MAV_CMD_DO_MOUNT_CONTROL not changing yaw

[Pedals2Paddles]

Issue details

The solo is sends mavlink mount control commands to ArduPilot. This is how it controls copter yaw and gimbal pitch in many of the smart shots. The companion computer interpolates gimbal pitch angles and copter yaw angles, then uses mavlink to command said movements. This all worked fine on the old ArduPilot Solo firmware from the 3.2/3.3 days. However on master the copter yaw response is no longer working. The gimbal pitch still responds, but not the copter yaw. So we know the message/command is going through. Just copter yaw is not responding.

• The companion computer uses msg_mount_control to send gimbal pitch and copter yaw. This can be found in copter GCS_Mavlink here: https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/GCS_Mavlink.cpp#L1904.
  It then goes to the function in the mount backend found here: https://github.com/ArduPilot/ardupilot/blob/master/libraries/AP_Mount/AP_Mount_Backend.cpp#L37
  Where it then goes into the control function in the mount backend here:

  ardupilot/libraries/AP_Mount/AP_Mount_Backend.cpp

  Line 45 in 1297a7d

  void AP_Mount_Backend::control(int32_t pitch_or_lat, int32_t roll_or_lon, int32_t yaw_or_alt, MAV_MOUNT_MODE mount_mode)

  
  Which finally concludes here in the mount backend:
  https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/commands_logic.cpp#L1172

• I tried changing the solo's companion computer to use the newer mav_cmd_do_mount_control and the result is the same. Gimbal pitch responded, yaw did not respond. This one is in copter GCS_Mavlink here: https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/GCS_Mavlink.cpp#L1227
  Where it then goes directly into the same control function in the mount backend as the old one:

  ardupilot/libraries/AP_Mount/AP_Mount_Backend.cpp

  Line 45 in 1297a7d

  void AP_Mount_Backend::control(int32_t pitch_or_lat, int32_t roll_or_lon, int32_t yaw_or_alt, MAV_MOUNT_MODE mount_mode)

  
  Which finally concludes here in the backend, again same as the other method:
  https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/commands_logic.cpp#L1172

• I then setup and auto mission and tried using the Mission Command do_mount_control. That does actually work properly in all respects. It begins here in Copter commands_logic: https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/commands_logic.cpp#L111
  Then goes down to the bottom here:
  https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/commands_logic.cpp#L1172
  And concludes here, which is where all the other mavlink commands also conclude:
  https://github.com/ArduPilot/ardupilot/blob/master/ArduCopter/commands_logic.cpp#L1172

So this tells me at least from that last point onward at set_angle_targets, it works fine. So something is not working properly in the Mavlink commands, or the code processing those mavlink commands which lead up to it. The result is again copter yaw not responding. I don't see anything glaringly obvious, but I only know enough C++ and mavlink to break it.

Version

3.5 / 3.6dev

Platform

[ ] All
[ ] AntennaTracker
[X] Copter
[ ] Plane
[ ] Rover

Airframe type

Solo, but likely any vehicle using mavlink mount control

Hardware type

Cube 2.1

Logs

Dataflash log attached.
Solo Mount Control.zip

[Pedals2Paddles]

@tridge @rmackay9 this is what I talking about on the dev call. I did some further testing as you suggested to get to this point.

[rmackay9]

Ok, this sounds to me like Copter thinks the Solo gimbal supports yaw so it thinks the main vehicle doesn't need to yaw.

EDIT: but the code that turns on/off this behavior seems ok. https://github.com/ArduPilot/ardupilot/blob/master/libraries/AP_Mount/AP_Mount_SoloGimbal.cpp#L88

[Pedals2Paddles]

That would be way too easy for my luck!

[Pedals2Paddles]

I put a condition_yaw command immediately following mav_cmd_do_mount_control, which is a working workaround now. The condition_yaw command is working properly. This doesn't appear to degrade the vehicle's performance. But it is a somewhat ugly workaround. It does however confirm the problem is not how the solo is calculating its yaw.

This leads me to believe there is a problem in how ArduPilot is handling the mount_control message and command.