Cartesian vs Joint Representation: or why a Joint move to a trained PR is not moving to literal joint angles in the PR?
What's the scoop on training a Position Register with respect to Joint moves?
I had a mid-point I taught as a PR and used in a program with a joint move. Later, I wanted to flip the knuckle on this robot due to interference issues, so I retaught the mid-point PR in the same X,Y,Z position, but with the new joint orientation that flips the knuckle. So it was the same cartesian coords, new configuration, and I didn't check the joint angles saved, but apparently it doesn't update the joint angles in this case. When testing, it didn't initially have a problem with the Joint move to the mid-point position, but I guess when coming back from a location in the other quadrant, the end-effector started a 180-degree flip. When checking the joint-angles, they only get updated at teaching time if you switch to Joint Representation.
I'm not satisfied that I really know what is going on here. Should I leave the Representation of the Position Register in Joint. I think all of my program moves to this PR are Joint moves, but I'm currently surprised it didn't update the joint angles to the flipped position when first saving it, and I'm also surprised that the joint move in one direction worked fine, but the joint move in the opposite direction resulted in it trying to move to the literal joint angles. Why didn't the first joint move try to move to the literal joint angles? The behavior is seemingly inconsistent.