digital resolution/precision):
"Resolution is the number of bits that wiggle, precision is the number of bits that mean something."

With wooden parts all around, I would think the actual precision will be maybe a few mm.

Cool project nonetheless :)

jwatte
03-27-2014, 11:56 AM
You also have to consider dynamics, which includes both backlash, bowing, and inertial flex.

aggrav8d
03-31-2014, 04:26 PM
Updated video with quieter drivers.

https://www.youtube.com/watch?v=ObzyILAJFGQ

aggrav8d
04-02-2014, 10:45 AM
Updated the thingiverse files. Easier assembly and maintenance. Add bill of materials.

aggrav8d
04-10-2014, 05:03 PM
https://www.youtube.com/watch?v=nYONZPmyZyg
Training software synchronizes with the real machine. I'll be rolling out this software later to all my other robots, including hexapods, stewart platform, delta robots, etc.

http://learn.marginallyclever.com/index.php/Arm3_v1
Exploded assembly diagrams

5505
updated picture.