DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

Mavic 2 Zoom flyaway (for real!) - help explain?

bzmot

Well-Known Member
Joined
May 27, 2018
Messages
167
Reactions
96
Age
44
Hey guys, for starters, I want to say that I did not believe in flyaways, mostly. I, like many in this community, leaned towards the view that the term was overused by people who crashed their drone due to pilot error. However, after what happened to me last Saturday I stand corrected.

I had a Mavic 2 Zoom which I had been flying since it was released in 2018 so a pretty long time. It was not my first drone either. I always read manuals and never had any safety problems. I used it recreationally only but I logged many hours and flew confidently.

I should mention here that for some weeks prior to the incident I used the Insta360 OneR Aerial Kit on the drone. It's a special mount that lets you fit the Insta360 OneR camera on the Mavic to achieve the so-called "invisible drone" effect. Now, the manufacturer clearly warns buyers that the kit will leave your Mavic 2 without a GPS signal (strangely, GPS on the Mavic Pro / Platinum would be unaffected), so you have a to fly ATTI. No problem for me, I've done it for a while and it's not difficult for me.

The first few tests were encouraging. The drone did lose the GPS fix (strangely though it was not the bulky mount that cut off the GPS signal; the bird lost GPS only after I turned the camera on so interference must have been electronic rather than physical). However, if I had a good fix prior to turning on the camera, the bird stayed in P mode. Technicaly it was showing "ATTI" (or sometimes also "Visual mode", I don't know why) but it held its position against wind (so hover ok) and it did show me airspeed. It didn't show me positioning and no RTH of course. So I guess it was half-GPS flying. Still, not bad.

One very strange thing I noticed though: once paired, the camera can be controlled from an Apple Watch via an Insta360 app. You can start recording at a distance for example. So one day I let my drone hover a few meters above and pressed record on the watch. It did start recording but at the same time... the drone started landing (the app on the controller said: "Landing initiated"). What the fudge?? This did not happen consistently, maybe in 50% of the cases where I used the Apple Watch, but it kept happening and I thought that was worrying.

However, other than that the system worked well, the footage was interesting, if not without its flaws, and I flew more and more confidently.

Fast forward to the incident day: Last Saturday, sunny weathter, a bit of wind, maybe 10-15 kph, I was on the shore of a local lake and decided to fly the drone. I launched it with the camera already recording. The first few seconds were ok. Then I suddenly lost video feed from the drone’s camera and then the drone started turning around and speeding away from me pretty fast. It didn’t react to any stick inputs, anything I did on the joysticks. Nothing. Now, I was a bit panicked but certainly I didn't order it to land. I tried turning it around and come back to me but to no avail. No video feed AND no GPS, well I was blind there. With increasing frustration and anguish I was watching it get smaller and smaller. Then it disappeared against the backdrop of some trees at the other shore of the lake. Friends who were with me and have keener eyesight than I do told me they saw it spin around wildly for a few seconds and then take a dive in the lake.

We did a small search party in the bushes and the treeline on the other shore (the lake is not that big) but to no avail.

Now, I am uploading my controller DAT and txt files. Maybe one of you can read something into this. Seeing as there was no GPS signal, though, I am not hopeful. I for one am completely stumped.
The drone was no longer under warranty so understandably DJI said there's nothing they can do. Insta360, however, told me if it was their camera's fault, they will provide a "reasonable solution", but they need to hear what DJI will say first. Now, I asked DJI to investigate further but just in case they ignore me, I am posting this here in the hope that someone can shed some light on this unfortunate situation (yep, close to $2k down the drain....)
 

Attachments

  • DJIFlightRecord_2020-05-30_[18-33-36].txt
    254.6 KB · Views: 37
  • 20-05-30-08-13-15_FLY039.DAT
    871.5 KB · Views: 8
  • 20-05-30-08-15-08_FLY040.DAT
    2 MB · Views: 3
  • 20-05-30-09-21-32_FLY041.DAT
    814.5 KB · Views: 2
  • 20-05-30-09-23-26_FLY042.DAT
    955.6 KB · Views: 2
  • 20-05-30-09-25-38_FLY043.DAT
    131.3 KB · Views: 2
  • 20-05-30-09-26-05_FLY044.DAT
    4.9 MB · Views: 11
Well I'm confused. This was your flight:

grab37.png

It appears to have landed in the lake. What confuses me is that the aircraft responded precisely to all your stick commands, including the descent into the lake:

Pitch.png

Roll.png

Climb.png

It appears to me that you lost orientation awareness and, as a result, flew it into the lake. And it was only spinning at the end because you were applying full rudder:

Yaw.png
 
Wow.... just wow... I just can't believe this. Thanks for the detailed analysis, as always.

1. Is it not possible that my commands were somehow warped or distorted, amplified perhaps, because of the Insta360 camera interference? Like in the case when my Apple Watch triggered the landing procedure of the drone.
2. Is there any hint in there as to why I would have lost video feed?
 
Sounds like interference between your add on camera/watch/drone. As soon as you hit record and it makes the drone descend I'd have taken it off and not used it again.
Not helpful I'm afraid, but hope you get sorted- not sure who'll take the blame but doubt DJI would
 
Bonus questions:
1. I am 100% positive that at first the drone sped away from me (100-200 m perhaps) and only then did I start to apply any stick commands. Is it possible that some of those inputs that you see as logged were in fact fake (yes, I realize probably every single pilot in my place would ask this question...)
2. How is it that GPS data was recorded but I didn't see in the app?
 
Wow.... just wow... I just can't believe this. Thanks for the detailed analysis, as always.

1. Is it not possible that my commands were somehow warped or distorted, amplified perhaps, because of the Insta360 camera interference? Like in the case when my Apple Watch triggered the landing procedure of the drone.
2. Is there any hint in there as to why I would have lost video feed?

No - the stick inputs look real to me. The only times I've seen spurious stick inputs it's been due to faulty or uncalibrated controllers, and the signature is very obvious.

I don't know about the loss of video feed - the telemetry link was solid until splashdown.
 
Bonus questions:
1. I am 100% positive that at first the drone sped away from me (100-200 m perhaps) and only then did I start to apply any stick commands. Is it possible that some of those inputs that you see as logged were in fact fake (yes, I realize probably every single pilot in my place would ask this question...)
2. How is it that GPS data was recorded but I didn't see in the app?

Since it was in ATTI mode it can drift, but there was no significant motion before you applied elevator input:

vel.png


The app doesn't display location data until the navHealth value reaches 4 - you never got there. The raw GNSS data are in the DAT file though.

Status.png
 
No - the stick inputs look real to me. The only times I've seen spurious stick inputs it's been due to faulty or uncalibrated controllers, and the signature is very obvious.

I don't know about the loss of video feed - the telemetry link was solid until splashdown.

Ugh....

It didn't drift because it hovered in a stable way right in front of me before I applied the first thrust. Is there any chance that the aircraft, with a weak GPS signal, somehow started flying on its own because its collision sensors were blinded by the sun. It was more or less exactly in the direction of the sun that it flew away.

It's still very hard for me to believe I may have caused this....
 
Ugh....

It didn't drift because it hovered in a stable way right in front of me before I applied the first thrust. Is there any chance that the aircraft, with a weak GPS signal, somehow started flying on its own because its collision sensors were blinded by the sun. It was more or less exactly in the direction of the sun that it flew away.

It's still very hard for me to believe I may have caused this....

GPS was never in use:

gnss.png

VPS was used for positioning for the first 57.5 seconds of the flight:

vps.png

I'm sorry to be the bearer of bad news, but to me it looks like you just panicked when it didn't stop when you released the elevator at 54 seconds.
 
>>> GPS was never in use:
>>> VPS was used for positioning for the first 57.5 seconds of the flight:

It's so weird that you would say that because like I said, in my previous flights with the aerial kit on the drone there was perfect hover despite the wind, and I flew at altitudes were VPS wouldn't work anymore (50-60 m). So the bird must have somehow used GPS...
In fact, I could post those logs here for comparison, what do you think?

>>> I'm sorry to be the bearer of bad news,

Well it's like those military guys who go door to door notifying next of kin. Nobody wants that job but... ;)

>>> but to me it looks like you just panicked when it didn't stop when you released the elevator at 54 seconds.

Yeah well I guess it was the perfect storm, coupled with video feed loss, it was very difficult to maneuver.
Anyway, thanks for your help and your time. You should really get paid for this...

[/QUOTE]
 
>>> GPS was never in use:
>>> VPS was used for positioning for the first 57.5 seconds of the flight:

It's so weird that you would say that because like I said, in my previous flights with the aerial kit on the drone there was perfect hover despite the wind, and I flew at altitudes were VPS wouldn't work anymore (50-60 m). So the bird must have somehow used GPS...
In fact, I could post those logs here for comparison, what do you think?

>>> I'm sorry to be the bearer of bad news,

Well it's like those military guys who go door to door notifying next of kin. Nobody wants that job but... ;)

>>> but to me it looks like you just panicked when it didn't stop when you released the elevator at 54 seconds.

Yeah well I guess it was the perfect storm, coupled with video feed loss, it was very difficult to maneuver.
Anyway, thanks for your help and your time. You should really get paid for this...

If you want to post one of the previous logs - it will need to be a DAT file - then I'd be happy to take a look.
 
  • Like
Reactions: NedKelly
If you do find yourself in ATTI mode with no camera view, first, climb to avoid any nearby obstacles, second, rotate the aircraft so that the orientation arrow is pointing towards you (that requires knowing the direction of the aircraft from you since there is no home point reference), then apply forwards elevator to bring it back close enough to descend and land.
 
If you do find yourself in ATTI mode with no camera view, first, climb to avoid any nearby obstacles, second, rotate the aircraft so that the orientation arrow is pointing towards you (that requires knowing the direction of the aircraft from you since there is no home point reference), then apply forwards elevator to bring it back close enough to descend and land.

Thanks! I'll try to post the dats later. I will need to see what the DJI reply to me first. I know you're certain I caused this but I'm still doubtful. Anyway, thanks for all the help!
 
Thanks! I'll try to post the dats later. I will need to see what the DJI reply to me first. I know you're certain I caused this but I'm still doubtful. Anyway, thanks for all the help!

Certainty is always hard to achieve, but what I would say here is that all the data are self-consistent and mostly consistent with your recollection of the event, so it seems like a pretty solid conclusion. I'm just not seeing any counter-evidence at all.

I can tell you right now that DJI will look at the txt log and say that it was in VPS, then went to ATTI, then you flew it into the water. Actually they won't know it was water unless they consult the DAT file, which they may not bother doing.
 
One more thing I am curious about and I forgot to ask... were there moments in the log where the drone is moving - be it on the wind or otherwise - but there are no stick inputs?
 
One more thing I am curious about and I forgot to ask... were there moments in the log where the drone is moving - be it on the wind or otherwise - but there are no stick inputs?

Of course - that's what happens in ATTI mode. It makes no attempt to stop. See the first graph in post #7.
 
That WiFi or Bluetooth signal from the camera was in such close proximity to the bird it jammed all other signals. Me telling what you might end up with at distance.
 
Lycus Tech Mavic Air 3 Case

DJI Drone Deals

New Threads

Forum statistics

Threads
130,984
Messages
1,558,558
Members
159,973
Latest member
rarmstrong2580