search

PX4 / PX4-Autopilot

PX4 Autopilot Software
https://px4.io
BSD 3-Clause "New" or "Revised" License
8.12k stars 13.34k forks source link

Drone rotates towards setpoints in Precision Land flight mode in 1.9.x and 1.10.0 #13498

Open dvornikov-aa opened 4 years ago

dvornikov-aa commented 4 years ago

Drone rotates towards setpoints and subsequently fails to land on the target if precision landing starts outside of the mission (by command commander mode auto:precland) . The drone yaw was always 0 in 1.8.x. The worst part is that I can't reproduce this behavior in SITL or HITL.

Steps to reproduce the behavior:

  1. Takeoff above IR-Lock beacon.
  2. Call commander mode auto:precland.
  3. Watch the precision landing process.

Expected behavior The drone yaw must always be constant to perform a successful precise landing.

Log Files and Screenshots Log

Yaw Angle

Drone: F450-based Quadrotor x

photo_2019-11-18_15-39-53
stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. Thank you for your contributions.

julianoes commented 4 years ago

FYI @bresch

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. Thank you for your contributions.

julianoes commented 4 years ago

@dvornikov-aa is this still an issue or have you found a solution or workaround?

kpetrykin commented 4 years ago

@julianoes yes, the problem is still there. The solution was not found yet.

julianoes commented 4 years ago

And is it also there in master/1.11?

@dusan19 do you have an idea about this?

dusan19 commented 4 years ago

I see that, precland doesnt set the triplet yaw.

The triplets were updated only 3 times in this log, each time the yaw was invalid which should result in the flight task calling _set_heading_from_mode(). So the yaw sp should be towards setpoint until the drone is close horizontally to the position setpoint when the yaw sp should become locked. Im not sure why this doesnt happen.

I see that the triplet.current.valid is true but triplet.current.position_valid is false, but the triplet position is finite so the flight task still uses these finite values (checks for is_finate rather than the valid flag). Also _set_heading_from_mode() computes the distance to _target, but the _target is set after the _set_heading_from_mode() is called, so _target from the previous iteration is used, this might be relevant in this case when the triplet is not updated all the time.

Anyway it would require more detailed debugging, also I never used precland outside of the mission

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. Thank you for your contributions.

amefabris commented 5 months ago

I am facing the same issue on v1.14.0 with a custom drone in simulation (gazebo-classic), but it's really hard to track down the issue. The yaw of the setpoint triplets is always 0, however the drone is changing the yaw and moving towards the target.