Flight Termination triggered due to attitude failure – aircraft disarmed in mid-air

PX4 Autopilot
1

Hi everyone,

I’m trying to understand why PX4 triggered Flight Termination during flight.

Setup

  • PX4 Version: 1.16.2
  • Flight Controller: MicoairH743
  • Vehicle: Quadcopter

What happened

The vehicle was flying normally when PX4 reported an Attitude failure (pitch). Immediately after that:

  • Flight Termination was triggered.
  • The aircraft disarmed in mid-air.
  • The motors stopped, and the vehicle fell to the ground.

The aircraft did not visibly roll or pitch before the disarm.

What I observed

  • Flight Review shows the estimated roll suddenly jumps to ~170°, while the roll setpoint remains near 0°.
  • A vibration spike occurs at the same time.
  1. What could cause the EKF to suddenly estimate a ~170° roll when the vehicle did not actually roll?
  2. Can accelerometer clipping or vibration cause this type of attitude failure?
  3. Are there any additional log topics I should check to identify the root cause?
2

It looks like:

FD_FAIL_P and FD_FAIL_R is set to 60

3

I assume you are wondering why the drone flipped/ the estimator thinks the drone flipped, and not why it subsequently flipped, right?

Did you observe the drone during this maneuver? What makes you think the estimator is wrong?

4

Helllo,

What about other details including battery charge state. Discharge state, remote controller position state, distance of RC from the drone and other states to understand what triggered such events. . Kindly try the same setup on Px4 v1.15.2 stable and Px4 v1.13.x stable and check if such ekf movements remain the same ! !

5

I don’t know why the estimator is wrong. It was hovering for almost a minute and then suddenly estimator detects high degree of roll/pitch and it throws altitude failure, and triggers flight termination.

I tried it with Px4 v1.15.2 and the result is same

.

6

Do I undersand that it has happened twice. Once with PX4 Version: 1.16.2 and once with PX4 Version: 1.15.2

7

Can you provide the log for the Px4 v1.15.2 stable version

8
9

Hello,

After initial review of your log, it seems your CCW Motor # 3 has failed and therefore everything dropped from the sky…Need to inspect Motor # 3 …check for loose props…Good luck !

10

Yes, that happened because there was an erroneous roll/pitch estimation. To compensate for the perceived attitude error, the motors were driven at a high PWM. However, no actual roll or pitch was observed. This ultimately led to flight termination, as the Failure Detector (FD) detected a roll/pitch angle exceeding the configured threshold of 60°. After this, Ekf losses confidence on attitude estimator and triggers flight termination via FD.

I was not able to know why EKF estimates the high degree of roll/ pitch.

Propellers were tight while checking after the crash. Same has been happened multiple times with both PX4 v1.15.2 and 1.16.2.

11

Why would your IMU perceive or produce a roll/pitch angle that did not occur?

When was the last time you calibrated your gyroscopes and acceleromters?

I don’t see the exagerated estimated roll and pitch angles on your second flight review that I see on the first.

Screenshot 2026-07-02 101022
Screenshot 2026-07-02 1010221023×443 41 KB

Screenshot 2026-07-02 101009
Screenshot 2026-07-02 1010091013×441 22 KB

12

I’m not sure. I calibrated all the sensors just before the flight. I believe that if we can figure out why it overestimates the roll and pitch, we can avoid the flight termination. This happens with different flight controllers of the same model as well as different models.

13

Hello,
Motor 3 or props may have failed causing your pitch and roll to give weird readings !!

14

Well its good you can rule out sensor calibrations.

The estimated roll and pitch angles are different between the two flight reviews. However, there is high anglualr accelerations on both flight reviews. If there is too much distance between your Center of Mass and IMU, you could get higher than expected accelerations.

ArudPiolot documentaion about Center of Mass and IMU placement problems.“For example if the autopilot was placed on the nose of a vehicle and the vehicle suddenly leans back (i.e. rotates so that its nose points up) with no offset compensation the vehicle velocity estimate would momentarily show the vehicle is climbing when it’s not.” Sensor Position Offset Compensation — Plane documentation

Well its going to be a process of elimination, and there are people here that know more about PX4 autopilot than me.

Let us know what the fix is.