APM Copter and the "bad" YAW control, a very old story...

[Marco Robustini]

Hi all, I would treat a theme already well-known, long ago we discussed what signaled to me several times without finding a solution that would solve the problem completely.
APM Copter suffers from always a bit of a bad YAW control YAW, often the quad tilts abnormally during a change of rotation performed on the YAW, as often rises to share.

This problem is easily verifiable, try to turn the quad with YAW (in Alt-Hold is easy), it does not matter too quickly, then take an opposite command to change the direction of rotation, if you do this several times you will see the quad tilt or otherwise behave abnormally.
With the same hardware I found that other flight control board (WKM, Mikrokopter, A2) do not have this behavior.
I tried a lot to play with the parameters without ever having acceptable results.
I thought a lot about what might be, in my opinion is not managed well the YAW between deceleration and acceleration, ie in the downturn there is a thrust equal to the same that is when the quad is changing the direction of rotation, and this is not correctly.
Expect your impressions on the issue raised by me, I can testify with video what I'm writing.

Bests, Marco

[Jason Short]

I've seen this too and I wonder if it is why the quad can't do position hold and yaw at the same time. If I yaw 180 the quad will drift quickly about two meters and slowly come back. I assume it tilted while yawning and blew itself off course. 

Could it be a CG issue where the access are not in the center and seeing centrifugal forces? Or level errors are building up as it rotates. 

Jason

Sent from my iPhone

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Olivier ADLER]

I've seen that too (Y6 frame), even on 3.1 RC4 i did try today. I did try to yaw in Loiter + simple mode, i had about 3 - 4 meters of drift and slow come back when releasing the Yaw stick.

At first i thought that the problem source was I terms not rotated at all, or not rotated fast enough. Could we have a remaining bug here ? Or a filter that should be dynamically changed during Yaw to allow for a faster correction ?



Olivier

[Marco Robustini]

Something around there is definitely wrong, the problem is more clearly if you give opposites yaw commands fast enough, for example stopping the quad every 180° with opposite command.
It's obvious that with continuous yaw rotation if the GC is not perfect there may be a drift, this makes it even WooKong, but I'm talking about another problem.

Marco

[Marco Robustini]

Imho the forces applied to the motors during yaw command are not well balanced and is much more noticeable on heavy quad with large props, maybe there is something wrong in the code.

Marco

On Sunday, October 20, 2013 2:01:05 PM UTC+2, Marco Robustini wrote:

[Olivier ADLER]

If this is an accelerometer shift induced by centripetal forces during yawing, then perhaps it could be possible to compensate for it using a calibration procedure ?



Olivier

[matt...@gmail.com]

Me too seeing the drift in Loiter when trying to Pan.

[Darrell Burkey]

On Sunday, 20 October 2013 23:01:05 UTC+11, Marco Robustini wrote:

With the same hardware I found that other flight control board (WKM, Mikrokopter, A2) do not have this behavior.

I've always wondered if the other controllers mentioned also configure the motors the same way the WKM does. With a coax the WKM spins all the motors the same direction. So all the upper props go the same direction and all the lower props go the same direction. Unlike the APM which has two props going one way and two going the other way. I'm assuming DJI had a good reason for doing it that way. Or does it make no difference?  

[Randy Mackay]

     If someone wants to do an objective test with the same frame but just changing the order, I'm happy to make the code changes available for the test.  It's a very easy change..just a couple of lines in the AP_Motors classs for the frame in question.

-Randy

[Leonard Hall]

There are a couple of different issues.

Marco is talking about a motor mixing issues relating to motor mixing, the stability patch, and motor torque/lift linearity. The end result of not properly understanding, implementing, or tuning results in an increase in total lift. This tends to cause an increase in altitude after aggressive yawing. I have a theory on this problem and a yaw controller design improvement that I hope will correct this. However, I can't do this until Randy and I fix what we are calling the onion.

The second issue related to the inertial navigation and is cause by the fact we don't rotate the acceleration corrections to the body frame. Randy initially did this but we removed it and kept it in the body frame because of the interactions between the horizontal inertial feedback and the altitude feedback. This was messing up Alt Hold. The solution is to rotate the accelerations to the horizontal body frame. We have known about this problem and the likely solution for a while but there is only so much time in the day.

Leonard

[Marco Robustini]

100% Leonard, this is what i mean...

Marco