[LukeGaines]

The standard RDK calibration procedure uses joint positions to move the robot to different poses and take measurements. I understand why this is done but is it possible to run the same calibration routine but instead of using joint positions, drive the robot to a cartesian coordinate?

We would like to do this because our robot comes with a kinematic calibration of it's own that is on all of the time. Therefore, we would like to use RDK's calibration on top the controller's existing kinematic model to essentially "fine tune" it's calibration. This won't happen if the calibration poses are commanded via joint positions. It essentially bypasses the controller's kinematic model when done that way. We would love a way to test this out. Is this possible?

[Albert]

I understand you want to calibrate the robot on top of an existing calibration, correct? In my experience this won't work well. Added to that, it is not possible to load custom accurate kinematics in RoboDK for automatic filtering.

It would be better to use the nominal model as a reference and then apply one calibration.

What robot controller are you using? There is probably a way to disable the existing calibrated model.

[LukeGaines]

Ok. I had assumed it might just be a marginal improvement if anything but I was still curious to try. The original calibrated kinematics are from the factory (ABB) and run on the IRC5 controller. I have already disabled them to run the RoboDK calibrated model and this works well. I just wanted to see if there was any opportunity for further improvement. Thanks for your help.