When I am placing objects in the scene, I find it extremely cumbersome to transform all of their shapes and subframes to the position I want the object to be in. It's really something that the user shouldn't have to take care of. My current workaround is to publish a temporary TF frame for the object, which is awkward, requires waiting for the frames to propagate, and isn't very stable.

I would like to replace the header object in the CollisionObject message to be a PoseStamped object called object_pose. All the shapes and subframes would be defined relative to that pose.

The current behavior would be easy to replicate by defining a neutral PoseStamped with the same header. With the change, it would be easier to shift the whole object, and in the future, attaching the object to a scene graph would be simple.

Does anyone see a problem with this design direction? I may start developing this soon otherwise.