Help...Mavic went haywire and ignored my controller, crashing somewhere out of reach

[timonoj]

Hi guys,

My Mavic Pro crashed last week on the edge of a cliff out of reach for anyone. Everything was going according to plan, until I softly hit forward for less than half second, to brake from a backwards panning I was doing. It didn't stop there, and kept pushing forward on its own. I pressed back several times, but it kept accelerating forward, up to the point it hit some bushes in a completely unreachable area. I had the DJI care refresh, but since I came empty handed, they won't provide another one. DJI also is refusing to give a replacement, even though it clearly malfunctioned (it shows clearly in the logs I'm pulling backwards for several seconds while in the video you can see it lunging full throttle towards a mountain). Is there anything I can do? They're only offering me a discount to buy a new one (even though in my opinion its a clear failure...within the first year warranty period).

I have a video mixing the logged controller input and the video image where it clearly shows I'm pulling backwards, yet it keeps going forward faster and faster. I'm considering to upload it later, if DJI keeps ignoring me.

EDIT: For full disclosure, the GPS wasn't getting a fix, and DJI is telling me "next time try to be careful and fly with GPS on". Sure, but that's no help. This drone is supposed to be able to fly in ATTI mode without ignoring controller input. Sure, you're liable to have some drift and not having such a stable flight (plus, no RTH if you lose it out of sight, etc)...but again, you're still expecting it to obey controller input. Forward is still supposed to be forward, and backwards is still supposed to be backwards.

 

[msinger]

Check out your TXT flight log to see if it explains what happened. You can upload and view it online here. If you'd like other people to review and comment on your flight log, then please post a link back here after you upload it.

 

[pixl]

Hi guys,

My Mavic Pro crashed last week on the edge of a cliff out of reach for anyone. Everything was going according to plan, until I softly hit forward for less than half second, to brake from a backwards panning I was doing. It didn't stop there, and kept pushing forward on its own. I pressed back several times, but it kept accelerating forward, up to the point it hit some bushes in a completely unreachable area. I had the DJI care refresh, but since I came empty handed, they won't provide another one. DJI also is refusing to give a replacement, even though it clearly malfunctioned (it shows clearly in the logs I'm pulling backwards for several seconds while in the video you can see it lunging full throttle towards a mountain). Is there anything I can do? They're only offering me a discount to buy a new one (even though in my opinion its a clear failure...within the first year warranty period).

I have a video mixing the logged controller input and the video image where it clearly shows I'm pulling backwards, yet it keeps going forward faster and faster. I'm considering to upload it later, if DJI keeps ignoring me.

Yup...there are a lot of experienced pilots here that can help you possibly determine what happened, and suggest a course of action. Once you upload the .txt file as directed by Mavic Help, post the link here and the community will be helpful as we can.

 

[timonoj]

Thanks! I checked it with the app. Since I have no GPS fixation, not sure it's very useful. I'll try again to take a look if it shows anything more useful.

Wow, thanks for the fast response. I'm at work now, let me see if I can upload them from the phone.

EDIT: Here's the link. Since GPS didn't have a fix, I'm not sure the information from this viewer is very useful.
DJI Flight Log Viewer - PhantomHelp.com

 

[pixl]

Thanks! I checked it with the app. Since I have no GPS fixation, not sure it's very useful. I'll try again to take a look if it shows anything more useful.

Wow, thanks for the fast response. I'm at work now, let me see if I can upload them from the phone.

You were in ATTI mode? what device were you using to run the app? I'm not sure DJI will be very helpful to you if you weren't in GPS.

 

[timonoj]

Yup, right you are, they are not being very helpful. Yeah, ATTI mode. I was running the DJI Go 4 app, all stock, no changes done. Running in a powerful Galaxy S7 with no lag.

 

[pixl]

Yup, right you are, they are not being very helpful. Yeah, ATTI mode. I was running the DJI Go 4 app, all stock, no changes done. Running in a powerful Galaxy S7 with no lag.

still, do the upload and show the community. We all love to trouble shoot.

 

[timonoj]

Sorry, I posted the logs but by editing a previous post, so maybe it didn't show immediately on your browser. Here it goes:
DJI Flight Log Viewer - PhantomHelp.com

 

[pixl]

Sorry, I posted the logs but by editing a previous post, so maybe it didn't show immediately on your browser. Here it goes:
DJI Flight Log Viewer - PhantomHelp.com

sure...lemme tag two other folks here that like to do forensic flight analysis: @lisadoc and @sar104

 

[pixl]

two messages that show up I find curious: "Braking now! Return sticks to midpoints first." and "Motor Obstructed". There would be a lot of uncorrected drift associated with ATTI braking. Motor obstruction could cause it to fly off in a direction that wasn't intended, and with the Mavic's need to have GPS to know whether it's holding position...maybe that contributed. It all happened in the last couple seconds, and it's really just speculation on my part.

Unlike with the inspire, where you can fly in ATTI by choice and get some practice with the performance, the Mavic is so automated that it just doesnt do well without GPS.

I'd do my best to retrieve it from where it is. Perhaps this video will inspire you:

 

[Robbyg]

Most likely not enough stick input was added to prevent the forward drift or you had wind pushing you into cliff and did not use enough power to overcome it.

The Motor obstruction is most likely when you hit the bushes, the whole set of messages happened in about 2 seconds so I figure it was during impact.

Rob

 

[timonoj]

two messages that show up I find curious: "Braking now! Return sticks to midpoints first." and "Motor Obstructed". There would be a lot of uncorrected drift associated with ATTI braking. Motor obstruction could cause it to fly off in a direction that wasn't intended, and with the Mavic's need to have GPS to know whether it's holding position...maybe that contributed. It all happened in the last couple seconds, and it's really just speculation on my part.

Unlike with the inspire, where you can fly in ATTI by choice and get some practice with the performance, the Mavic is so automated that it just doesnt do well without GPS.

I'd do my best to retrieve it from where it is. Perhaps this video will inspire you:

Hmm interesting. I never actually tried to go ATTI by choice. Might be a good training for the next one...

Most likely not enough stick input was added to prevent the forward drift or you had wind pushing you into cliff and did not use enough power to overcome it.

The Motor obstruction is most likely when you hit the bushes, the whole set of messages happened in about 2 seconds so I figure it was during impact.

Rob

It wasn't very windy in that location (it was a bit windy though). Levers were full backwards for several seconds (two short pushes full backwards, then after I realize something's up, full continuous backwards push for around 4-5 seconds, until it crashed...I released the lever after I heard it hit the bushes), but it made no attempt to fight any wind. It was just moving forwards. In the last seconds I was directly looking at the drone, and in the last push I could see it taking a full forward attack position (some inclination around 30º forward). It wasn't trying to go backwards, it was decidedly going forward, like on purpose. However that´s just my word, the drone´s angle changes are not noticeable on video...They never were, thanks to the gimbal. And I don't think the log files logged that level of detail.

 

[sar104]

It's tricky to diagnose without the GPS data, but that would require the DAT file from the aircraft. Looking at the relationship between elevator/pitch and aileron/roll is the best that can be done to see if the aircraft was responding to the controls.



It bears some resemblence to your description but there are some notable discrepancies. Considering pitch, the aircraft had positive pitch (i.e. was likely flying backwards) in response to back elevator from 53 s to around 59 s. At around 61 s you applied full forward elevator for 2.5 seconds (not "soft" elevator for less than 0.5 s) and it responded appropriately with significant negative pitch (accelerating it forwards). You then centered the elevator for just over 3 s and the pitch stabilized around zero, after which you applied full forward elevator again for another 2.5 s, to which the aircraft again responded correctly with more negative pitch. Those two periods of full forward elevator likely imparted significant forward velocity to the aircraft. Nearly 2 s after that you applied sporadic full back elevator for just 1 s and then centered the elevator again for another second or so, followed by another second of full back elevator, with the aircraft applying appropriate, though brief, positive pitch on both occasions. It's unlikely that would have killed the forward motion. Then it appears to have crashed at just past the 79 s mark.

Note also that you were putting rather random-looking inputs on the aileron throughout the last 20 s, with appropriate response from the aircraft to those inputs. I cannot tell if those had any relevance to the crash though, since I don't know the actual velocity vectors.

My initial comment, based on those limited data, is that I don't see any attitude responses that are not consistent with the stick inputs, and you seem to have applied far more forward elevator than you remember that, in ATTI mode, led to the forward motion leading to the crash. Can we see the video? That might make things clearer.

 

[sar104]

two messages that show up I find curious: "Braking now! Return sticks to midpoints first." and "Motor Obstructed". There would be a lot of uncorrected drift associated with ATTI braking. Motor obstruction could cause it to fly off in a direction that wasn't intended, and with the Mavic's need to have GPS to know whether it's holding position...maybe that contributed. It all happened in the last couple seconds, and it's really just speculation on my part.

Unlike with the inspire, where you can fly in ATTI by choice and get some practice with the performance, the Mavic is so automated that it just doesnt do well without GPS.

I'd do my best to retrieve it from where it is. Perhaps this video will inspire you:

Those messages appear to be just before and after impact, respectively. I don't think that they are relevant.

 

[timonoj]

Wow...I'm impressed with the level of detail you pulled. Thanks. Screw it, since DJI didn't reply in the last 24h I'll upload the video. Give me a few minutes.

Here it is:

The levers reaction is a bit out of sync for the first 30 seconds (that's the app recording at a different framerate), but in the important section when it starts misbehaving, it's on sync.

Please let me know your thoughts.

 

[timonoj]

Also, here's the unedited video the app recorded:
Private Seafile

It's tricky to diagnose without the GPS data, but that would require the DAT file from the aircraft. Looking at the relationship between elevator/pitch and aileron/roll is the best that can be done to see if the aircraft was responding to the controls.

View attachment 25445

It bears some resemblence to your description but there are some notable discrepancies. Considering pitch, the aircraft had positive pitch (i.e. was likely flying backwards) in response to back elevator from 53 s to around 59 s. At around 61 s you applied full forward elevator for 2.5 seconds (not "soft" elevator for less than 0.5 s) and it responded appropriately with significant negative pitch (accelerating it forwards). You then centered the elevator for just over 3 s and the pitch stabilized around zero, after which you applied full forward elevator again for another 2.5 s, to which the aircraft again responded correctly with more negative pitch. Those two periods of full forward elevator likely imparted significant forward velocity to the aircraft. Nearly 2 s after that you applied sporadic full back elevator for just 1 s and then centered the elevator again for another second or so, followed by another second of full back elevator, with the aircraft applying appropriate, though brief, positive pitch on both occasions. It's unlikely that would have killed the forward motion. Then it appears to have crashed at just past the 79 s mark.

Note also that you were putting rather random-looking inputs on the aileron throughout the last 20 s, with appropriate response from the aircraft to those inputs. I cannot tell if those had any relevance to the crash though, since I don't know the actual velocity vectors.

My initial comment, based on those limited data, is that I don't see any attitude responses that are not consistent with the stick inputs, and you seem to have applied far more forward elevator than you remember that, in ATTI mode, led to the forward motion leading to the crash. Can we see the video? That might make things clearer.

I think the problem starts after the 61s. When the drone pitch reports to zero, it's actually moving forward at a rather decent speed. Any correction pushes afterwards seem to be ignored.

 

[Rnl]

Go get it rope and a hook it's close to the edge

 

[sar104]

Wow...I'm impressed with the level of detail you pulled. Thanks. Screw it, since DJI didn't reply in the last 24h I'll upload the video. Give me a few minutes.

Here it is:

The levers reaction is a bit out of sync for the first 30 seconds (that's the app recording at a different framerate), but in the important section when it starts misbehaving, it's on sync.

Please let me know your thoughts.

The video seems consistent with the log and explains the aileron inputs, most of which appear to be intentional although the right aileron in the last few seconds was definitely the opposite of what was needed. The effect of the two periods of full forward elevator are clear. What was needed was full backward elevator with no aileron as it headed for the rocks, but it appears to me that it was too little too late.

Unfortunately I cannot see DJI coming to any other conclusion if they review the data.

The location is a bit challenging but it looks as if the aircraft is likely not very damaged. No chance of rappelling down from above to retrieve it?

 

[timonoj]

The video seems consistent with the log and explains the aileron inputs, most of which appear to be intentional although the right aileron in the last few seconds was definitely the opposite of what was needed. The effect of the two periods of full forward elevator are clear. What was needed was full backward elevator with no aileron as it headed for the rocks, but it appears to me that it was too little too late.

Unfortunately I cannot see DJI coming to any other conclusion if they review the data.

The location is a bit challenging but it looks as if the aircraft is likely not very damaged. No chance of rappelling down from above to retrieve it?

Thanks for the detailed reply!
Sorry if I'm a bit thick, but I don't quite follow. Let me see...There's a full forward very short altogether with full left panning for a while, so I initiate a panning move (1m02 in the app). I follow this by short rotations to the right so it keeps us in frame as it moves. Then at 1m10s I do a full forward for about 2 seconds, intending to stop the panning. From there, it starts moving forward, catching speed. I do a couple of tentative braking (full lever, but very short periods). After I notice there's no effect and the drone keeps catching speed, I go full backwards at 1m16, and I keep it that way until it crashes at 1m21. It had nearly 5 seconds to brake at full power backwards, and yet it kept catching speed.

Whenever I did braking maneuvers before, from full speed, it usually took a full second in getting to a complete halt. During that time, the distance covered is still long, which is a different thing, but it usually brakes to a halt in about a second. It had nearly 5 seconds to brake, and when it crashed it was still accelerating. The final aileron movement right before the crash was a final useless desperate attempt to get it to change course...but too little too late.
Can you explain how I did wrong? I mean, if I knew something was up earlier, I might have pushed back the lever 4 seconds earlier...but I still feel the drone ignored the input after 1m10s, and specially after I start pushing backwards full time.

 

[fiberdude]

In the atti mode, anything can happen — drifting at a large speed over dozen meters is common.