Improve front transition tracking & fix flyaway #14727
Conversation
|
In order to reproduce the first bug addressed by this PR use plain master and then:
You will see the vehicle perform a flyaway. What we should also fix: |
|
The figure below shows a takeoff followed by a loiter around the home position. After issuing a backtransition command the vehicle transitions while still flying the circle. Once transitioned it moves back to the home point (which is the loiter target).
|
|
Fixes #14718 |
|
Mission: |
|
Solved my issue 👍 |
| Vector2f prev_wp{0.0f, 0.0f}; | ||
| if (!PX4_ISFINITE(_transition_waypoint(0))) { | ||
| double lat_transition, lon_transition; | ||
| waypoint_from_heading_and_distance(_current_latitude, _current_longitude, _yaw, 1000.0f, &lat_transition, |
There was a problem hiding this comment.
A small comment where this 1000 is coming from would be useful here. Maybe best define a separate variable for it, e.g. const float transition_destination_dist = 1000.0f.
There was a problem hiding this comment.
@sfuhrer yes, I can do that
|
@RomanBapst this makes sense to me. Beside the minor proposal for better commenting it looks good to me, and also does in sitl wthat it's supposed to do. Do we want real flight testing before merging? |
@sfuhrer Would be good to get some testing for the release. |
Signed-off-by: RomanBapst <[email protected]>
…ition - this fixes the case where the navigator publishes a loiter waypoint or any waypoint which is too close to the vehicle. Signed-off-by: RomanBapst <[email protected]>
RomanBapst
force-pushed
the
pr_flyaway_front_trans
branch
from
2a9f02a to
f1cb5f3
Compare
This PR contains the following improvement:
At the start of a front transition the fixed wing position controller creates a virtual position setpoint which is located 1000m in front of the vehicle. This avoids the controller tracking waypoints which are too close to the vehicle, e.g. loiter waypoints.
It would of course be nicer if the navigator would automatically publish this setpoint. However, this would take quite some changes and I'm not comfortable with the attempt just before a release.
Also, the logic in the position controller is straight forward and can be removed very easily.
This PR fixes the following bug:
A takeoff followed by a transition command lead to the vehicle doing a flyaway. This was because the takeoff code did not set valid latitude and longitude for the takeoff waypoint but still set the _can_loiter_at_sp flag to true. As a result the loiter code did not set a position setpoint but just copied over the NAN values.
Finally, the position controller did not know what to do with that and just did nothing.
I think this bug was introduced with flighttasks as they support of NAN position setpoint.