Make moveit_config compatible to moveit_configs_builder

[fmauch]

As mentioned in #917 our current config is not compatible to the moveit_configs_builder.

This PR aims at making our config compatible with it.

The current state is a working version which allows loading the description from the robot_state_publisher's topic and can be used in the moveit_configs_builder

Things still need to be done

• Lazy starting of move_group. Currently, it crashes when the description topic isn't yet published
• [ ] Add upstream issue about the crash when no description topic is provided
• Add moveit_servo

Things that significantly change in this PR:

• the tf_prefix has been removed. It didn't work correctly, anyway. Since there are joints inside config files such as the joint_limits.yaml and they to my knowledge cannot handle argument substitution during loading inside the configs builder, this doesn't work.

Things suggested in #917 (comment) I did not implement

• I did not add different IK solutions, as it seems not to be possible to pass them to the configs builder dynamically. As this is merely an example repo, I'd rather not add this. Users can still define their own when they implement their own launchfiles with the configs_builder in them.

[gavanderhoorn]

Unorthodox perhaps, but what about not using the config builder at all and creating a more ROS-1-like MoveIt config pkg? Tyler made a nice example showing how to do this, right around the time I was also looking into it.

See ROSCON 2023 - Retro ROS 2 Launch (and the package itself).

I'm wondering whether hiding what goes on in a launch file -- by encapsulating its 'insides' in a separate library -- and making it harder to customise a MoveIt configuration isn't actually doing a disservice to users.

Whenever we feel the need to start creating a reusable Python package because it feels like things are getting more verbose than XML, we may have a bit of a problem ..

[fmauch]

Thank you for the feedback @gavanderhoorn!

We have considered for quite a while now whether we want to migrate our config package or not. Personally, I am also not the greatest fan of the way the configs builder hides the required config files away. Though once looking at its source code it became less hidden for me, so a better documentation could potentially improve things here. In my opinion there are a couple of arguments that justify making this step:

• configs-builder seems to be the standard used by MoveIt! nowadays. Of course, that doesn't mean that we'll have to follow that and we do a couple of things differently than a standard moveit_config package would, but it might help new users to get familiar with the structure more easily. And I expect that package to be used by less experienced MoveIt! users anyway, as more experienced users would probably create their own moveit_config package rightaway.
• Reusing this to write your own launchfiles launching your custom nodes should be much more convenient, especially if you are used to using the configs builder. The main point about this PR is changing the structure, not necessarily using the configs builder in our own launchfiles. That's also (from my understanding) what the OP in Adding MoveItConfigsBuilder compatible example to ur_moveit_config #917 was expecting.
• Tyler's package I think is also from a time where the configs builder didn't exist, yet. The launchfiles become much less cluttered once the description is re-used from the description topic.
• Using one big config file for everything isn't necessarily reducing complexity in my opinion.

[fmauch]

I decided against an upstream issue regarding the description topic as it seems intentional to have a default of 10 seconds timeout.

[VinDp]

Works nicely and looks good to me, I only have some small suggestions.

[VinDp]

Suggested change

	<buildtool_depend>ament_cmake_python</buildtool_depend>
	<exec_depend>launch</exec_depend>
	<exec_depend>launch_ros</exec_depend>

[VinDp]

Suggested change

install(DIRECTORY config DESTINATION share/${PROJECT_NAME})