Attached bodies in RViz don't show a trail

[Jntzko]

At the moment attached bodies are not displayed in the trail of the generated robot trajectory.
However, it would be nice to see them there.

[davetcoleman]

This should be possible since that Rviz display plugin takes as input a moveit_msgs/DisplayTrajectory.msg that includes a moveit_msgs/RobotState.msg that includes a vector AttachedCollisionObject[]

You simply need to add this notion to moveit/moveit_ros/visualization/rviz_plugin_render_tools/src/trajectory_visualization.cpp

[thummel26]

Presently working on this

[davetcoleman]

i suspect this image is misleading and collision objects are never attached - rather they are only in the RobotState for start and goal

[rhaschke]

@davetcoleman There is a fundamental issue in the design of moveit_msgs/DisplayTrajectory.msg: There is only a single field for the (initial) RobotState. However, if we a have a more complex trajectory, e.g. due to pick + place, the RobotState, particularly the set of attached objects, might change multiple times along the whole trajectory.
Hence, I suggest to extend DisplayTrajectory.msg to take an array of RobotStates (of same dimension as trajectory). What do you think? Probably this will break some things elsewhere?

[v4hn]

Hence, I suggest to extend `DisplayTrajectory.msg` to take an array of `RobotState`s (of same dimension as `trajectory`).

I gave this a bit of thought and I like the idea. Thoughts from my side: - Filling in the array should be optional and partially filled-in arrays should be supported to - I would like to encourage users to set `is_diff == true` for all RobotState msgs in there in a comment. Also to make clear that they *don't* have to fill in the whole robot state if they want to specify a changed attached object - The entries in the array should be applied (usually as diffs) at the end of each trajectory segment in `trajectory` - I would propose `changed_state` as a name @rhaschke It would be nice to have this, but someone has to add support for it to rviz. I don't see merit in adding it to `moveit_msgs` before there's a patch for the rviz interpretation too. Otherwise it's just bloat in the msg. If anyone wants to continue on the change proposed by @rhaschke please create a new issue or a pull-request for it. This issue is about a more basic bug.

[v4hn]

@thummel26 Did you manage to implement a patch for this issue last month?

[felixvd]

I think this can be closed.

[rhaschke]

Yes, fixed in #1766