[whansen]

I followed the instructions to update kinematics on our UR10 robot (created a program from scratch on the teach pendant, saved to USB drive, ran UR_LoadControllerKinematics.py script) and got a message "Warning! Something does NOT look right..."

The Robot Parameters screen tells me that default kinematics will give a 36.489mm numerical error relative to the updated kinematics.

If I align my end effector with the robot J6 flange using default kinematics, and then enable updated kinematics, the end effector in my RoboDK station appears to be offset approx. 22mm in the +X direction and rotated approx. -12 degrees about the Z-axis.

I have tried re-running the updated kinematics procedure several times and with two different programs created on the teach pendant, and the results are the same.

使用updated kinematics definitely improves robot positioning accuracy, but it is still imperfect. I realize some of the error is due to discrepancy between the station model geometry and actual as-built geometry, but I am wondering if there is a problem with the calculated DH parameters. Some of the values seem ridiculously large. I attached an image of the modified DH table.

My questions:
Is the offset appearance of the end effector normal behavior when using updated kinematics?
Is there some way to verify the accuracy of the calculated DH parameters?

I am running RoboDK v.4.0.0 13188-Release date 2019-09-27
UR software (PolyScope) 3.5.1.10661 (Dec 13 2017)

[Albert]

The accuracy of the kinematics extracted by RoboDK is the 36 mm numerical error that you see. It should be 0.000 mm.

Can you confirm you tried this on a UR10 (not a UR10e)?
Can you provide the URP file you tried to load?

I would recommend you to try again setting all position parameters to 0 (you currently have it setup to +60,-61,10 mm approximately).

[whansen]

谢谢,艾伯特。

• I can confirm this is a UR10, not UR10e.
  
• I am getting 0.000mm numerical error when I check the box for Use Controller Kinematics; I see the 36.489mm error when I uncheck the box.
  
• I reran the script with all robot position parameters at zero using two different .urp files, and I get the same result. I'm attaching an image of the generated DH file and one of the .urp files.
  

update_kinematics.urp(Size: 1.59 KB / Downloads: 395)

[Albert]

Great. It looks like the kinematics is imported successfully.

You see a warning message because the real controller kinematics deviates a lot from the norminal kinematics (36 mm, this is why you see the message "Warning something does not look right"). This will mostly impact the location of the robot flange, not matching the 3D model. However, the forward/inverse kinematics will match the robot controller's forward/inverse kinematics (same numbers you see on the teach pendant).

If you locate the UR_LoadControllerKinematics.py script (available in C:/RoboDK/Scripts) you'll see that this warning is automatically triggered when the real kinematics deviates more than 20 mm with respect to the nominal kinematics. You can change this tolerance if you need.

[whansen]

So if the X,Y,Z and rotation numbers on the teach pendant display match the numbers in RoboDK, then the kinematics must match.
By making my TCP coincident to the UR10 flange and setting my reference frame to "UR10 Base" in RoboDK, I was able to compare directly to the numbers on the teach pendant. Everything matched closely.

Thanks for your help!