Hello together,

we just started to test and integrate a UR10 and have several questions. 

Our setup:
UR10
Polyscope 3.5.1.1
CB3.1
Robotiq FT300
vacuum gripper with 4  suction cups
mounted on a rigid table


Task:
Apply a glass plate (10") with adhesive tape perfectly centered in a plastic frame. The visual gap on each side of the glass in relation to the frame depends on the manufacturing tolerances of the frame and yields to 0.5 to 1.0mm on each side.

Our plan:
1. manual: Fix plastic frame on a part support
2. manual: place a glass plate in the frame (already equipped with adhesive tape and a applied protection layer.)
3. start the process
4. Move to Point 10mm above the glass
5. Search in z until 20N reached due to let the suction cups compress a little bit
6. switch on vacuum
7. Search again in z due to the fact that the glass sometimes lift a little bit so we are on zero plane again.
8. Lift the glass 0.5mm to not contact and also overlap the glass and the frame a little bit
9. Check if there are any forces in x or y du to a contact at the edgeds. If so, move 0.2mm in the opposite direction to ensure there are no stress forces in the system.
10. Save this point.
11. Move to a external position where the worker can remove the protective layer from the adhesive tape.
12. Move to the saved point (remember: 0.5mm above the level where the adhesive tape sticks).
13. Check again for any stress forces in x and y and  if necessary release them.
14. Average the rotaion in Z (0.4Nm detection force)
15. Average the orientation in X (1.5N detection force)
16. Average the orientation in Y (1.5N detection force)
17. Force Move in Z to bond the glass to the frame.
18. Release vacuum
19. Move back to idle position

Our main problem:
We are not able to align the gripper perfectly planar to the glass plate. This causes that when we press down the glass panel one edge will contact to the frame prior to the rest. This leads to stress forces which displace the whole system a tiny little bit and the glass is not centered perfectly anymore. 

Our question:
Any advice in general ;-)
Especially: How could we realize a function that level the plate to the frame planar prior to average the center point. For instance search in Z for a small force an try to release rotation forces in X and Y. Any advice how to program that?

Salute!
Lars

@Lars have you tried monitoring moments in X and Y to re-align the end-effector? Do you have anything that gives compliance in your setup or is everything rigid? In my experience, you would need both compliance and moments monitoring to align with precision.

This was also our approach. But we couldn´t realize a function which gives us a good result. for example: 
If Moment in X; Rotate -X for 0.1°; move -Z; check again
How does the function work where one ore several movements are free to follow a unknown contour? Could this be a solution for us in any way?

We tried to get the systems as rigid as we can since we had more problems with a optoforce f/t sensor which is really flexible and not as precise as the robotiq sensor.