Camera as flange child

[marrts]

This allows for motion planning with the camera as a TCP

[marip8]

Why did this not work prior to this change? It seems like this change shouldn't be necessary

[marrts]

The tool_flange is the manipulator tip. The tool_flange is also a child of tool0 and the camera frame is a child of tool0. So Tesseract can't find the camera frame as a valid tcp link for the manipulator group ending with the tool flange. Alternatively the manipulator group can be changed to end at tool0 and I think everything should work. This was primarily an issue when IK Fast was being used, since it was generated relative to the tool_flange, but I think just KDL is used now, so the manipulator tip shouldn't matter.

[marip8]

I see; let's close this PR in favor of changing the tip link of the KDL solver.

[gavanderhoorn]

High-level comment: tool0 is not supposed to be a parent for any other frames. It's a leaf frame.

Other tool frames should be children of flange or a frame with a similar role.

[marip8]

There is a REP (or REP proposal) that specifies that, right? I think about 95% of the URDFs I've made or used have not followed that convention

[gavanderhoorn]

Yeah: Coordinate Frames for Serial Industrial Manipulators.

Specifically: Coordinate Frames: tool0 and Application-Specific Tool Frames.

I think about 95% of the URDFs I've made or used have not followed that convention

well it's not an active REP, so it's really just a convention written down in some .rst somewhere.