Plane: Fix level-off and throttle without airspeed sensor #28776
Conversation
menschel
force-pushed
the
pr-fix-auto-mode-takeoff
branch
4 times, most recently
from
7fbc5fc
to
8b3f392
Compare
|
I'm satisfied with the takeoff behavior now. There is no visible difference from 4.5.7. The parameter needs to be discussed though. I argue the parameter should be "inverted", so any user migrating from 4.5.7 to 4.6 experiences the known behavior. |
This patch intends to fix two issues with auto takeoff mission item for planes without an airspeed sensor. Level-off fix: - Correct the level-off condition by factor 2 because during level-off, the pitch and therefore the climb-rate decreses linearly. Mathmatically the plane therefore only climbs half of the altitude in the expected time and is not guaranteed to converge to target altitude by itself. - Add flag for level-off stage and let TECS handle the final climb. Throttle without airspeed sensor: - Remove conditional evaluation of airspeed sensor to enable traditional TECS-controlled throttle range also without airspeed sensor. This restores the behavior of ArduPlane 4.5.7. It is to be discussed if the flag THROTTLE_RANGE in parameter TKOFF_OPTIONS is to be reversed to keep the behavior from Plane 4.5.7 consistent with Plane 4.6 onwards. It could be something like FORCE_MAX_THROTTLE instead.
menschel
force-pushed
the
pr-fix-auto-mode-takeoff
branch
from
8b3f392
to
4ad2972
Compare
| @@ -309,13 +308,17 @@ int16_t Plane::get_takeoff_pitch_min_cd(void) | |||
|
|
|||
| // are we entering the region where we want to start levelling off before we reach takeoff alt? | |||
| if (auto_state.sink_rate < -0.1f) { | |||
| float sec_to_target = (remaining_height_to_target_cm * 0.01f) / (-auto_state.sink_rate); | |||
| float sec_to_target = (remaining_height_to_target_cm * 0.01f * 2.0f) / MIN(TECS_controller.get_max_climbrate(), -auto_state.sink_rate); | |||
There was a problem hiding this comment.
The decrease of pitch is not linear; that much we've seen from the SITL tests.
But in any case, can you provide some justification for this formula?
There was a problem hiding this comment.
I make the case that the scalar creates the linear dependency between remaining height and pitch.
Integrating A*X results in (A/2)*X² and that's where the factor 2 comes from.
Setting both equal results in X=1/2 for the gained altitude.
The MIN() is a sanity check to limit climb-rate to the max value known to TECS.
It is not dependent on angle but on the power to weight ratio.
I don't know what happens in 4.5.7 but there it is not linear.
There was a problem hiding this comment.
Thank you for your effort @menschel !
As you have noticed, we chose to address the level-off bug in a much simpler way, implemented here: #28792
Regarding the throttle behaviour, we intentionally changed it from 4.5 to 4.6 here: #27758
In 4.5, the throttle during takeoff would be managed by TECS in TAKEOFF mode but set fixed to TKOFF_THR_MAX in AUTO mode takeoffs.
We opted to keep the 2nd behaviour.
If you really want the old throttle behaviour for TAKEOFF mode back, we would probably need to make use of another TKOFF_OPTIONS bit. That would need to be done carefully, testing all corner cases.
It's not high on my priority list. If you want to implement that, you're very welcome of course.
If you believe the 4.6 behaviour is dangerous for some reason, naturally we'd want to hear your argument.
|
Ahem, you have mixed Takeoff-mode and Auto-mode takeoff mission item. |
This PR does intend to fix Issues #28685 and #28775
This patch intends to fix two issues with auto takeoff mission item for
planes without an airspeed sensor.
Level-off fix:
the pitch and therefore the climb-rate decreses linearly.
Mathmatically the plane therefore only climbs half of the altitude in the expected time
and is not guaranteed to converge to target altitude by itself.
Throttle without airspeed sensor:
TECS-controlled throttle range also without airspeed sensor.
This restores the behavior of ArduPlane 4.5.7.
It is to be discussed if the flag THROTTLE_RANGE in parameter TKOFF_OPTIONS is to be reversed to keep
the behavior from Plane 4.5.7 consistent with Plane 4.6 onwards.
It could be something like FORCE_MAX_THROTTLE instead.
Latest measurements with this change cherry-picked onto ArduPilot-4.6:
24-12-01_09-12-18.bin
24-12-01_09-16-52.bin
24-12-01_09-21-56.bin