I have saved multiple Point features in my Installation. From what I understand, these are cartesian poses, without any prefered IK solution saved unlike a Waypoint.
There seems to be an issue with Polyscope, selecting “Move here” in the Installation->Feature tab will select random IK solutions, often with very large joint reorientations and often with flipped axes, even if the active TCP is close from this Point. It makes it impossible to jog to these Points reliably.
For double checking I wrote a quick program with a single Script block and using:
movel(get_inverse_kin(myPoint, get_actual_joint_positions()), 0.25, 0.25, 0, 0)
And the robot will consistently move perfectly to the closest IK solution, while clicking the “Move Here” button in the Feature tab for this Point uses random IK solutions. I guess this is a bug, or rather forgetting setting the qNear as current joint pose? Seems like it is using a random seed instead of the current joint positions as seed, because it’s also not deterministic, and clicking again often gives a different IK solution… (I guess it’s a numerical solver, since you can pass it position and orientation error thresholds?)
Polyscope 5.19