Unknown cause for GPS glitch and fly away

Hello guys

Please lend me an eye and help me figure out what caused this flyaway.

I was flying my drone, a octoquad with pixhawk 2.4.8, with 2 RTK GPS antennas injected through HIL GPS. We lost control of the drone and had a bad flyaway about 3 months ago, and we were only able to retrieve it yesterday.

We’ve been pouring over the data, and we think our RTK GPS worked the way it should (both RTK GPS antennas recorded same position data, and we inject only one of the antenna to the pixhawk), and the pixhawk seems to be at fault. But in the pixhawk flight log, we are not able to figure out what caused the glitch exactly. Therefore I uploaded the log hoping someone can help figure out the issue.

In the pixhawk log, please look at the first 110K lines, especially before the “ERR: GPS-2” came on. It’s a very long because it was stuck in the tree still recording.

Some relevant parameters:
EK2_POSNE_M_NSE is set to 0.1 to trust RTK GPS.
EK2_ALT_M_NSE is set to 8.5 to trust RTK GPS more.
All other EKF noise factors are default.

Many thanks!

We just found this interesting observation. If you look at the NKF1.PE, the movement does not agree with the GPS Longitude. The drone flew away towards the east, most likely because the EKF thought it was going west… But what caused the EKF to be so wrong??

It was not a GPS but a magnetometer failure. If you look at the log, your copter become unstable on the pitch axis right before you switched to RTL, at the same tame Magnetometer innovations on the X axis started to rise. This eventually lead to compass switch then EKF failsafe.

The GPS glitch message was triggered since the compass gave false direction.

The biggest problem is imho, that there is no such thing as 2.4.8 Pixhawk. The ones with this designation are usually cheap clones, with questionable quality sensors and inferior assembly. I would not trust anything valuable on it.

1 Like

Eosbandi
Thank you for that. If you look at the IMU1 and 2 GyrX, you can also see the difference. The other gyro axis don’t have that problem. You are right! What a weird porblem, never seen it before! Yeah, since that flyaway, we have moved to the Pixhawk 2.1