Mavic Mini out of control - saved by RTH

[Starz]

Yes that is correct. I took off from the ice quickly, and usually let it hover until everything is locked in. Since I was planning to just fly strait up and pan around directly above I wasn’t expecting it to start flying away like it did.

May not be anything inside the trailer. Outside where it launched from was a bit polluted for metal. Steel wheels, steel axle, utv.
May be just luck that it hasn’t happened previously.

 

[BudWalker]

I have flown many flights from my ice fishing shack and never had an issue before. I did purchase a new fish finder that also has GPS lake mapping capability. Could a fish finder GPS interfere with my MM flying far away?

From the video it seems that the launch site was sandwiched between two large metallic vehicles.

 

[miker570]

Ok I understand now. The plow on the front of my ATV is clearly a large metal object. This was the first time I have launched from the ice directly in front of it. I did not see any compass warnings so I did not make that connection. Yes I was lucky it returned home safely.

 

[miker570]

Thanks to all that responded and this forum for the valuable service you provide!

 

[sar104]

I missed this thread at the time, and was pointed to it today by @slup in relation to another event. While the cause of the uncontrolled flight was an incorrectly initialized IMU yaw, it does not look like a magnetically distorted launch site. The compass was completely wrong for the entire flight - most likely due to being completely out of calibration with the magnetic state of the aircraft. The IMU yaw was initialized incorrectly by around 127° - the aircraft started off pointing due south, not north east as indicated by the compass. But the compass did not recover after takeoff. The FC did a reasonable job of tracking the rotation of the aircraft using a combination of the rate gyro inertial solution and its secondary optical rotation calculations, until 86 seconds (1:13 in the video) when it became confused, perhaps by the optical data, and recorded a 100° CW rotation that did not happen. It basically ignored the compass data completely.

The graph below shows that excursion, which was not present in either the video or the pure inertial calculation from integrating the vertical (not aircraft z-axis) gyro vector, both of which show continuous CCW rotation. Anyway - that error, combined with a small non-steady gyro bias error, randomly resulted in the IMU yaw solution becoming correct, even though the compass was still completely wrong. That was how it managed to limp back to the launch point. Pure luck.

 

[new2mavic]

.....tracking the rotation of the aircraft using a combination of the rate gyro inertial solution and its secondary optical rotation calculations.....

Mind explaining what "optical rotation calculations" is ?

 

[sar104]

Mind explaining what "optical rotation calculations" is ?

The Mavic 2 and Mini have some kind of "optical compass" that uses its VPS cameras to detect and measure rotation. It's recorded as VIO_yaw. The graph below shows the values from this flight, compared to the IMU yaw:



Note that it isn't a continuous record - it gets reset to match the IMU yaw if it loses track.

 

[DanMan32]

If launching from snow/ice, it's lucky VPS was able to help resolve the situation.

Although it was determined not to be the primary cause, don't launch without sufficient GPS lock. Without GPS, the only other means of attitude stability is VPS, and with ground being mostly white, that might not be reliable either. So while hovering waiting for a GPS lock, you're potentially subject to wind drift.

 

[sar104]

This event is still bothering me. Having made some more careful calculations of the pure inertial solution combined with the initial and final yaw values as determined by video, there is either a strange coincidence here or something I don't understand.



The pure inertial yaw solution and the VIO yaw solution agree very well, and if they are initialized with the real initial yaw determined from the video then they end up in good agreement with the yaw measured at the end of the video. That's nice, and is a testament to the effectiveness of the DJI optical yaw algorithm and the quality of the rate gyro data.

However, the IMU starts off completely wrong and stays wrong for the first 88 seconds - the compass data are useless and no help in correcting it. But at 88 seconds the IMU yaw discontinuously changes to match the actual yaw that the inertial solution and the VIO data predict - but only if you know the actual starting yaw, which the FC did not since the compass was incorrect. It's a little hard to see but note that the positive compass excursion at a similar time actually occurs just after the IMU yaw reset, so that wasn't the cause. That begs the question - was that IMU yaw reset caused by something else, and just fortuitously made the yaw correct, or is there something else going on here? Without a working compass, how did the FC know which way the aircraft was pointing, and if didn't, and it was just coincidence - which data caused that reset?

@BudWalker, @slup

 

[slup]

This event is still bothering me. Having made some more careful calculations of the pure inertial solution combined with the initial and final yaw values as determined by video, there is either a strange coincidence here or something I don't understand.

View attachment 95548

The pure inertial yaw solution and the VIO yaw solution agree very well, and if they are initialized with the real initial yaw determined from the video then they end up in good agreement with the yaw measured at the end of the video. That's nice, and is a testament to the effectiveness of the DJI optical yaw algorithm and the quality of the rate gyro data.

However, the IMU starts off completely wrong and stays wrong for the first 88 seconds - the compass data are useless and no help in correcting it. But at 88 seconds the IMU yaw discontinuously changes to match the actual yaw that the inertial solution and the VIO data predict - but only if you know the actual starting yaw, which the FC did not since the compass was incorrect. It's a little hard to see but note that the positive compass excursion at a similar time actually occurs just after the IMU yaw reset, so that wasn't the cause. That begs the question - was that IMU yaw reset caused by something else, and just fortuitously made the yaw correct, or is there something else going on here? Without a working compass, how did the FC know which way the aircraft was pointing, and if didn't, and it was just coincidence - which data caused that reset?

@BudWalker, @slup

Just quickly looked into the timing when the RTH were commanded & compered to the Desert event ... & scrolled through the EventLogPlayer in CsvView looking for any command just before & after RTH that said something that could point to some kind of reset ... no luck.

The 2 events don't behave similar in relation to the RTH either ... this case continue spiral away some time after the command but the Desert event settles down just before ...

Suspect this is way over my present knowledge level ...

 

[sar104]

Just quickly looked into the timing when the RTH were commanded & compered to the Desert event ... & scrolled through the EventLogPlayer in CsvView looking for any command just before & after RTH that said something that could point to some kind of reset ... no luck.

The 2 events don't behave similar in relation to the RTH either ... this case continue spiral away some time after the command but the Desert event settles down just before ...

Suspect this is way over my present knowledge level ...

Yes - I should have added that the RTH command definitely had nothing to do with the aircraft recovering stability.