Improve quadplane use of forward throttle in VTOL modes #13760

priseborough · 2020-03-08T21:25:40Z

This patch provides a more direct control of forward thrust motor throttle during quadplane Q modes. I have found it to provide more precise position control in varying wind conditions in Q_LOITER and better control with less pitch/dive when operating in QSTABILIZE and QHOVER and applying forward tilt to reposition or hold position in wind.

I would be interested to get feedback from other quadplane flyers.

Hwurzburg · 2023-06-09T09:25:42Z

Paul some questions on this...first ..is this adding forward motor when only pitch would be used is QMODEs to fly forward rapidly? like Mark manual forward throttle?

what modes is this active? what about VTOL waypoint moves is AUTO?

does 0 gain disable it? (if so should be noted in metadata)

priseborough · 2023-06-12T11:09:57Z

Paul some questions on this...first ..is this adding forward motor when only pitch would be used is QMODEs to fly forward rapidly? like Mark manual forward throttle?

It performs te following:

Limits forward tilt to a value that allows the wing to go to a zero lift angle of attack but no more
Replaces the forward acceleration that would have been produced by the tilt with thrust fro the forward motor

what modes is this active? what about VTOL waypoint moves is AUTO?

This method works in all VTOL modes other than auto-tune

does 0 gain disable it? (if so should be noted in metadata)

Yes it does and metadata has been updated to make this clear. Also I have updated metadata to make it clear that this method replaces the velocity controller method controlled by Q_VFWD_GAIN that is less accurate and only works on flight modes that use the velocity controller.

pompecukor · 2023-07-03T23:52:18Z

Going to test this soon. Will report back

pompecukor · 2023-07-04T15:22:37Z

This method works in all VTOL modes other than auto-tune

Correction, it also does NOT work in Qstabilize mode.... Which is a good thing.

priseborough · 2023-07-04T23:46:25Z

I've consolidated the patches, fixed the bug preventing use with tilt rotors and incorporated feedback on parameter names and descriptions.

pompecukor · 2023-07-05T16:06:38Z

@priseborough
Tested, seems to work very nice. Logs sent.

IamPete1 · 2023-07-07T23:50:55Z

This is essentially adding an acceleration P gain. I completely agree that this is the way we need to go.

However I don't think this is the correct place. I would like to see a full PID acceleration controller implemented in the position controller and integrated with the 6DoF work @lthall and @bnsgeyer are working on. It will take a bit longer to get there, but the final result should be better.

priseborough · 2023-07-08T06:30:06Z

This is essentially adding an acceleration P gain. I completely agree that this is the way we need to go.

However I don't think this is the correct place. I would like to see a full PID acceleration controller implemented in the position controller and integrated with the 6DoF work @lthall and @bnsgeyer are working on. It will take a bit longer to get there, but the final result should be better.

It is not adding a P gain, ie there is no new feedback loop. It is replacing the use of forward pitch to achieve acceleration with forward throttle to replace the acceleration lost due to pitch limiting,. As such I see this as being downstream of any acceleration feedback controller and belonging in a control mixer.

IamPete1 · 2023-07-08T15:01:16Z

It is not adding a P gain, ie there is no new feedback loop. It is replacing the use of forward pitch to achieve acceleration with forward throttle to replace the acceleration lost due to pitch limiting,. As such I see this as being downstream of any acceleration feedback controller and belonging in a control mixer.

Sorry, your of course right. I'm confusing what I had in mind and what you actually did.

However I still think this is in the wrong place, the position controller needs visibility of this extra degree of freedom. For example this lets the vehicle go quite fast while staying level, however that speed is still constrained by Q_ANGLE_MAX, dispute the pitch angle staying at 0.

To elaborate on the approach I was thinking about. We add a 2D acceleration controller, unlike the current position controller we align to vehicle yaw so we can have different forward/back and left/right gains. The closed loop nature of this controller means that we can just hard code the mixing that you have parameterized with Q_FWD_THR_GAIN. It should also deal with none-linearity in throttle -> thrust response such as those we fix with the expo parameter for vertical motors. It should also deal with moving between angle and control and the new degree of freedom. Having the acceleration controller also means we can define acceleration limits, which we cannot currently do. I hope that the new acceleration controller would be like our other position controllers that are quite robust and easy to tune to a reasonable level (half it until it stops wobbling).

priseborough · 2023-07-22T09:46:17Z

It is not adding a P gain, ie there is no new feedback loop. It is replacing the use of forward pitch to achieve acceleration with forward throttle to replace the acceleration lost due to pitch limiting,. As such I see this as being downstream of any acceleration feedback controller and belonging in a control mixer.

Sorry, your of course right. I'm confusing what I had in mind and what you actually did.

However I still think this is in the wrong place, the position controller needs visibility of this extra degree of freedom.

I would propose a different way of architecting the required functionality. How about making the position controller agnostic to the method that is used to generate a specific force in the Front Right Down directions and let the vehicle libraries to the required mapping as this is gong to be unique to different vehicle types. I don't think long term the position controller should be responsible for the mapping from specific force demand to actuation.

priseborough · 2023-07-23T10:14:35Z

These changes should not impact on future position controller restructuring. The attached sketch shows how the position controller (PSC) pitch setpoint is limited in the quadplane code and converted to the forward throttle demand in quadplane::calc_fwd_tilt_throttle() This is all done open loop.

Extending this to the tilt rotor case was trivial f896636 as the functionality required to convert a forward thrust demand to a rotor tilt already exisited.

What I would like to see us adopt in the future is a more generic position controller interface that outputs specific force demands in the forward and right direction. The vehicle code would then convert that to whatever combination of vehicle attitude, actuation angles and motor throttle required to produce the required force. This would require the vehicle code to provide the position controller with dynamically updated saturation limits for specific force in the front, back, right, left directions.

timtuxworth · 2023-07-23T16:14:11Z

I'm expecting to see a position controller class. Am I missing something?

priseborough · 2023-07-25T11:01:32Z

Here is an example acceleration into wind in QLOITER mode from @pompecukor 's testing with

Q_FWD_PIT_LIM 500
Q_FWD_THR_GAIN 2.0
Q_VFWD_GAIN 0.000000

The aircraft can now achieve over 30 m/s of airspeed in QLOITER without porpoising or losing height. The following figure shows a stick forward acceleration to maximum ground speed followed by a stick release decel.

Note the height gain during decel which is due to the pitch back and increase in wing lift. This PR does not modify that behaviour. Height control accuracy was good during the acceleration phase.

The following figure shows right and left rotor tilt servo PWM demand where forward tilt are larger values.

The position controller was able to control horizontal position accurately throughout.

IamPete1 · 2023-07-25T22:09:01Z

ArduPlane/mode_qloiter.cpp

@@ -92,7 +92,7 @@ void ModeQLoiter::run()

    // call attitude controller with conservative smoothing gain of 4.0f
    attitude_control->input_euler_angle_roll_pitch_euler_rate_yaw(plane.nav_roll_cd,
-                                                                  plane.nav_pitch_cd,
+                                                                  MAX(plane.nav_pitch_cd, quadplane.q_fwd_nav_pitch_lim_cd),


We should apply this up on 84, the set_VTOL_roll_pitch_limit call will be limiting to the wrong value other wise. So 84 would become:

plane.nav_pitch_cd = MAX(loiter_nav->get_pitch(), quadplane.q_fwd_nav_pitch_lim_cd);

IamPete1

The throttle handling stuff all looks good. But I think we need to be very careful where we apply the angle limit, since were now directly affecting what a mode it doing by taking away pitch forward rather than just adding extra throttle I think we should output the throttle in the same place we apply the angle limit such that we can never get one without the other.

I think we also need to make sure we update nav_pitch_cd to the limited value because it is used throughout the code in various odd places that might get upset if it does not represent what the vehicle is doing.

I'm happy to help with some re-work if that would be usefull.

ArduPlane/quadplane.cpp

IamPete1 · 2023-07-25T22:13:50Z

ArduPlane/quadplane.cpp

+    q_fwd_throttle = 0.0f;
+    q_fwd_nav_pitch_lim_cd = -aparm.angle_max;


maybe a "is_active" method rather than keeping this copy.

I don't understand.

IamPete1 · 2023-07-25T22:15:25Z

ArduPlane/quadplane.cpp

@@ -934,13 +953,13 @@ void QuadPlane::multicopter_attitude_rate_update(float yaw_rate_cds)

        if (use_yaw_target) {
            attitude_control->input_euler_angle_roll_pitch_yaw(plane.nav_roll_cd,
-                                                               plane.nav_pitch_cd,
+                                                               MAX(plane.nav_pitch_cd, q_fwd_nav_pitch_lim_cd),


I think we have to set nav_pitch_cd to this limit before getting into this call. For example tailsitter transitions need to break these rules. I know that tailsitters should never have this set, but they might.

ArduPlane/quadplane.cpp

IamPete1 · 2023-07-25T22:40:35Z

I just thought of something else that makes this even more tricky to implement. We need to tell the position controller when we have run out of throttle or tilt so it can stop winding up I term. We do have a set_externally_limited_xy() but I'm not sure how well that would work for long periods, currently its only used temporally in the transition and for touch down.

priseborough · 2023-07-31T08:15:59Z

I just thought of something else that makes this even more tricky to implement. We need to tell the position controller when we have run out of throttle or tilt so it can stop winding up I term. We do have a set_externally_limited_xy() but I'm not sure how well that would work for long periods, currently its only used temporally in the transition and for touch down.

I can achieve the required integrator protection by calling set_lean_angle_max_cd to provide a lean angle that corresponds to when forward throttle will be saturated. I would like to split set_lean_angle_max_cd into a separate call for forwards tilt. I'll give you a call to discuss.

priseborough · 2023-08-13T07:27:08Z

@IamPete1 I've reworked it to move the angle limiting and forward thrust calculation to just after we get the demanded pitch angle from the position controller so that a consistent pitch angle demand is guaranteed for all downstream processing.

…R_GAIN

and use a common function for the active method

Don't unnecessarily increase the forward pitch limit when position control forward action is saturated.

tridge · 2023-09-26T06:38:09Z

ArduPlane/mode_qstabilize.cpp

@@ -35,6 +35,11 @@ void ModeQStabilize::update()
        plane.nav_roll_cd = roll_input * plane.quadplane.aparm.angle_max;
        plane.nav_pitch_cd = pitch_input * plane.quadplane.aparm.angle_max;
    }
+
+    if (plane.control_mode->mode_number() == Mode::Number::QSTABILIZE) {


this check is preventing assign_tilt_to_fwd_thr() being called in QHOVER mode. I just tested QHOVER and the new functionality doesn't work with the RC switch
we can either remove the check and remove the call in qloiter, qrtl etc, or add the call into all the relevant modes
I think it is better to have it in one place?

tridge · 2023-09-26T07:04:24Z

I don't think we can fix the above issue by just always running in qstabilize for all modes as it would defeat the checks per-mode for things like throttle_wait, if a user put the nose down while in throttle wait it would run up the fwd motor
but that means we need to be sure we put plane.quadplane.assign_tilt_to_fwd_thr() in all modes that are missing it now, or we could end up without vfwd support where we have it now with the old system.

this prevents double calling and fixes qhover

tridge · 2023-09-26T08:57:01Z

@priseborough as discussed, I've fixed by moving the assign tilt to the run() function

priseborough added Enhancement VTOL-Plane labels Mar 8, 2020

priseborough requested a review from tridge March 8, 2020 21:25

priseborough self-assigned this Mar 8, 2020

tridge mentioned this pull request Jun 6, 2023

Plane: stable 4.5.x issues list #15941

Open

priseborough force-pushed the pr-QuadPlaneFwdThrottle branch from 8095af6 to b059077 Compare June 9, 2023 07:45

priseborough closed this Jun 9, 2023

priseborough reopened this Jun 9, 2023

priseborough force-pushed the pr-QuadPlaneFwdThrottle branch from 751a072 to 45f7699 Compare July 2, 2023 01:10

priseborough force-pushed the pr-QuadPlaneFwdThrottle branch 2 times, most recently from 77667b9 to d247fab Compare July 4, 2023 23:45

IamPete1 reviewed Jul 25, 2023

View reviewed changes

priseborough force-pushed the pr-QuadPlaneFwdThrottle branch from 59dfc69 to 918b46f Compare August 13, 2023 07:23

priseborough and others added 22 commits September 26, 2023 16:10

Plane: if statement cleanup in QuadPlane::assign_tilt_to_fwd_thr

b913f32

Plane: Fix use of q_fwd_nav_pitch_lim_cd class variable

b8a3a75

Plane: prevent wind up of Q-mode velocity controller I term

3bd151f

Plane: Default Q_FWD_THR_GAIN to 0 and update meta data

3b14edd

Plane: Use Q_FWD_THR_USE parameter to control modes that use Q_FWD_TH…

a39568d

…R_GAIN

Plane: Add arming check for incompatible Q_VFWD_GAIN and Q_FWD_THR_USE

3b66034

Plane: Add Q_FWD_THR_GAIN to table of transmitter tuning parameters

47c23fc

Plane: Remove duplicated parameter check

05d1f96

AP_ICEngine: added ICE_DISABLED state

eafc5d3

RC_Channel: added VFWD_THR_OVERRIDE

406295a

Plane: implement methods for aux switch control of vfwd code

d936794

and use a common function for the active method

Plane: fixes from review

b986a24

AC_PosControl: Add monitoring and reporting of forward accel saturation

c27e9cc

Plane: Add protection for failed quadplane forward thrust motor

bea29c2

Plane: fix centi-degree scaling bug

6963257

Plane: explicitly type cast pitch limit

a919943

Plane: Prevent quadplane.assign_tilt_to_fwd_thr() being called twice

ef77cc9

Plane: Restructure throttle scaling and pitch limiting to enable logging

5ccd190

Plane: Add diagnostics logging to QuadPlane::assign_tilt_to_fwd_thr

fa5fceb

Plane: Rework pitch limiting in QuadPlane::assign_tilt_to_fwd_thr

d53d512

Don't unnecessarily increase the forward pitch limit when position control forward action is saturated.

Tools: Add test for quadplane forward motor use and pitch limiting

ddd3e93

Plane: fix metadata units for Q_FWD_PIT_LIM

de100ce

tridge requested changes Sep 26, 2023

View reviewed changes

Plane: moved assign tilt to the run() function

1d4c738

this prevents double calling and fixes qhover

tridge force-pushed the pr-QuadPlaneFwdThrottle branch from 68c9928 to 1d4c738 Compare September 26, 2023 08:56

tridge approved these changes Sep 26, 2023

View reviewed changes

tridge added the MergeOnCIPass label Sep 26, 2023

tridge merged commit babdb36 into ArduPilot:master Sep 27, 2023

		q_fwd_throttle = 0.0f;
		q_fwd_nav_pitch_lim_cd = -aparm.angle_max;

Improve quadplane use of forward throttle in VTOL modes #13760

Improve quadplane use of forward throttle in VTOL modes #13760

Uh oh!

Conversation

priseborough commented Mar 8, 2020

Uh oh!

Hwurzburg commented Jun 9, 2023

Uh oh!

priseborough commented Jun 12, 2023

Uh oh!

pompecukor commented Jul 3, 2023

Uh oh!

pompecukor commented Jul 4, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

priseborough commented Jul 4, 2023

Uh oh!

pompecukor commented Jul 5, 2023

Uh oh!

IamPete1 commented Jul 7, 2023

Uh oh!

priseborough commented Jul 8, 2023

Uh oh!

IamPete1 commented Jul 8, 2023

Uh oh!

priseborough commented Jul 22, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

priseborough commented Jul 23, 2023

Uh oh!

timtuxworth commented Jul 23, 2023

Uh oh!

priseborough commented Jul 25, 2023

Uh oh!

IamPete1 Jul 25, 2023

Choose a reason for hiding this comment

Uh oh!

IamPete1 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

IamPete1 Jul 25, 2023

Choose a reason for hiding this comment

Uh oh!

priseborough Jul 31, 2023

Choose a reason for hiding this comment

Uh oh!

IamPete1 Jul 25, 2023

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

IamPete1 commented Jul 25, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

priseborough commented Jul 31, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

priseborough commented Aug 13, 2023

Uh oh!

tridge Sep 26, 2023

Choose a reason for hiding this comment

Uh oh!

tridge commented Sep 26, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tridge commented Sep 26, 2023

Uh oh!

Uh oh!

pompecukor commented Jul 4, 2023 •

edited

Loading

priseborough commented Jul 22, 2023 •

edited

Loading

IamPete1 commented Jul 25, 2023 •

edited

Loading

priseborough commented Jul 31, 2023 •

edited

Loading

tridge commented Sep 26, 2023 •

edited

Loading