[WIP] VTOL coordinate FW and MC rate controllers only #9190

dagar · 2018-03-29T18:11:42Z

In current VTOL during transitions the VTOL "attitude controller" determines the attitude setpoint and both full FW & MC attitude controllers run respecting that setpoint. There are various issues with this especially for tailsitters (#9189).

In this PR only the rate controllers are allowed to run simultaneously. During transition that means the MC attitude controller is running, but the FW attitude controller is now in a state where it's only running the rate controller and respecting published rate setpoints (coming from the MC controller).

@RomanBapst I believe this was working with VTOL standard, but the rebase was nasty so proceed like it's completely untested.

dagar · 2018-03-29T18:33:56Z

If we go forward with this I'd also like to consider removing the thrust setpoint from the attitude setpoint and rates setpoint messages.

vehicle_thrust_setpoint

float thrust[3]

Then I'd like to use it to initialize the position controllers correctly.

TECS after transition to FW
TECS after FW manual, acro, rattitude, or stabilize mode
MC thrust after acro or rattitude

RomanBapst · 2018-04-03T14:10:15Z

@dagar I have now the fw rate controller running in hover for the tailsitter. In SITL we don't model any slipstream so I got the following funny behavior.
https://www.youtube.com/watch?v=9fkTu6M0Dzc&feature=youtu.be

dagar · 2018-04-03T16:05:40Z

src/modules/fw_att_control/FixedwingAttitudeControl.cpp

-					_roll_ctrl.set_bodyrate_setpoint(_rates_sp.roll);
+					vehicle_rates_setpoint_poll();
+
+					if (_parameters.vtol_type == vtol_type::TAILSITTER) {


Can we handle all of these adjustments right after the orb_copy? Then the actual controller code can be generic.

RomanBapst · 2018-04-03T16:09:25Z

@dagar Successfully tested in SITL for standard and tiltrotor, tailsitter works in manual but still got to find a bug in mission. Once we get the desired behavior we should discuss some specifics we might want to solve cleaner.

RomanBapst · 2018-04-05T14:05:47Z

@dagar This now contains changes required to use the duorotor tailsitter in stabilized mode (including transitions and fw flight).
There are a few things I needed to hack and which probably break the other vtol types- we should have a look at them and discuss how to solve the properly.

RomanBapst · 2018-04-05T14:07:25Z

src/modules/fw_att_control/FixedwingAttitudeControl.cpp

@@ -574,6 +602,10 @@ void FixedwingAttitudeControl::run()
 					perf_count(_nonfinite_input_perf);
 				}

+				if (_vehicle_status.in_transition_mode || !_vehicle_status.is_rotary_wing) {


@dagar This is kind of a gain scheduling. In transition and fw mode the fw controller gains get scaled down as if the vehicle was flying at max airspeed.

Should be tailsitter only though right? I think we should get all the vtol (and tailsitter) special cases centralized in a function outside of the main loop. Then a single check for tailsitter that corrects everything (attitude, control mode, airspeed).

RomanBapst · 2018-04-05T14:08:56Z

src/modules/mc_pos_control/mc_pos_control_main.cpp

 			_att_sp.timestamp = hrt_absolute_time();

 			/* reset velocity limit */
 			_vel_max_xy = _params.vel_max_xy;
 		}

 		/* reset setpoints and integrators VTOL in FW mode */
-		if (_vehicle_status.is_vtol && !_vehicle_status.is_rotary_wing) {
+		if (_vehicle_status.is_vtol && (!_vehicle_status.is_rotary_wing || _vehicle_status.in_transition_mode)) {


@dagar This was required to reset the yaw_sp after a transition, otherwise the vehicle would mostly do a 180 in yaw.
Need to be done properly, other vtol types need the controller to run during transition.

Oh okay, let's get it back in.

dagar · 2018-04-05T15:11:01Z

src/modules/fw_att_control/FixedwingAttitudeControl.cpp

 	vehicle_control_mode_poll();
 	vehicle_manual_poll();
 	vehicle_status_poll();
 	vehicle_land_detected_poll();

+	// set initial maximum body rate setpoints
+	_roll_ctrl.set_max_rate(_parameters.acro_max_x_rate_rad);


The existing set_max_rate() update code below doesn't handle this?

@dagar Nope, it does not. The problem is that for VTOL the condition
_vcontrol_mode.flag_control_attitude_enabled != _flag_control_attitude_enabled_last
does not hold because since the mc attitude controller is in charge _vcontrol_mode.flag_control_attitude_enabled is false in the beginning.
So you either have to switch to fw mode first of you have to change a parameter.
Without this change all of the max rates of the rate controller were set to zero, so it essentially did nothing.

dagar · 2018-04-05T18:56:55Z

TODO: review FW integrator reset. https://github.com/PX4/Firmware/blob/d011aa7d47262042e41b28146e6f8ffd0f971355/src/modules/fw_att_control/FixedwingAttitudeControl.cpp#L648

dagar · 2018-04-12T05:43:25Z

Take a look at the MC offboard attitude control CI failure?
http://ci.px4.io:8080/blue/organizations/jenkins/Firmware/detail/PR-9190/11/pipeline

jlecoeur · 2018-04-12T08:29:31Z

Good to see that you are pushing towards separated XYZ thrust setpoints. I am sure that will make the transition phase easier to handle. For quadplane, it may simplify things like X thrust aid to fight headwind during hover, or Z thrust aid to prevent falling after wing stall.
I have few comments:

Thrust frame
The X Y and Z thrust components should be in front-right-down body frame. That means you will need to multiply thrust_z by -1.
Actuators[3]
Should actuator_controls also contain X Y and Z thrust? As of now it contains THROTTLE. I think this is fine to update it in a later pass, but then we have to agree on the sign. I would say that throttle is always positive, even if thrust_z is negative. (i.e. actuator_controls.controls[3] = + thrust_x and actuator_controls.controls[3] = - thrust_z). The advantage is that the mixers do not have to be changed. Let me know if you want to add XYZ thrust in the actuator_controls because I did it in another branch.
Tailsitter body frame
This more a question than a comment. Why do tailsitters use different body frames during hover and forward flight? Hovering with body X axis pointed up should not be a problem if the mc attitude controller uses quaternions all the way.
Simultaneous rate controllers
Is it planned to also prevent fw rate controller and mc rate controller from running simultaneously? In the tailsitter, both control the same set of actuators, so it seems that the original issue VTOL: Control architecture limitation for tailsitters #9189 is unsolved?

RomanBapst · 2018-04-12T11:50:48Z

@jlecoeur Good to see you also participating in this!

Thrust frame
The X Y and Z thrust components should be in front-right-down body frame. That means you will need to multiply thrust_z by -1.

Yes, I totally agree. This is still on my TODO list for this PR.

Actuators[3]
Should actuator_controls also contain X Y and Z thrust? As of now it contains THROTTLE. I think this is fine to update it in a later pass, but then we have to agree on the sign. I would say that throttle is always positive, even if thrust_z is negative. (i.e. actuator_controls.controls[3] = + thrust_x and actuator_controls.controls[3] = - thrust_z). The advantage is that the mixers do not have to be changed. Let me know if you want to add XYZ thrust in the actuator_controls because I did it in another branch.

I think it should. So what if we want to support reverse thrust as well? In that case throttle would need to cover negative ranges as well, right? I'm also thinking about vehicles with variable pitch props.
Can you give me the link to your branch so that I can have a look?

Tailsitter body frame
This more a question than a comment. Why do tailsitters use different body frames during hover and forward flight? Hovering with body X axis pointed up should not be a problem if the mc attitude controller uses quaternions all the way.

The reason why it was done like this was because it was the easiest solution at that time I guess.
But I would also prefer if we don't have to transform attitudes between attitude controllers.
Given that the fixed wing attitude controller is euler angle based we should definitely choose the 'fixed wing' frame as reference (neutral attitude is when fixed wing is level).
Should we discuss this together? I'm available for a skype call or we can discuss it on the dev call.
We should also sync with @dagar

Simultaneous rate controllers
Is it planned to also prevent fw rate controller and mc rate controller from running simultaneously? In the tailsitter, both control the same set of actuators, so it seems that the original issue #9189 is unsolved?

For the tailsitter, why do you think they control the same sets of actuators? In that case the multirotor rate controller controls motors and the fw rate controller controls elevons.
The idea was to run them simultaneously either because it's needed (e.g. a duorotor tailsitter needs elevons in hover to control pitch and yaw) or because it might be beneficial (quadplane elevons can help stabilize during transition).

jlecoeur · 2018-04-12T12:49:09Z

all right!
Yes, I would rather have XYZ thrust added to actuators_controls, and throttle removed. Reverse thrust would be done with negative thrust_x, and inverted multirotor flight with positive thrust_z.
Here is the branch where I added 3d thrust: Multirotor mixers 6dof jlecoeur/Firmware#1.
let's discuss during the dev call
ok my bad I was under the impression that both rate controllers controlled the motors. So in forward flight is this the mc rate controller that provides thrust command?

dagar · 2018-04-12T13:21:22Z

Yes, thanks for clarifying this. My first pass was lazy find and replace.
I'm interested in reverse thrust for planes in the near future.

Should we implement a vehicle_thrust_setpoint message in this PR and get the thrust out of the attitude and rate setpoints entirely?

I was putting off the actuator control changes for now as that will have ramifications for the mixers (depending on how we do it).

tailsitters Signed-off-by: Roman <bapstroman@gmail.com>

- TODO: we should define a frame, e.g. NED. Signed-off-by: Roman <bapstroman@gmail.com>

… frame Signed-off-by: Roman <bapstroman@gmail.com>

Signed-off-by: Roman <bapstroman@gmail.com>

…tter - enable yaw control via mc roll actuator to coordinate turns in fw mode Signed-off-by: Roman <bapstroman@gmail.com>

Signed-off-by: Roman <bapstroman@gmail.com>

- this means hover and transition modes for vtol Signed-off-by: Roman <bapstroman@gmail.com>

- we cannot have position control running during transitions because it's not aware of the forward flight capabilities. Therefore, tell the controller to just control altitude. Signed-off-by: Roman <bapstroman@gmail.com>

…ules for vtol Signed-off-by: Roman <bapstroman@gmail.com>

- thrust_z should be negative for positive throttle Signed-off-by: Roman <bapstroman@gmail.com>

dagar · 2018-09-16T19:14:01Z

@RomanBapst I rebased on master and fixed a few minor issues. The remaining SITL test failures are real.

jlecoeur

LGTM except for the minor comments.

@dagar I guess that the next logical step relative to 3D thrust will be to pass the 3 thrust components through actuator_controls and let the mixer allocate it to lifting/pushing actuators?

jlecoeur · 2018-09-17T08:25:09Z

msg/vehicle_attitude_setpoint.msg

-float32 thrust			# Thrust in Newton the power system should generate
+float32 thrust_x					# Throttle command in body x direction [-1,1]
+float32 thrust_y					# Throttle command in body y direction [-1,1]
+float32 thrust_z					# Throttle command in body z direction [-1,1]


Consider adding ", thrust_z is negative for upwards thrust."

jlecoeur · 2018-09-17T08:25:48Z

msg/vehicle_rates_setpoint.msg

-# TOPICS vehicle_rates_setpoint mc_virtual_rates_setpoint fw_virtual_rates_setpoint
+float32 thrust_x	   # Throttle command in body x direction [-1,1]
+float32 thrust_y	   # Throttle command in body y direction [-1,1]
+float32 thrust_z	   # Throttle command in body z direction [-1,1]


same comment as in vehicle_attitude_setpoint

jlecoeur · 2018-09-17T08:33:19Z

src/modules/mc_pos_control/mc_pos_control_main.cpp

@@ -749,7 +749,7 @@ MulticopterPositionControl::task_main()
 			matrix::Quatf q_sp = matrix::Eulerf(_att_sp.roll_body, _att_sp.pitch_body, _att_sp.yaw_body);
 			q_sp.copyTo(_att_sp.q_d);
 			_att_sp.q_d_valid = true;
-			_att_sp.thrust = 0.0f;
+			_att_sp.thrust_z = 0.0f;


General comment: should we ensure that thrust components X and Y are equal to 0 in MC (and that components Y and Z are null in FW)?

// MC _att_sp.thrust_x = 0.0f; _att_sp.thrust_y = 0.0f; _att_sp.thrust_z = value;

@jlecoeur good point! Will do.

dagar · 2018-09-19T18:16:12Z

@RomanBapst do you have an example failure log review (possible stability regression mentioned on the dev call)?

stale · 2019-01-20T05:06:08Z

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

stale · 2019-02-04T00:04:28Z

Closing as stale.

dagar added Admin: Enhancement (improvement) 💡 Hybrid VTOL 🛩️🚁 Multirotor + Fixedwing! labels Mar 29, 2018

dagar assigned dagar and RomanBapst Mar 29, 2018

dagar requested review from RomanBapst and bresch March 29, 2018 18:11

dagar force-pushed the pr-vtol_rates branch from b521a59 to d0c8fec Compare March 29, 2018 18:31

dagar force-pushed the pr-vtol_rates branch from d0c8fec to 4093529 Compare March 29, 2018 18:39

dagar requested a review from jlecoeur March 29, 2018 18:39

dagar force-pushed the pr-vtol_rates branch from 4093529 to 609d88a Compare April 1, 2018 23:28

RomanBapst force-pushed the pr-vtol_rates branch from 609d88a to b8b65af Compare April 3, 2018 08:54

dagar commented Apr 3, 2018

View reviewed changes

RomanBapst force-pushed the pr-vtol_rates branch from 9afef2a to d011aa7 Compare April 5, 2018 14:03

RomanBapst reviewed Apr 5, 2018

View reviewed changes

dagar commented Apr 5, 2018

View reviewed changes

RomanBapst force-pushed the pr-vtol_rates branch from d011aa7 to 7cc679a Compare April 11, 2018 13:12

thomasgubler added this to the Release v1.7.4 milestone Apr 11, 2018

RomanBapst force-pushed the pr-vtol_rates branch from dab01c1 to 375954b Compare April 12, 2018 08:02

RomanBapst force-pushed the pr-vtol_rates branch from 375954b to cc9d3bf Compare April 12, 2018 11:33

RomanBapst and others added 17 commits September 16, 2018 12:40

FixedWingAttitudeControl: small gain scheduling hack for airspeed less

7b100aa

tailsitters Signed-off-by: Roman <bapstroman@gmail.com>

msg: update comment about throttle setpoint components

f863792

- TODO: we should define a frame, e.g. NED. Signed-off-by: Roman <bapstroman@gmail.com>

FixedWingAttitudeControl: convert desired bodyrates between mc and fw…

1b2ed9a

… frame Signed-off-by: Roman <bapstroman@gmail.com>

tailsitter: use mulirotor throttle control during backtransition

2567b7b

Signed-off-by: Roman <bapstroman@gmail.com>

vtol: moved diff_thrust parameter to vtol class since used for tailsi…

9f1e24a

…tter - enable yaw control via mc roll actuator to coordinate turns in fw mode Signed-off-by: Roman <bapstroman@gmail.com>

tailsitter: transition code cleanup

3d40eaa

Signed-off-by: Roman <bapstroman@gmail.com>

tailsitter: only use elevons lock parameter in multirotor mode

50b5845

Signed-off-by: Roman <bapstroman@gmail.com>

vtol_att_control cleanup init and default destructors

f812e44

mc_pos_control: renamed thrust to thrust_z

2038477

Signed-off-by: Roman <bapstroman@gmail.com>

FixedWingAttiudeControl: only copy manual control topic if it updated

f489ba9

Signed-off-by: Roman <bapstroman@gmail.com>

tailsitter.h: include matrix lib

1b57bcc

Signed-off-by: Roman <bapstroman@gmail.com>

flight_tasks: only run when rotary wing:

6c0ab2d

- this means hover and transition modes for vtol Signed-off-by: Roman <bapstroman@gmail.com>

flight tasks: make transitions in altitude control

3949ea8

- we cannot have position control running during transitions because it's not aware of the forward flight capabilities. Therefore, tell the controller to just control altitude. Signed-off-by: Roman <bapstroman@gmail.com>

FixedWingAttitudeControl: added parameter to allow airspeed scaling r…

f7aaf73

…ules for vtol Signed-off-by: Roman <bapstroman@gmail.com>

fixed sign of attitude setpoint thrust_z

829c28b

- thrust_z should be negative for positive throttle Signed-off-by: Roman <bapstroman@gmail.com>

fw_att_control move static constexpr out of header

1ccfa3d

vtol mark all virtual functions override

0b29136

dagar force-pushed the pr-vtol_rates branch from f63069d to 0b29136 Compare September 16, 2018 17:15

uorb graph remove rates_sp_id special case

340df53

jlecoeur requested changes Sep 17, 2018

View reviewed changes

RomanBapst mentioned this pull request Nov 8, 2018

VTOL rate control improvements #10819

Merged

stale bot added the Admin: Wont fix label Jan 20, 2019

weekly-digest bot mentioned this pull request Jan 27, 2019

Weekly Digest (20 January, 2019 - 27 January, 2019) #11307

Closed

stale bot closed this Feb 4, 2019

dagar deleted the pr-vtol_rates branch February 24, 2019 17:20

xdwgood mentioned this pull request Jan 15, 2020

[Two motor tailsitter]Controlling the servo through the output from the mc controller #13917

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] VTOL coordinate FW and MC rate controllers only #9190

[WIP] VTOL coordinate FW and MC rate controllers only #9190

dagar commented Mar 29, 2018 •

edited by AuterionWrikeBot

Loading

dagar commented Mar 29, 2018 •

edited

Loading

RomanBapst commented Apr 3, 2018

dagar Apr 3, 2018

RomanBapst commented Apr 3, 2018

RomanBapst commented Apr 5, 2018

RomanBapst Apr 5, 2018

dagar Apr 5, 2018

RomanBapst Apr 5, 2018

dagar Apr 5, 2018

dagar Apr 5, 2018

RomanBapst Apr 11, 2018

dagar commented Apr 5, 2018

dagar commented Apr 12, 2018 •

edited

Loading

jlecoeur commented Apr 12, 2018

RomanBapst commented Apr 12, 2018

jlecoeur commented Apr 12, 2018

dagar commented Apr 12, 2018

dagar commented Sep 16, 2018

jlecoeur left a comment

jlecoeur Sep 17, 2018

jlecoeur Sep 17, 2018

jlecoeur Sep 17, 2018

RomanBapst Sep 17, 2018

dagar commented Sep 19, 2018

stale bot commented Jan 20, 2019

stale bot commented Feb 4, 2019

[WIP] VTOL coordinate FW and MC rate controllers only #9190

[WIP] VTOL coordinate FW and MC rate controllers only #9190

Conversation

dagar commented Mar 29, 2018 • edited by AuterionWrikeBot Loading

dagar commented Mar 29, 2018 • edited Loading

RomanBapst commented Apr 3, 2018

Choose a reason for hiding this comment

RomanBapst commented Apr 3, 2018

RomanBapst commented Apr 5, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dagar commented Apr 5, 2018

dagar commented Apr 12, 2018 • edited Loading

jlecoeur commented Apr 12, 2018

RomanBapst commented Apr 12, 2018

jlecoeur commented Apr 12, 2018

dagar commented Apr 12, 2018

dagar commented Sep 16, 2018

jlecoeur left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dagar commented Sep 19, 2018

stale bot commented Jan 20, 2019

stale bot commented Feb 4, 2019

dagar commented Mar 29, 2018 •

edited by AuterionWrikeBot

Loading

dagar commented Mar 29, 2018 •

edited

Loading

dagar commented Apr 12, 2018 •

edited

Loading