I am looking for help teaching an arbitrarily angled plate as a feature plane for my UR5e with high accuracy. I am working on integrating a Cognex vision system on a pick and place and have to use a calibration grid on my angled parts plate. To teach the robot to this grid, I use a point end effector (screw in a plate) and teach the plane using the feature creation installation tab. I can tell just by teaching the TCP and the plane that these are not exact. Anybody have experience teaching feature planes with exact origins and axes with their robot. Thank you!
1 Like
What kind of resolution are you looking for?
The resolution on the camera is 5320x3032. In terms of placement accuracy as close as possible. Ideally +/- 100um in x and y. +/-250um may be acceptable.
1 Like
I have never tried this kind of precision. Is there a way to have a mechanical aid/guide for placement to allow for less strict location?
You can also try to edit and refine the plan manually.
I working on a position calibrating system for mobile robots.
It can align to a arbitrarily angled plate with a repeatability of less than 0.1 mm (x,y,z) and less than 0.3 mRad (Rx,Ry,Rz)
More info and video here:
1 Like
Teaching a high-accuracy plane feature can be tricky, especially when working with angled plates. One approach is to use a precise calibration grid in conjunction with your Cognex vision system. Ensure the TCP is accurately set and verify it by performing a test alignment. When defining the plane, use multiple reference points and validate them using a measurement tool to check for consistency. Small errors in the TCP or reference points can cause inaccuracies, so double-check all settings and recalibrate if needed.
The camera-TCP must be set with an accuracy that matches the accuracy you want for the plane.
This high accuracy Eye-Hand calibration must be performed before you have a chance to align your robot to the plate.
The Selfie Aligner comes with a new feature to automatically perform Eye-Hand calibration.