Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5

Not enough valid calibration points

#11
Just finished a 60 point calibration where 9 of the poses hid the SMR from the tracker. I had to raise the tolerance to 100mm to get it to calculate, leaving me with a mean error of 21mm. (report attached)

The manual has a page showing a "make mastering program" method using a button that I don't have. I'm wondering if I need to master this arm better first, or what else is going on. Maybe I should update the kinematics in the controller and try again?  I have a couple clone drives so I'm not worried about messing one up.

I just checked my SMR by mounting it on the tracker base nest and measuring it, then rotating it in the nest to various angles and the measurements are all within 0.02mm, at any angle or straight on.




   


Attached Files
.pdf   Robot calibration 2022-08-02 10h28m15s.pdf (Size: 37.77 KB / Downloads: 218)
.rdk   kuka22 - R4.rdk (Size: 1.05 MB / Downloads: 182)
#12
I used the Faro tracker utility to manually measure points on the tool flange, indexing A5 3 times, then index A3 once, index A5 again 3 times etc, and plotted the points in solidworks to measure my A3-A5 length (came out to 1001.95mm, compared to 1001mm expected accounting for the 45mm offset), I'll have to do it again with more points to get the A2-A3 length and the true offset but having the real thing closely match the model in RoboDK is assuring.

Still not sure why the calibration program isn't giving me good numbers, but I'll try mounting the tool again on Monday and put the SMR on it a little further from the A6 axis. Maybe I'll update the RoboDK model with my manual length measurements as well.
#13
Today I made another new rdk cell, moved the SMR further from A6 axis and tried again without the set measurements reference option. The base and tool went fine. Creating a new calibration and running it with the Follow Tracker option, the pointer was about 1.5" away from where it was "supposed" to be.

   

I tried the same method above only moving axes 2 and 3, to find the length between them. That came out to be 1046.4mm (nominal is 1050mm).

I updated that parameter here, is that right?: 

   

But rerunning the claibration, it still expects the SMR to be 1.5 inches away from where it is.

If it's a good check to use before running 60 measurements, that's convenient. I now believe the nominal lengths between A2-A3 and A3-A5 are accurate, my laser tracker passes startup and operational/health checks. What else can I do here to get calibration to work?


Attached Files
.rdk   kuka22-R5.rdk (Size: 1.05 MB / Downloads: 178)
#14
I just tried the compensations in this section https://robodk.com/doc/en/Robot-Calibrat...rackerTest and my tracker passed, so I don't believe it's the tracker at this point.
Results:
   

Rerunning the entire calibration starting with Base, then Tool, then letting the tracker lead me to the first calibration pose, I'm nearly 6" away this time. If I remember right, I should be 2-5 mm away.

I guess now I'll triple check the mastering, unless somebody has a better idea.

Mastering for J2 and J3 is as close as I can get it with a manual caliper.
#15
I tried creating a tcp on the robot outside of robodk and the error was 33mm, apparently my mastering was garbage. Running through that again today, until I can create a TCP on the teach pendant, then I'll come back to robodk and try again.
#16
Solved!!

I checked angles from manually jogging, and A5 was at 90deg when the controller thought it was at 70. I had the incorrect $machine.dat - corrected that based on the nameplate on the arm, and we're good to go. Check out these results!

   

Thanks for the help earlier Arnold, it made me suspect something was wrong with my physical arm versus the program.
#17
Excellent! Thank you for your feedback. These results make more sense. I understand joint 5 was way off it's zero value.

Do you have any suggestions for us to improve the calibration documentation? We are aware it is a bit outdated as we kept adding features without updating the documentation.
#18
(08-11-2022, 06:40 AM)Albert Wrote: Do you have any suggestions for us to improve the calibration documentation? 

It's actually pretty well laid out, I just rush through things when I'm reading. The mastering program procedure confused me (didn't know I had to click parameters, mastering, then OK to make the button appear), and the "follow tracker" checkbox isn't explained, but it essentially validates my cell on the first calibration pose.

It would be really handy to have a sample .rdk file to download to see how a cell is set up for calibration, with the measurements ref, tracker ref etc organized correctly WRT to any robot arm. 

I'm still not sure what to do with the base measurement reference that you suggested to ignore, at this point I'm skipping it entirely but the manual says to do that as step 1.

The KRC2 ethernet connection instructions are for a KRC4, but that's less and less relevant as time goes on! I just use the exported programs. Overall, it's a great manual.
  




Users browsing this thread:
1 Guest(s)