Coordinate Frames "Pushing Down" with transforms

Feature request

• transforms function the same way as namespaces, allowing for pushing down of frame_ids
• loading in static_tf_params from a yaml file not directly from a launch file

Feature description

With ROS 1.0 namespaces you could "push down" different nodes into unique namespaces so that you could have multiple of the same-named nodes in different namespaces (i.e. /left_arm/hand, /right_arm/hand).

This concept is addressed in the tf2/CommonQuestions page however I don't necessarily agree with the argument here.

When creating a static transform structure for robot A, and you want to create another instance of that same robot but call it robot B - with the current naming structure you are required to rename every single coordinate frame and make it unique instead of just changing the name to robot_B. Why cant we use by default the same structure of pushing down like we do with namespaces?

Also, if we could add a feature to static_tf_params.cpp to load in the xyzrpy values from a yaml it would help a lot too I think. I've had a modified version of static_tf_params.cpp for the last year and it has helped standardize my tf structure and launch files - allowing me to use the same launch file for many different robots.

Implementation considerations

• follow the same structure as ros namespaces, just for the frame_id by default use the full namespace name
• add in functionality to static_tf_params.cpp file to load in the xyzrpy values from the parameter server....letting users load in a yaml file at launch to the parameter server using rosparam load

┆Issue is synchronized with this Asana task