Help with log analysis for crashed flight : GPS glitch

ArduCopter Copter 3.5
1

Hello,

I need help with log analysis for a crashed flight.The crash resulted in the battery catching fire.
This is for quadcopter that uses Pixhawk. We use a 6S Li Ion battery.
The flight log is available at the following link

Context:
After take-off, at about 16m altitude AGL, the AV went into AltHold mode due to GPS Glitch. (We have set the FAILSAFE response for GPS Glitch as AltHold)
No. of satellites for GPS1 was always higher than that of GPS2 except once.
HDOP for GPS2 was always higher than that of GPS1.
But no change was seen on either when the AV went into AltHold mode.

The pilot was unaware that AV went into AltHold mode, and couldn’t take control.
EKF variance occurred immediately after this.
If u compare the Lat/Long of GPS1, GPS2 and POS, you can see that POS plot switches from GPS1 to GPS2 right before AltHold mode. There is considerable variation in POS plot thereafter.

Capture
Capture.JPG1702Ă—512 68.8 KB

The pilot switched the mode to LAND (Because the GPS Glitch cleared for sometime allowing the pilot to switch to LAND)
However, this made the AV extremely unstable and it flew in an arc to the left, with high speed, hit a building and crashed which led the battery to catch fire.

If u check the vibration logs, you can see that there is incremental increase in VIBE values especially on Z axis.

Capturevibe
Capturevibe.JPG1689Ă—529 70.8 KB

Questions:

  1. Is there something that we are missing while analysing the log that will lead us to the root cause?
  2. What would have caused the POS variance? Both GPSs have been observed to give similar values, and they were verified to be correct. We assume that when the pilot switched to LAND, the AV thought that it was in a different location, and flew to land at the location where it was commanded to land.
  3. Will that be an issue with IMU output or Compass?
  4. Would Li Ion battery cause issues with IMU? We have seen issue with mag calibration (LIS3MDL) before, but we used HMC5883L for the flight under question and this issue was not seen.

Please help!

2

Nothing to do with GPS glitch. Very likely a failed compass, maybe it got loose. Look at the increased vibration before “the glitch” and even the accelerometer logging paused for a moment.
Unfortunately there are no MAG log, but for me it looks like a failed or loose magnetometer.
GPS glitch detected because the failed magnetometer contributed wrong data to the EKF position estimator, and it thought a different position than the gps. So the message is kind of misleading.

image
image.png1594Ă—494 167 KB

1 Like
3

Hi, Sorry for the late reply.
I would like to point out that it is not just the accelerometer that paused logging. We had seen that the break in logging can be seen in other parameters as well. And such breaks are there at various points of the plot.
So, we had assumed that it is a logging error.
A possibility is that the entire autopilot functioning was affected by some event (may be due to vibration) which lead to malfunctioning of the accelerometer.
Is it possible for this to happen?
Also, after the GPS glitch, the POS plot switches momentarily starts tracing GPS2 before deviating altogether.