Immanuel Martini

I can confirm that this fixes #11917.

The problem I observed was not the controller state but the state of the robot `Diffbot`, that is used in the tests. `Diffbot` has a start/stop functionality. Stop triggeres `Nan`...

Hi. > I think, the current tests should be sufficient. However, I was unable to execute them since I got a lot of unrelated errors. I tried to make some...

These lines introduced the bug: if not self._reference_frame == _DEFAULT_BASE_LINK: return _to_robot_reference(robot, self._reference_frame, self._goal) The actual change is that now the goal is always transformed to the root frame of...

It does change the behaviour of the API. Now it behaves as promised by the API docu.

@jschleicher We discussed several concepts for a backwards compability and scheduled its implementation. Until then, merging this PR is postponed.

I would suggest that we do not treat this as a bug but as a missing feature. Whilst I agree on the expected behaviour, this cannot be realized in the...

Shall we apply this on package or repo-level? I think we cannot rely on the stability of dependent packages. So the stability has to be ensured by our maintenance. Do...

Is it intended that we require `prbt_base` and `world` to be identical? I am not sure if I fully understand the meaning of these frames. At least it should be...

Yes, that sounds very reasonable. Thanks for the input. Actually, our Readme states that `prbt_base` is used as default.