PixRacer 1.5.5 fly away


After good results on previous flights, today ended up being rather scary.
After a manual takeoff, I switched to the auto hold mode. Initially, it appeared to be working as it did previously. However, while apparently maintaining position, the drone kept climbing. Once realising this I attempted to take back manual control and subsequently pulled the manual kill switch on the RC. No reaction at all from the quad.

Only difference to previous flights is the addition of a servo on AUX2 and the extras.txt file with the following content:

#start mount driver
vmount start

The log appears to be stuck in the queue (http://logs.uaventure.com/view/qfy8LGvcBLGjkvtCJmFvz8) so here is the px4log file: https://www.dropbox.com/s/0asoyi92umdp58j/15_46_05.px4log?dl=1

I also have a video recording, at about 20 seconds in I switched to auto hold. I cut out about 3 minutes of constant climb. https://www.youtube.com/watch?v=HXISXxW-BhA

The RC transmitter was showing a solid connection the entire time and none of the fail-safes nor the geofence worked at all. After it crashed the quad was still powered but the safety button led was off.

The px4log from dropbox only contains 16 seconds and manual mode. Any chance you have more of the log?

That’s the complete file as I found it on the sdcard.

Are there any other logs that might be of use?

16 seconds sounds about the time when I switched to auto hold mode.

Is there anything else on the sdcard? Did you save the mavlink log in QGC?
Unfortunately it sounds like the autopilot stopped responding. Would you be able to do some bench testing with an identical setup to see if you can reproduce it?

The only other file related to that flight I can find is the preflight_perf:

I was not connected to a ground station, so sadly no logs on that front.
I did a few quick tests on the bench without props but did not manage to reproduce the issue yet. I’ll keep at it though.

It seems like at least the stabilisation and position hold was still working up till the point when the battery was depleted. Any ideas of what could lead to such an outcome?

More on accident than really planning on it, it seems that I’m now able to replicate the issue reliably. That said I’m not 100% sure it’s the exact same issue that caused the fly away but seeing that it only occurs when the vmount driver was started seems to be pointing that way.


Two tests: once with and once without the extras.txt, after take-off (manual mode) the uart example program is started.
Only when the vmount driver was not started the connection stays active and manual control can be regained at any time. If the vmount driver was running the drone stops responding as soon as the uart example program is launched. No reaction to kill switch, Qgroundcontrol looses connection …

It would be great if someone could give this a shot and see if they can reproduce this behaviour.

I was not able to reproduce this on the 1.6.0rc beta.


I have the same behaviour using the vmount driver in Follow Me mode (Pixhawk 2.4.5, PX4 current master, image < 1Mb), but only when MNT_MODE_IN is set to MavLink ROI mode. MavLink Do Mount does nothing and in RC mode all gimbal axes are controllable.

Manual, Attitude, PosHold work, but in the moment I switch to Follow Me mode (QGC on Android), the controller seems to crash and failsafe kicks in until batteries are drowned, no usable log data (stops in the moment of switching the mode). Can replicate on ground without props mounted.

Follow Me mode basically works without setting MavLink ROI mode.


Instructions for reproducing are really helpful. Can we move this over to a github issue to get a developer familiar with vmount looking at it?

Done https://github.com/PX4/Firmware/issues/6783

I’ve provided feedback for an initial attempt at fixing it on the Github thread. Other than ruling out a stack overflow its worth checking what else could go wrong at the global position calculation step.