Landing horizontal velocity compensation / unsteady landing

[Claudio-Chies]

Solved Problem

There was a problem where triggering Land, while having a forward velocity, the landing algorithm didn't compensate it, meaning the vehicle remained drifting while getting closer to the ground , and in some cases jerking unsteadily

Solution

i set the target landing location based on the current velocity, maximum acceleration and jerk. this way the drone is able to compensate the velocity and land smoothly at the target location.

Changelog Entry

• changed the lat and long in the triplet generation process in the navigator to be at a location in front of the vehicle, dependent on its current state.
• set the previous setpoint triplet to be the same as the target one, this way we avoid getting into offtrack mode.

Bugfix: Unsteady landing with horizontal velocity
Documentation: the position set point for landing gets set to a location in front of the vehicle where its able to halt before. greatly improving landing performance when triggering landing while moving

Test coverage

SITL and on Hardware

[sfuhrer]

@Claudio-Chies looks clean but I need to look at it in more detail. Do you know which change caused this regression (assuming it is a regression)?

[Claudio-Chies]

@sfuhrer Not yet, i can have a look into it.
I realized just now, that this might enable us zo get rid of the special case for vtol where the landing position is hardcoded 1m infront of the current position.

[bresch]

_target is the _triplet_target pre-processed by the flight task. The general rule is that we shouldn't use _triplet_target in the code and only use _target as this is used everywhere else otherwise there will be an inconsistency between the different parts of FlightTaskAuto.

If I understand correctly using _triplet_target fixes the issue because it skips _updateInternalWaypoints. Could you rather find why _updateInternalWaypoints changes it and fix the problem there?

[Claudio-Chies]

_target is the _triplet_target pre-processed by the flight task. The general rule is that we shouldn't use _triplet_target in the code and only use _target as this is used everywhere else otherwise there will be an inconsistency between the different parts of FlightTaskAuto.

If I understand correctly using _triplet_target fixes the issue because it skips _updateInternalWaypoints. Could you rather find why _updateInternalWaypoints changes it and fix the problem there?

good point, didn't know that. so the issue seems that it goes into offtrack mode.
with the change proposed it avoids offtrack mode by setting the previous triplet to be the same as the target.