Crash: Pixhawk 2.1 (Cube) with v1.7.4 Beta firmware

[r0gelion]

Flying an F450 with Pixhawk 2.1 (Cube) and Here GPS, loaded v.1.7.4 Beta (e63f9d9).

The pilot reported the vehicle ended a mission and it was hovering in position hold and suddenly it flew to a side in a straight line descending and it crashed on a roof.

Here is the log:
https://review.px4.io/plot_app?log=34c46166-8fee-4cb8-b7de-38517de4122b

Thank you!

[RomanBapst]

@r0gelion The log ends when the pilot was maneuvering the drone in position control mode. Until then I cannot see any strange behavior, if you actually use the 3D view you can see that the vehicle followed the pilot's commands very well.
Is it possible for the pilot to have a look at the 3D view and maybe estimate the time between when the log stops and when the vehicle started to drift away?

[RomanBapst]

@r0gelion Do you happen to have a mavlink log of this flight?

[r0gelion]

@RomanBapst this is the log:

2018-04-05 12-35-45.zip

[RomanBapst]

@dagar @LorenzMeier This is a csv file which contains the relevant data from this flight. We analyzed it with @bkueng and found that the mavlink connection must have dropped about 4 seconds after the ulog stopped. This somehow makes sense because the logger writes to the SD card in certain intervals.

Line 202: this attitude sample roughly corresponds to the last attitude sample in the ulog.
Line 314: this is the last message that was received from the autopilot. After that there are only RADIO and RADIO_STATUS messages which are generated by the radio module.
This only tells us that the system was still powered.

While in the incident reported by @JohnSnowball the mavlink connecting did not seem to be interrupted after the log stopped in this case it clearly did.

telem_log.csv.tar.gz

[dagar]

Could someone go over the timeline? How much longer do we think the vehicle was in the air after the log stopped?

We only have about 2 seconds of RADIO_STATUS data from the vehicle after the last regular mavlink message.

[dagar]

What was the state of the vehicle after the crash?

[LorenzMeier]

Closing: #9271 (comment)