Flight termination

PX4 Autopilot
1

why i had flight termination during this flight?

2

I’m not so sure, but I have come across a similar report recently: Sudden Motor Stop Mid-Air [Autopilot disengaged, switching to terminate]

For some reason it could no longer go into Stabilized mode.

image
image1311×591 91.2 KB

It basically had no longer an attitude estimate:

image
image1136×593 33.8 KB

Not sure why that would happen though.

What I can see is that you have both high accelerometer bias warnings and at the same time mag warnings. The mag warnings are due to the fact that the compass is measuring a lot of interference from your motor/ESC wiring, as you can see given the correlation here:

image
image1132×506 124 KB

3

what can cause bias warning?

4

Bad calibration and/or bad vibrations.

5

@ShlomiRR thanks for sharing your experience.

We had a very similar issue; in our case we used a VTOL in standard quadplane configuration and we experienced the issue during FW mode. We were using PX4 v1.15.4. It looks like flight termination was triggered in flight, leading to loss of control and subsequent crash. Unfortunately we could not recover the .ulg log, but looking at the telemetry .tlog file logged by QGC we were able to reconstruct the following:

  • “Preflight Fail: High Accelerometer Bias”
  • In the next seconds, attitude estimate was declared invalid (we got NaN values in the ATTITUDE_QUATERNION mavlink message), and immediately after, also the angular velocity estimate was invalid
  • “Failsafe activated”
  • “Preflight Fail: Flight termination active”

I did some research and I found this other issue which looks related: Autopilot disengaged, switching to terminate @diana_diaz

Based on your log I see that at a certain point the flag estimator_sensor_bias/accel_bias_stable goes to 0, and from that point on the accel_bias estimates diverge and saturate to the limit value.

We never experienced this kind of issue before with previous PX4 versions; I guess it might be related to the fact a new EKF implementation was introduced in v1.15, @JulianOes what do you think? I also suspect it might be related to IMU autocalibration.

Screenshot 2025-08-17 114018
Screenshot 2025-08-17 1140181306×901 37.1 KB

6

Hi there. As the topic came up again and I myself had such an unfortunate experience before I looked more into it again and have the impression that using the quick mag calibration that was enabled in v.1.15 could have led to this result. I just recently red more into it and it seems that quick mag calibration can’t replace the normal one (as I assumed). And as far as information out there are concerned it can lead to yaw estimate errors and also issues with ekf. Maybe that helps?

7

@MikeRogerIt thanks for sharing your experience.

Based on my understanding, there are two issues to be addressed here:

  1. investigate why the EKF fails, resulting in an invalid attitude estimate;
  2. discuss whether it is appropriate that the failsafe state machine does lead to flight termination in case of loss of attitude estimate (for instance in manual flight, fixed wing mode, one would still be able to fly even without attitude estimate available).

In my opinion point #1 is the most critical one - losing the attitude estimate would be catastrophic with multirotors in particular. To be honest, I do not feel much confident flying knowing that the EKF may fail at any time…

There have been already multiple crash reports which look related to this issue (and there may be even more unreported…), and it might occur again if this is not addressed. I think this is an issue of concern and should be investigated more in depth.

Any comments? @JulianOes

8

Looks like this issue