WIP: Fix FMU MC failsafe on RC loss #5388

kd0aij · 2016-08-26T00:11:04Z

This fixes some problems with RC loss failsafe when in manual modes on fmu-v4, but may break other things since it's not clear to me how it was intended to work.

It appears that failsafe would never be entered if in manual mode with a GPS connected.
Also, px4fmu quit running the output mixer when RC was lost.

LorenzMeier · 2016-08-26T09:35:18Z

Please rebase on master to fix CI.

LorenzMeier · 2016-08-26T11:12:50Z

@julianoes Can you have a look?

julianoes · 2016-08-26T12:35:23Z

src/modules/mc_att_control/mc_att_control_main.cpp

@@ -889,6 +891,12 @@ MulticopterAttitudeControl::task_main()
 			vehicle_status_poll();
 			vehicle_motor_limits_poll();

+			/* Simple handling of failsafe: kill motors if failsafe is on */
+			if (_termination_enabled && _v_control_mode.flag_control_termination_enabled) {
+				_armed.lockdown = true;


This won't have any effect. The _armed topic is subscribed here, not published.

Right, I guess all we can do here is deploy a parachute, as in the FW case.

kd0aij · 2016-08-26T23:00:46Z

The remaining open question on this fix is how to repair the commander logic which sets flags.gps_failure. Since this is never set if there is no GPS module (one known problem case), there is currently no way to configure RC loss to cause motor shutdown (termination). My proposed fix works, but precludes a controlled descent on RC loss. Of course at this point (line 2703 of commander.cpp) we're already in termination state, not land or descend...

Also, even if GPS were available, would it be wise to execute a powered descent, given that we have no idea what is below the vehicle?

mhkabir · 2016-08-27T04:46:25Z

Even without GPS, the failsafe configuration should have an openloop descent as default. With GPS available, we can just close the altitude control loop. Even if we have no idea about what's below / exactly how high we are, it's safer than killing the motors.

kd0aij · 2016-08-27T12:49:23Z

@mhkabir That's debatable. The FAI rules (http://www.fai.org/downloads/ciam/SC4_F3FPV_2016) for FPV racers state

The model must be equipped with a fail-safe device, the triggering of which stops the motorization.

and

In case of emergency, the helper is authorised to shut off the transmitter in order to trigger the fail-safe
device.

mhkabir · 2016-08-27T15:59:27Z

I was talking about the general config (for most users). For racers, killing motors is obviously something we will have to support, as an extension of the flight termination state.

kd0aij · 2016-08-27T18:47:28Z

Right now, all users will see a fly-away, and at least one user (besides me) would like to be able to prevent such out-of-control situations:
#5357 (comment)

kd0aij · 2016-08-28T16:28:41Z

I'll close this in another day or two unless there is further useful discussion

mhkabir · 2016-08-28T16:48:11Z

@kd0aij, I will be joining this effort to fix failsafe handling. Just reviewing the state machine code now.
Thanks a lot for the initiative!

julianoes · 2016-08-31T08:12:43Z

Sorry @kd0aij for the lack of work and communication here. This is something that needs to be fixed for sure. I'll assign myself to keep it on the radar.

kd0aij · 2016-09-20T15:54:46Z

@mhkabir @julianoes any progress on this?

julianoes · 2016-09-27T10:39:40Z

@kd0aij sorry I'm pretty busy. Will tackle it when I get a chance.

LorenzMeier · 2016-10-08T11:12:22Z

@julianoes Any chance to get back to this?

julianoes · 2016-10-10T20:45:22Z

@kd0aij Ok I started debugging this with a Pixracer and a FrSky D4RII receiver and the first thing of master I tested was:
Flying in manual without GPS connected and switching off RC. It successfully enters "Land" mode and lowers overall PWM trying to descend. At some point it switches to min PWM and detects landed.
Does this fit with your observation?

kd0aij · 2016-10-10T20:52:19Z

I believe so. My PR addressed the case with GPS valid:
#5388 (comment)

On Mon, Oct 10, 2016 at 2:45 PM Julian Oes notifications@github.com wrote:

@kd0aij https://github.com/kd0aij Ok I started debugging this with a
Pixracer and a FrSky D4RII receiver and the first thing of master I tested
was:
Flying in manual without GPS connected and switching off RC. It
successfully enters "Land" mode and lowers overall PWM trying to descend.
At some point it switches to min PWM and detects landed.
Does this fit with your observation?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#5388 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/ACMZPQixkuz254gnE9IBNYJF5xkhNWhzks5qyqPkgaJpZM4Jtn_m
.

kd0aij · 2016-10-10T20:54:25Z

I'm sure I never (intentionally) flight tested RC loss under any
circumstances since I wasn't satisfied with the documentation.

On Mon, Oct 10, 2016 at 2:52 PM Mark Whitehorn kd0aij@gmail.com wrote:

I believe so. My PR addressed the case with GPS valid:
#5388 (comment)

On Mon, Oct 10, 2016 at 2:45 PM Julian Oes notifications@github.com
wrote:

@kd0aij https://github.com/kd0aij Ok I started debugging this with a
Pixracer and a FrSky D4RII receiver and the first thing of master I tested
was:
Flying in manual without GPS connected and switching off RC. It
successfully enters "Land" mode and lowers overall PWM trying to descend.
At some point it switches to min PWM and detects landed.
Does this fit with your observation?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#5388 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/ACMZPQixkuz254gnE9IBNYJF5xkhNWhzks5qyqPkgaJpZM4Jtn_m
.

julianoes · 2016-10-10T20:55:39Z

Ok, I'll test with GPS next up. Unfortunately, I don't have the cable to wire the GPS up with the Pixracer. I'll have to do the test with a Pixhawk.

julianoes · 2016-10-11T10:12:00Z

@kd0aij: This fixes some problems with RC loss failsafe when in manual modes on fmu-v4, but may break other things since it's not clear to me how it was intended to work.

Which behaviour did you want and test?? RTL or all props off?

kd0aij · 2016-10-11T12:19:01Z

In non-assisted modes, this PR ensures that flight_termination state is entered, and that results in "lockdown", which is all motors stopped. No RTL.

kd0aij · 2016-10-11T12:21:15Z

The remaining open question on this fix is how to repair the commander logic which sets flags.gps_failure. Since this is never set if there is no GPS module (one known problem case), there is currently no way to configure RC loss to cause motor shutdown (termination). My proposed fix works, but precludes a controlled descent on RC loss. Of course at this point (line 2703 of commander.cpp) we're already in termination state, not land or descend...

Also, even if GPS were available, would it be wise to execute a powered descent, given that we have no idea what is below the vehicle?

julianoes · 2016-10-11T16:28:31Z

Also, even if GPS were available, would it be wise to execute a powered descent, given that we have no idea what is below the vehicle?

According to the commander state machine that's what we want to do. And I think it's what most other drones would do as well. If a drone comes down slowly it gives people the chance to step away. If it falls like a brick that's harder.

I think I now understood that the RC loss issue you're addressing, is for the case where you want termination for racers. I think this is not properly implemented (and you're fixes might be the missing piece). I'll look closer at this case but I need to find the correct params first. As you said, documentation is hard to find about this 😢.

PX4BuildBot · 2016-10-19T19:52:42Z

Can one of the admins verify this patch?

LorenzMeier · 2016-10-19T20:27:29Z

Jenkins add to whitelist.

LorenzMeier · 2016-11-11T07:50:37Z

@kd0aij Can you rebase?

kd0aij · 2016-11-11T16:19:14Z

The remaining open question on this fix is how to repair the commander logic which sets flags.gps_failure. Since this is never set if there is no GPS module (one known problem case), there is currently no way to configure RC loss to cause motor shutdown (termination). My proposed fix works, but precludes a controlled descent on RC loss. Of course at this point (line 2703 of commander.cpp) we're already in termination state, not land or descend...

And for FPV racers, RC loss MUST result in motor shutdown, if I understand FAI rules correctly.

gregd72002 · 2016-11-11T17:55:10Z

I would also like to see motor shutdown option on RC loss +1

kd0aij · 2016-11-11T19:05:13Z

retesting: enters "LAND" mode on RC loss (Pixracer, no GPS, FrSky RX)
Required a mod to commander to force lockdown in addition to failsafe (not sure what the force_failsafe flag accomplishes).
Note that the default value for CBRK_FLIGHTTERM disables this behavior.

Needs testing on fmu-v2/v3

kd0aij · 2016-11-11T20:43:52Z

Jenkins test this please

kd0aij · 2016-11-13T15:29:20Z

@julianoes What do you think of this as an interim workaround for motor shutdown? If flight termination is disabled, the copter should just switch to LAND mode without shutting down the motors.

kd0aij · 2016-12-17T20:18:42Z

superseded by #5863 ?

kd0aij changed the title ~~WIP: Fix fmu mc failsafe on RC loss~~ WIP: Fix FMU MC failsafe on RC loss Aug 26, 2016

kd0aij force-pushed the fix_fmu_mc_failsafe branch from b4e574a to f7979c2 Compare August 26, 2016 12:24

julianoes reviewed Aug 26, 2016
View reviewed changes

julianoes self-assigned this Aug 31, 2016

kd0aij added 3 commits November 11, 2016 08:00

always enter failsafe if in manual mode and RC lost

0b5ffd6

fix logic error which prevented mixing when RC lost

ed2565c

add block for handling flight termination nav state

8fdb5f9

kd0aij force-pushed the fix_fmu_mc_failsafe branch from 31ab75d to 8fdb5f9 Compare November 11, 2016 16:10

run astyle

757dea5

force lockdown on RC loss if flight termination enabled

a1d835b

kd0aij mentioned this pull request Nov 16, 2016

Feature/First iteration on MC failsafe #5863

Closed

kd0aij closed this Dec 17, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP: Fix FMU MC failsafe on RC loss #5388

WIP: Fix FMU MC failsafe on RC loss #5388

kd0aij commented Aug 26, 2016

LorenzMeier commented Aug 26, 2016

LorenzMeier commented Aug 26, 2016

julianoes Aug 26, 2016

kd0aij Aug 26, 2016

kd0aij commented Aug 26, 2016 •

edited

mhkabir commented Aug 27, 2016

kd0aij commented Aug 27, 2016 •

edited

mhkabir commented Aug 27, 2016

kd0aij commented Aug 27, 2016

kd0aij commented Aug 28, 2016

mhkabir commented Aug 28, 2016

julianoes commented Aug 31, 2016

kd0aij commented Sep 20, 2016

julianoes commented Sep 27, 2016

LorenzMeier commented Oct 8, 2016

julianoes commented Oct 10, 2016

kd0aij commented Oct 10, 2016

kd0aij commented Oct 10, 2016

julianoes commented Oct 10, 2016

julianoes commented Oct 11, 2016

kd0aij commented Oct 11, 2016

kd0aij commented Oct 11, 2016

julianoes commented Oct 11, 2016

PX4BuildBot commented Oct 19, 2016

LorenzMeier commented Oct 19, 2016

LorenzMeier commented Nov 11, 2016

kd0aij commented Nov 11, 2016

gregd72002 commented Nov 11, 2016

kd0aij commented Nov 11, 2016

kd0aij commented Nov 11, 2016

kd0aij commented Nov 13, 2016

kd0aij commented Dec 17, 2016

WIP: Fix FMU MC failsafe on RC loss #5388

WIP: Fix FMU MC failsafe on RC loss #5388

Conversation

kd0aij commented Aug 26, 2016

LorenzMeier commented Aug 26, 2016

LorenzMeier commented Aug 26, 2016

julianoes Aug 26, 2016

Choose a reason for hiding this comment

kd0aij Aug 26, 2016

Choose a reason for hiding this comment

kd0aij commented Aug 26, 2016 • edited

mhkabir commented Aug 27, 2016

kd0aij commented Aug 27, 2016 • edited

mhkabir commented Aug 27, 2016

kd0aij commented Aug 27, 2016

kd0aij commented Aug 28, 2016

mhkabir commented Aug 28, 2016

julianoes commented Aug 31, 2016

kd0aij commented Sep 20, 2016

julianoes commented Sep 27, 2016

LorenzMeier commented Oct 8, 2016

julianoes commented Oct 10, 2016

kd0aij commented Oct 10, 2016

kd0aij commented Oct 10, 2016

julianoes commented Oct 10, 2016

julianoes commented Oct 11, 2016

kd0aij commented Oct 11, 2016

kd0aij commented Oct 11, 2016

julianoes commented Oct 11, 2016

PX4BuildBot commented Oct 19, 2016

LorenzMeier commented Oct 19, 2016

LorenzMeier commented Nov 11, 2016

kd0aij commented Nov 11, 2016

gregd72002 commented Nov 11, 2016

kd0aij commented Nov 11, 2016

kd0aij commented Nov 11, 2016

kd0aij commented Nov 13, 2016

kd0aij commented Dec 17, 2016

kd0aij commented Aug 26, 2016 •

edited

kd0aij commented Aug 27, 2016 •

edited