Enable support for global position setpoints with SET_POSITION_TARGET_GLOBAL_INT

[Jaeyoung-Lim]

Describe problem solved by the proposed pull request
Currently offboard position setpoints can only be passed in local coordinates with SET_POSITION_TARGET_LOCAL_NED. However, when running multiple vehicles at the same time, this is not scalable as setpoints need to be transformed to each vehicle's local frame.
Also, there has been a lot of workarounds for external applications to use global coordinates and converting it to local coordinates to just enable offboard mode.

Fixes #11625

Describe your preferred solution
Added a message handler in mavlinkreceiver for SET_POSITION_TARGET_GLOBAL_INT which converts the setpoint to local coordinates.

Test data / coverage
This log shows the vehicle position being set to altitude of 12m with the same lon / lat with the home position.
I have also tested with multiple vehicles flying in a circular pattern: https://youtu.be/tj3FcbmsVpw

Additional context
Tools such as mavros convert global position setpoints into local coordinates as a workaround. There has been confusion from mavros users of px4 not supporting this message(mavlink/mavros#1256, mavlink/mavros#903, mavlink/mavros#728 )

[hamishwillee]

Just a reminder to update http://docs.px4.io/master/en/flight_modes/offboard.html when this goes in.

I assume this only really affects offboard mode?

[Jaeyoung-Lim]

@hamishwillee Ofcourse! I will prepare a PR for the docs.

Yes it only affects offboard mode

[hamishwillee]

Thanks @Jaeyoung-Lim . Not everyone "remembers" docs. I'm pleased you do.

[LorenzMeier]

Awesome! These small fixes are so critical to keep adopters happy.

[LorenzMeier]

Please have a quick look at the CI failures

[julianoes]

Thanks for this! I made some comments how I think this could be improved.

Also, it would be nice to add an integration test to MAVSDK to test it.

[Jaeyoung-Lim]

@julianoes I would be interested in adding the unit tests to mavsdk. Could you provide any pointers?

[julianoes]

@Jaeyoung-Lim that would be great.

Actually, MAVSDK doesn't support global position offboard yet, so it would have to be added there as well, like the implementation for local position:
https://github.com/mavlink/MAVSDK/blob/e84034a468d67be6666296fadcb95312d77cc20c/src/plugins/offboard/offboard_impl.cpp#L112-L137

And then a test would look like this:
https://github.com/mavlink/MAVSDK/blob/e84034a468d67be6666296fadcb95312d77cc20c/src/integration_tests/offboard_position.cpp#L43-L51

[Jaeyoung-Lim]

@julianoes Thanks for the pointers!

[Jaeyoung-Lim]

@julianoes Could we merge this since the documentation has been updated already? I would prefer making the improvements regarding the indentation and logic together with the message handling of SET_POSITION_LOCAL_NED message, since the code is nearly identical.

@hamishwillee

[julianoes]

@Jaeyoung-Lim absolutely!

[TSC21]

@Jaeyoung-Lim are you open to add support for this on the MAVROS setpoint_position plugin? Currently we have the /setpoint_position/global topic that does the conversion from LLA to ENU and sends it as local position setpoints. We can keep that but for the current callback, the topic can be renamed to /setpoint_position/global_to_local and add a new callback that subscribes to /setpoint_position/global of type geographic_msgs/GeoPoint (or GeoPose if you set the orientation as well).

[Jaeyoung-Lim]

@TSC21 Thanks for the suggestions! I would be interested to add support on MAVROS. I was super confused with that plugin when I was trying to use global setpoints on mavros for that reason(hence, this PR).

Could you explain on why we would want to use geographic_msgs/GeoPoint compared to the mavros_msgs/GlobalPositionTarget? IMO it would make sense to replace the call back to simplify the plugin.

[Jaeyoung-Lim]

@TSC21 Ah I misunderstood what you were saying. So I should replace the current callback to the set_position_target_global_int and move the current callback to a new topic. Makes sense.

[TSC21]

@Jaeyoung-Lim what I am saying is that inside the setpoint_position we are handling the data being published in /setpoint_position/global in a callback that transforms LLA to ENU local setpoints. We rather want to handle it in a callback that directly uses the LLA and publishes it as set_position_target_global_int.

[TSC21]

Also mavros_msgs/GlobalPositionTarget is used by the setpoint_raw plugin. On the setpoint_position plugin we want to use std ROS msgs, not mavros_msgs.