Estimator-real position inversion

[iandareid]

The Estimated position North and East is opposite to what it should be based on the output of path_planner and the full_estimator (MAVSim estimator with a ROS wrapper) says it should be. In order to match truth in Simulation the estimator North and East had to be inverted. This issue became clear while testing path following, in a recent flight test.

[iandareid]

• Gazebo's /fixedwing/truth/NED does not agree with /state (a repackaging of Gazebo truth to a ROSplane state message. /fixedwing/truth/NWU also does not agree. However I know that /state as it stands is correct (uninverting N and E). This is because it agrees with /full_estimator/estimated_state and /path_planner's output.
  • The path planner could be faulty and the full state estimator from MAVSim.
    • This seems very unlikely. These bugs would manifest in MAVSim which they do not.
  • Alternatively Gazebo is publishing NED incorrectly. This issue would have arisen in ROScopter as well.
    • This also feels unlikely. Though investigating how ROScopter deals with this could be enlightening.
  • The last alternative is that I am publishing the state wrong. However, looking through the code it is just a item by item copy of the /fixedwing/truth/NED.
    • For now I am going to fix the issue in Gazebo state transcription node by inverting N and E, but the issue needs to be fixed more permanently.
    • Testing waypoint functionality in sim is necessary.

[iandareid]

Turns out that the fix was to invert the rotation that aligned the spherical coordinate frame for the LLA. It was rotated by 90 degrees to align the axes, but it should have been rotated in the opposite direction by 90.