Data_ekf_yaw_reset

Hello, Could someone please explain to me the meaning of those ev 62 DATA_EKF_YAW_RESET ? I’m getting a few of them and don’t know if they are OK or if they mean something that I should mind.

It means the EKF has reset what it knows as the yaw value. It shouldn’t be happening in a regular basis during one flight, but it’s not a major cause of concern if it happens occasionally.
EKF resets the yaw (among other times) when the compass is bad and changes to a different one.

1 Like

Thanks Francisco.

I ran an autotune session yesterday and it happened twice, however prior to that i’ve recalibrated the compass using the internal procedure and redid the compassmot. I will follow it closely to see if it keeps happening.

I’m not 100% sure, but I think it always happen a little time after you take off. So if you only see it that the beginning of a flight it should be normal.

It is a known issue. I think Randy said it will be fixed in rc3.

1 Like

Thanks for sharing wjaxxx. Is it somehow dangerous ?

I really do not know. Maybe Paul or Randy can tell us more.
Apm 3.4 rc2 seems like a very mature release. Hope to see soon the Rc3 and a final one.

I’m the one who was having that issue all the time…
Due to my GPS not maintaining a good communication speed, which I haven’t solved yet… (replacing the GPS didn’t fix it!)
It’s dangerous in that all of a sudden your copter might be facing the wrong direction! It’s usually only a few degrees, but I’ve had it quickly spin around 180 degrees.

-rc3 will be coming this week for sure.
I’m not sure if this reset issue fix will be included or not. Paul has a fix but it needs more developer testing.
As @OXINARF says, it’s normal for it to happen after each takeoff. In the problem log (from @wicked1 I think) it appeared many dozens of times.

1 Like

The EKF uses a simpler magnetic heading calculation when on the ground and before climbing above 5m due to the possibility of ground based magnetic interference. When it climbs above 5m, it switches to a more accurate 3D method of using magnetometer data and also resets the yaw angle (usually by less than 5deg). This reset event is communicated to the yaw controller which adjusts the demanded yaw to prevent any change in yaw angle that could confuse the pilot.

A yaw reset can also occur earlier if the copter detects the magnetic field changing significantly as it climbs away from the ground. This is done to prevent a large ground based magnetic anomaly (eg taking off from a car roof) from causing toilet bowling after takeoff.

The problem for the pilot arises if the EKF using IMU1 runs into a significant sensor error and there is switch-over to the EKF using IMU2. This can cause significant yaw step events that are not communicated to the control loops. There is a PR that fixes this vulnerability which is being reviewed.

1 Like

Thanks for the PR and explanation Priseborough!

Is this problem fixed? I flew a brand new quad with arducopter 3.3 with no problems. Then I updated to 3.4.1 and got this error dozens of times in a short period of time. The drone also flew very erratically for several seconds. I was in loiter both times. I didn’t autotune with 3.4. Nor have I reloaded 3.3.

Here is the log if anyone is interested.

Thanks,

Davehttps://www.dropbox.com/s/vx67zcyt2p398dh/2016-11-13%2011-55-39.bin?dl=0

1 Like

Yes, fixed. I can’t view the log file now. I had it once when my gps stand got loose… I’ll analyze it later and will get back

Thanks for the reply. Certainly appreciate any thoughts.

Dave

skyveyor,

I’m sure you’ve seen it but Paul’s responded on the other thread you opened. Intermittenly erratic flight behavor

We discussed the EKF lane switching today on the dev call a fair bit. It’s still under discussion but it seems while many users are fine, some are not. I think Paul will likely be making some changes to the switching logic in a future patch release.

Randy,

Thanks for passing this one on. I saw the post from Paul. Great insight.

Dave