WIP: Control Allocation #13351

jlecoeur · 2019-11-01T13:27:37Z

Work in progress control allocation module.

Edit: We decided to bring it in with multicopter support only for the start, and bring in FW/VTOL support in a followup. I've grouped the WIP items accordingly.

Minimal multicopter support (with this PR):

Followup PR:

jlecoeur · 2019-11-01T13:30:11Z

Many Ci failures will be fixed by #13336

src/modules/control_allocator/ControlAllocation/ControlAllocation.cpp

src/modules/control_allocator/ControlAllocator.cpp

LorenzMeier · 2019-11-17T10:16:33Z

You might want to rebase on master to fix the Mac OS build.

jlecoeur · 2019-11-18T17:58:47Z

Here is a demo of what can be done with this new module:

Motor failure handling!

On an hexacopter robot, two motors are manually disabled with minimum impact on the flight: the robot keeps hovering without drifting.

As motors are disabled (simulating motor failures), a new allocation matrix is automatically computed and the hexacopter keeps flying as if it was an asymmetric pentacopter, then a quadcopter.

The robot is not disturbed because the asymmetry is immediately compensated for in the allocation module. Without the logic implemented in the allocation module, the robot would drift sideways and downward while the integrators build up to compensate.

jkflying · 2019-11-19T20:38:36Z

On an hexacopter robot, two motors are manually disabled with minimum impact on the flight: the robot keeps hovering without drifting.

So the only piece missing from a real vehicle would be detecting the motor is missing and removing it from the allocation table?

jlecoeur · 2019-11-19T20:54:42Z

@jkflying Yes!
However note that it would not look as nice in a real world scenario. Until the failure is detected, the platform would be disturbed and integrators would build up, then, once the allocation table is updated, the integrators need to be emptied. In other words the disturbance would last a little longer than the failure detection time.

src/lib/mixer/mixer_allocated_actuator.cpp

platforms/common/include/px4_platform_common/px4_work_queue/WorkQueueManager.hpp

stale · 2020-02-29T18:49:24Z

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

…out-of-memory

…n updated Reduces stack + RAM usage

@1KHz

And limit the operations to the number of configured outputs. Only using the number of configured actuators reduces CPU load by ~2% on F7 @1KHz.

Mostly important in cases where the ouputs would change (e.g. param change)

…ting the matrix Just to be on the safe side.

LorenzMeier · 2021-01-01T15:15:26Z

@jlecoeur @bkueng @RomanBapst Let's take the time to wrap this up!

- control allocation module with multirotor, VTOL standard, and tiltrotor support - angular_velocity_controller - See #13351 for details Co-authored-by: Julien Lecoeur <julien.lecoeur@gmail.com> Co-authored-by: Silvan Fuhrer <silvan@auterion.com> Co-authored-by: Roman Bapst <bapstroman@gmail.com>

- control allocation module with multirotor, VTOL standard, and tiltrotor support - angular_velocity_controller - See #13351 for details Co-authored-by: Silvan Fuhrer <silvan@auterion.com> Co-authored-by: Roman Bapst <bapstroman@gmail.com>

mengchaoheng · 2021-01-20T07:08:15Z

Here is a demo of what can be done with this new module:

Motor failure handling!

On an hexacopter robot, two motors are manually disabled with minimum impact on the flight: the robot keeps hovering without drifting.

As motors are disabled (simulating motor failures), a new allocation matrix is automatically computed and the hexacopter keeps flying as if it was an asymmetric pentacopter, then a quadcopter.

The robot is not disturbed because the asymmetry is immediately compensated for in the allocation module. Without the logic implemented in the allocation module, the robot would drift sideways and downward while the integrators build up to compensate.

This is really cool! But I have a question. How does the system know that a motor has failed? Does it imply that there is fault detection in the system?

jlecoeur · 2021-01-20T09:28:08Z

@mengchaoheng yes you would need a detection mechanism. The failure is set manually in the demo above.

julianoes · 2021-10-20T20:56:27Z

ROMFS/px4fmu_common/init.d-posix/10017_iris_ctrlalloc

+# @name 3DR Iris Quadrotor SITL
+#
+# @type Quadrotor Wide
+#
+# @maintainer Julian Oes <julian@oes.ch>


I don't think so 😄

JacobCrabill · 2021-10-27T23:24:22Z

src/modules/control_allocator/ActuatorEffectiveness/ActuatorEffectivenessStandardVTOL.hpp

+	 */
+	void setFlightPhase(const FlightPhase &flight_phase) override;
+
+	int numActuators() const override { return 7; }


Suggested change

int numActuators() const override { return 7; }

int numActuators() const override { return 8; }

This may explain why standard_vtol never worked with control allocation.

sfuhrer · 2022-01-07T16:44:40Z

I'd say with #16528 and #18776 being merged we can close this one.

dagar self-requested a review November 1, 2019 17:52

dagar added the Admin: Enhancement (improvement) 💡 label Nov 1, 2019

jlecoeur force-pushed the pr-control_allocation branch 2 times, most recently from 798399f to 6a7f1b3 Compare November 2, 2019 11:11

RomanBapst reviewed Nov 4, 2019

View reviewed changes

src/modules/control_allocator/ControlAllocation/ControlAllocation.cpp Outdated Show resolved Hide resolved

RomanBapst reviewed Nov 4, 2019

View reviewed changes

src/modules/control_allocator/ControlAllocator.cpp Show resolved Hide resolved

RomanBapst reviewed Nov 4, 2019

View reviewed changes

src/modules/control_allocator/ControlAllocator.cpp Outdated Show resolved Hide resolved

jlecoeur force-pushed the pr-control_allocation branch from 59d024d to 4646f2a Compare November 8, 2019 10:56

jlecoeur force-pushed the pr-control_allocation branch 2 times, most recently from 2ca4fd2 to 38a5391 Compare November 16, 2019 19:09

jlecoeur force-pushed the pr-control_allocation branch from 8ac3340 to d5a7215 Compare November 17, 2019 10:34

This was referenced Nov 19, 2019

Update matrix submodule and add tests #13517

Merged

[WIP]: Rate control: introduce torque_setpoint, angular_accel_setpoint, and inertia matrix #13219

Closed

Multirotor mixers 6dof jlecoeur/Firmware#1

Closed

dagar reviewed Nov 22, 2019

View reviewed changes

src/lib/mixer/mixer_allocated_actuator.cpp Outdated Show resolved Hide resolved

jlecoeur force-pushed the pr-control_allocation branch from 8c2f446 to 9c0b4af Compare November 23, 2019 19:16

jlecoeur mentioned this pull request Nov 23, 2019

WIP: Add mixer type AllocatedActuatorMixer #13581

Closed

jlecoeur added the [ARCHIVED] mixer Archived after v1.14 release where Mixer was removed label Nov 23, 2019

jlecoeur mentioned this pull request Nov 25, 2019

WIP: Vehicle model estimator (onboard mass & inertia estimation) #13585

Closed

jlecoeur force-pushed the pr-control_allocation branch 2 times, most recently from 5a712f2 to f45c17f Compare November 28, 2019 19:07

dagar reviewed Nov 29, 2019

View reviewed changes

platforms/common/include/px4_platform_common/px4_work_queue/WorkQueueManager.hpp Outdated Show resolved Hide resolved

jlecoeur mentioned this pull request Feb 14, 2020

Inquiry and discovery of Fault Tolerant control system in PX4 in case of Actuator Faults mid flight. #13857

Closed

stale bot added the stale label Feb 29, 2020

bkueng added 9 commits January 1, 2021 16:11

fix control_allocator: use 'delete' instead of 'free', guard against …

703fb00

…out-of-memory

fix control_allocator: set _last_run at the correct place

54dfdb0

refactor control_allocator: directly get the effectiveness matrix whe…

de10a3c

…n updated Reduces stack + RAM usage

control_allocator: change SequentialDesaturation to existing MC mixer

8c4cfe6

And limit the operations to the number of configured outputs. Only using the number of configured actuators reduces CPU load by ~2% on F7 @1KHz.

control_allocator: avoid vector copies and sqrt()

fde6fb8

control_allocator: inline one-line setters & getters

f8d245d

control_allocator: ensure unused outputs are initialized to min

bdacca3

Mostly important in cases where the ouputs would change (e.g. param change)

control_allocator: set unused actuators to trim instead of min

2d3d6e8

ActuatorEffectivenessMultirotor: set effectiveness to zero when compu…

014d6ac

…ting the matrix Just to be on the safe side.

dagar mentioned this pull request Jan 10, 2021

initial control allocation support (safe to merge) #16528

Merged

julianoes reviewed Oct 20, 2021

View reviewed changes

JacobCrabill reviewed Oct 27, 2021

View reviewed changes

sfuhrer closed this Jan 7, 2022

junwoo091400 mentioned this pull request May 6, 2022

[Project Tracker] Control Allocation User experience improvement #19444

Open

19 tasks

dagar deleted the pr-control_allocation branch October 9, 2022 19:11

VedantBhosle31 approved these changes Oct 24, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP: Control Allocation #13351

WIP: Control Allocation #13351

jlecoeur commented Nov 1, 2019 •

edited by dagar

Loading

jlecoeur commented Nov 1, 2019

LorenzMeier commented Nov 17, 2019

jlecoeur commented Nov 18, 2019

jkflying commented Nov 19, 2019

jlecoeur commented Nov 19, 2019

stale bot commented Feb 29, 2020

LorenzMeier commented Jan 1, 2021

mengchaoheng commented Jan 20, 2021

jlecoeur commented Jan 20, 2021

julianoes Oct 20, 2021

JacobCrabill Oct 27, 2021

JacobCrabill Oct 27, 2021

sfuhrer commented Jan 7, 2022

	int numActuators() const override { return 7; }
	int numActuators() const override { return 8; }

WIP: Control Allocation #13351

WIP: Control Allocation #13351

Conversation

jlecoeur commented Nov 1, 2019 • edited by dagar Loading

jlecoeur commented Nov 1, 2019

LorenzMeier commented Nov 17, 2019

jlecoeur commented Nov 18, 2019

jkflying commented Nov 19, 2019

jlecoeur commented Nov 19, 2019

stale bot commented Feb 29, 2020

LorenzMeier commented Jan 1, 2021

mengchaoheng commented Jan 20, 2021

jlecoeur commented Jan 20, 2021

julianoes Oct 20, 2021

Choose a reason for hiding this comment

JacobCrabill Oct 27, 2021

Choose a reason for hiding this comment

JacobCrabill Oct 27, 2021

Choose a reason for hiding this comment

sfuhrer commented Jan 7, 2022

jlecoeur commented Nov 1, 2019 •

edited by dagar

Loading