Mavic went ROGUE, and is gone

[Thorster]

I had the same thing happen while filming over water in Norway. I had hours upon hours on the drone with no issues. Suddenly, on this particular flight the Air decided to do its own thing. I was able to find the drone, and I did have DJI Refresh, but when I sent the drone to them with the telemetry I pointed out that I had not crashed the drone, as was evident from the video footage that I had, as well as the controller input vs. actual travel. The nail in the coffin in my argument was that upon hitting the ground the drone was flying at almost 85 miles an hour (as per their own logging/telemetry)... which is well above any speed I should ever be able to fly the Air, as it is speed limited.

So, to my surprise, they sent me a new one and it didn't reflect on my DJI Refresh account. It was clearly a hardware/software failure of some sort.

So, don't give up hope and good luck. The replacement unit has been working fine ever since.

 

[JDawg]

Who do you purchase the runaway insurance from?

State Farm.

 

[JDawg]

@Thorster, Unfortunately this one is pilot error, sounds like yours was not.

 

[EagleFlyFree]

@ac30 my Mavic Air did the same last year.
I was on one of the mountains around Vancouver.
I started it, GPS strong, every calibration ok, no metal around (maybe some cougars but I didn't see any ) .
Just taking off it went up a bit, then straight to its own left side , fast like a bullet.
It crashed on the rocks luckily close by.
Had only to change the wings, but if it had gone straight instead of by its side , bye bye camera.
I think it was related to that firmware version, 2 weeks later a new firmware came out indeed.
Sorry for your lost was it a Air?

 

[John Suhr]

You can try to get it under warranty if you want but most likely will be denied which is why I and most other users on here have insurance. It's times when we least expect it that the worst happens and that's when it is nice to have insurance for those times, especially when it's that cheap. I'm sad to say but it seems like you will be out $700 or however much you paid for the Mavic. Next time, get insurance. It's a cheap price to pay for something that you don't have 100% control of. Trust me, I crashed my Phantom 3 advanced and had to pay for repairs and ended up losing money. With my Mavic Air, I crashed it and was able to get a brand new drone thanks to DJI Care Refresh.

I thought it only works if you return the drone (which he can’t do) or are you talking about a State Farm insurance? They will cover lost?

 

[Radovan]

So i bought a mavic air back in August, for a trip to Europe in September. Did some practice flights, no problems, got comfortable. Have been doing some local filming since i returned.
Flew over some big cities in Europe and over water, in wind between mountains, through forests, again no problems. I was cautious of buildings and obstacles, wind and air moisture, wifi range/obstruction etc.

Today i tried to film a quick shot, Mavic went up in the air for about 10-15 seconds, above my head while i adjusted light/aperture settings ,
It suddenly veered 150m out over the water, did a sharp U-turn, banked hard right, all within 10 seconds
and i believe it hit the side of a bridge before falling most likely into water.

I didnt find it.

Reviewing the flight record and video feed, you can see i'm not touching the joysticks after initial take off... for 10 seconds until the very last second in an attempt to avoid the crash.
For a split second before impact it did have interference warning and high wind warning, BUT:
-It was already off-track before the warnings appeared
-The wind would have been blowing from the west off the ocean, and my Mavic Air veered westward (into the wind). It was also travelling 40km/h, wind that day was recorded at mid-20, i have recordings of trees in the area and they're barely blowing. Altitude was only 5-6 meters
-It banked HARD right suddenly, before losing transmission as if someone hit it out of the sky

I've had my drone up over the English Channel, Swiss Alps, Mediterranean Sea, over Vancouver's Mountains, and it had no problem staying level against wind gusts.
I've certainly never had it behave so erratically and go so far, so fast in such a short period of time,

My theory is that, a motor may have malfunctioned on one side of the drone. I briefly saw a "motor warning" of some kind once last month, but didn't have enough time to read the warning before it was off the screen, and the drone acted normal again. For the past 2-3 flights i've noticed it seemed to be floating sideways on its own, coming close to trees, and not being able to stay in one spot and i have to counter-hold the joystick slightly to manually stabilize.
Is it possible a motor died and the drone was unable to stabilize before crashing?
What are my warranty/replacement options?
Any other ideas what happened? will attempt to upload

My air seemed like yours. He turned slightly and shifted to the right, although I did not touch the controller. First I thought the wind was pushing him, but I quickly realized that the problem was in the controller. The dust probably got into the controller and caused unwanted contact. I solved the problem by balancing the controller.

 

[louderwo]

Do you have DJI care refresh or drone insurance?

Dji care won’t do anything unless you have the crashed unit to return to them. I know from experience. I had the flight records showing where it crashed and a video of the crash showing it was in the mountains above 13000 feet and can’t get to it. They will Do nothing. GPS location showed where it went down. Dji care is only good if you drop it and step on it!!

 

[Smithy12]

I would be intrigued to know a bit more about the interaction of compass and GPS lock in this case ... anyone know?
I’m an avionics engineer by trade and I cannot understand why an ‘incorrect’ compass value (caused by magnetic interference at the starting point) should be able to cause an out of control event when the GPS locks?
I can understand that the craft will be able to ‘see’ an error between magnetic heading and ground track (gps based) when it is moving but none of this should be relevant if stocks are centered and the craft is hovering?
I would have thought that then the compass heading of the craft is irrelevant and it’s the IMUs keeping it level and stable with GPS holding its position against drift?

Is the compass a 3D compass providing aiding to the IMUs maybe?

Can anyone confirm what additional information from the compass is needed in this case and how it can cause a flyaway condition if the data is incorrect?

Very intrigued!!!!

 

[Cameron Woods]

So i bought a mavic air back in August, for a trip to Europe in September. Did some practice flights, no problems, got comfortable. Have been doing some local filming since i returned.
Flew over some big cities in Europe and over water, in wind between mountains, through forests, again no problems. I was cautious of buildings and obstacles, wind and air moisture, wifi range/obstruction etc.

Today i tried to film a quick shot, Mavic went up in the air for about 10-15 seconds, above my head while i adjusted light/aperture settings ,
It suddenly veered 150m out over the water, did a sharp U-turn, banked hard right, all within 10 seconds
and i believe it hit the side of a bridge before falling most likely into water.

I didnt find it.

Reviewing the flight record and video feed, you can see i'm not touching the joysticks after initial take off... for 10 seconds until the very last second in an attempt to avoid the crash.
For a split second before impact it did have interference warning and high wind warning, BUT:
-It was already off-track before the warnings appeared
-The wind would have been blowing from the west off the ocean, and my Mavic Air veered westward (into the wind). It was also travelling 40km/h, wind that day was recorded at mid-20, i have recordings of trees in the area and they're barely blowing. Altitude was only 5-6 meters
-It banked HARD right suddenly, before losing transmission as if someone hit it out of the sky

I've had my drone up over the English Channel, Swiss Alps, Mediterranean Sea, over Vancouver's Mountains, and it had no problem staying level against wind gusts.
I've certainly never had it behave so erratically and go so far, so fast in such a short period of time,

My theory is that, a motor may have malfunctioned on one side of the drone. I briefly saw a "motor warning" of some kind once last month, but didn't have enough time to read the warning before it was off the screen, and the drone acted normal again. For the past 2-3 flights i've noticed it seemed to be floating sideways on its own, coming close to trees, and not being able to stay in one spot and i have to counter-hold the joystick slightly to manually stabilize.
Is it possible a motor died and the drone was unable to stabilize before crashing?
What are my warranty/replacement options?
Any other ideas what happened? will attempt to upload flight record

I am sorry you lost your drone. I don't have a solution for your problem but I wanted to add that this issue happened several times to me with my Mavic Air. The difference is I got mine back each time because when it happened I was flying high enough that I did not crash into anything and after a few seconds of not being able to control the drone I was able to use return to home to get it back. Attempts to control the drone did not work, it was like being in sport mode with the gimbal unstable.

The last time this happened to me I had a registered speed that the drone took off at 100 mph while I was on a trip to Hawaii! This was for a very short time then it stopped. I was over the ocean and used return to home to get it back.

DJI did offer a return but I kept the drone anyway thinking the same thing would just happen again. I did all the calibrations with the IMU, gimbal, remote and compass to try to fix this. I eventually sold it a bought a Mavic 2.

 

[sar104]

I would be intrigued to know a bit more about the interaction of compass and GPS lock in this case ... anyone know?
I’m an avionics engineer by trade and I cannot understand why an ‘incorrect’ compass value (caused by magnetic interference at the starting point) should be able to cause an out of control event when the GPS locks?
I can understand that the craft will be able to ‘see’ an error between magnetic heading and ground track (gps based) when it is moving but none of this should be relevant if stocks are centered and the craft is hovering?
I would have thought that then the compass heading of the craft is irrelevant and it’s the IMUs keeping it level and stable with GPS holding its position against drift?

Is the compass a 3D compass providing aiding to the IMUs maybe?

Can anyone confirm what additional information from the compass is needed in this case and how it can cause a flyaway condition if the data is incorrect?

Very intrigued!!!!

This is a common question.

Position and heading are initially determined by GPS, compass and barometer but, after that, are both primarily updated from the IMU rate gyros and accelerometers, with the GPS, compass and barometer only being used to correct for drift in the time-integrated rate sensor data via a technique called sensor fusion. That is how the aircraft is able to hover with sub-meter accuracy even though GPS and barometer data are not that accurate.

When you power up the aircraft the IMU yaw value is initialized by the compass heading. If the compass heading is wrong then the initial IMU yaw value is wrong, which is a problem because the primary computation of yaw is done by adding the time-integrated rate gyro z-axis data to the initial IMU yaw, with the compass data only used to correct for sensor drift.

In a hover it may be stable since there is no actual movement but, after takeoff, when the FC attempts to move the aircraft in response to stick inputs, it will move in the wrong direction as indicated by the GPS position and velocity data. That triggers heading errors. Secondly, when the aircraft moves away from the magnetic interference and starts to indicate the correct heading, it will now disagree with the IMU yaw too, leading to compass errors.

 

[Lonjsnyder]

I would be intrigued to know a bit more about the interaction of compass and GPS lock in this case ... anyone know?
I’m an avionics engineer by trade and I cannot understand why an ‘incorrect’ compass value (caused by magnetic interference at the starting point) should be able to cause an out of control event when the GPS locks?
I can understand that the craft will be able to ‘see’ an error between magnetic heading and ground track (gps based) when it is moving but none of this should be relevant if stocks are centered and the craft is hovering?
I would have thought that then the compass heading of the craft is irrelevant and it’s the IMUs keeping it level and stable with GPS holding its position against drift?

Is the compass a 3D compass providing aiding to the IMUs maybe?

Can anyone confirm what additional information from the compass is needed in this case and how it can cause a flyaway condition if the data is incorrect?

Very intrigued!!!!

Here's my very basic explanation:
1. GPS tells the aircraft its position on earth (doesn't control altitude, its barometer does that)
2. Compass tells the aircraft which way is North. They work together to keep the aircraft hovering in one place.
3. See attached diagram; In this example, the aircraft's has initialized thinking "North" is really "East". The aircraft starts in a hover at location #1 and is blown North by the wind to location#2. The aircraft then wants to return to its original position by traveling South. But it thinks South is really West and it moves to position#3. It will continue to spiral (toilet bowl) out of control.

 

[ac30]

not sure how any of that would make the drone decide to take off without any input.
Improper RTH maybe, Improper position on the map, ATTI mode should be the safety fall back... but not actually flying away

 

[Meta4]

not sure how any of that would make the drone decide to take off without any input.
Improper RTH maybe, Improper position on the map, ATTI mode should be the safety fall back... but not actually flying away

The short version is that with a yaw error, any time it moves, it doesn't go where it thinks it's going to go to.
GPS tells it where it actually is and it tries again and again, getting further from where it's trying to be.

 

[ac30]

So why does it try to go anywhere unless I hit the RTH button as a Plan B?
I'm perfectly capable of flying with the joysticks. Again there was certainly no wind at the location... I was hovering for only a few seconds among the trees. The drone can certainly hover properly without a gps/compass input. And the first few moments panning northward indicated no issues

The pilot should be in control and the AI features should be assisting, not overriding, so it's very difficult for me to understand that it would just decide to fly away. It seems like as soon as it picked up a GPS signal it just decides that the primary source of control isn't the pilot anymore, fitting the description of "rogue".

Again, I can see how the RTH location might be off, but it updated correctly upon take off right before the flyaway.

Very frustrating that "such smart tech" could be so... Inferior to human control.

 

[laurens23]

@Lonjsnyder nice ELI5
@sar104 nice ELI25

@Smithy12 one thing to clarify about why it flies fine without the GPS.

The uncontrolled flyaway is the automatic position correction of the Flight Computer (not stick inputs).
The automatic position correction uses GPS input and sends heading requests to the IMU.

For example GPS detects drifting East, FC sends command to IMU to compensate West. The IMU knows West based on the compass reading at the startup. If at the start the compass set a wrong heading in the IMU, then the IMU will angle the drone in the wrong direction, so instead of compensating and staying in position it will now fly the wrong way without user input. This quickly escalates and at that point the drone will detect something is wrong and give some compass errors. But the drone does not switch out of GPS mode (A design flaw in my opinion). This usually results in the GPS screaming for more and more automatic corrections and the IMU sending the drone faster and faster in the wrong direction until it crashes.

@ac30 Even very very very small position drift (i.e. basically no wind) it can quickly escalate if the IMU is sending the drone in the wrong direction. RTH also uses GPS. So the GPS tells the drone home is North, but the IMU sends it in another direction.

Without GPS the drone does not knows its position, Therefore there is no attempt to do any automatic position corrections. The IMU directly translated the stick inputs to changes in angles. A stick command "Yaw left 10deg/sec for 1 sec" gives the same result regardless where the IMU thinks is North. So, without GPS the drone responds and flies as expected.

TLDR: GPS works with heading, IMU works with angles. The compass is used to give the IMU a heading to work with. A wrong compass means wrong heading in IMU. A wrong heading in the IMU in GPS mode means the automatic position corrections are in the wrong direction, resulting in uncontrolled flight. No GPS means no automatic position corrections, so no heading information, only angles, Therefore the drone drone flies fine in ATTI (without GPS).

 

[Meta4]

The drone can certainly hover properly without a gps/compass input.

Are you sure of that?
The compass is very, very important and the drone can't do much without it.

The pilot should be in control and the AI features should be assisting, not overriding, so it's very difficult for me to understand that it would just decide to fly away. It seems like as soon as it picked up a GPS signal it just decides that the primary source of control isn't the pilot anymore, fitting the description of "rogue".

Post #33 explained quite well that your compass was screwed up which caused further problems for the flight controller.
It would be a good idea to read that again.

You did not have AI features over-riding anything.
You had a drone with the most basic navigation feature compromised such that it was not capable of flying straight or even hovering in place.

Again, I can see how the RTH location might be off, but it updated correctly upon take off right before the flyaway.

Correction ... You launched without waiting for GPS and the drone had no GPS at all for the first 47.9 seconds of the flight.
The home point was not recorded until 52.8 seconds into the flight.

Very frustrating that "such smart tech" could be so... Inferior to human control.

This incident had nothing to do with smart tech being inferior to anything.
The compass was screwed up so nothing could control it.

 

[sar104]

So why does it try to go anywhere unless I hit the RTH button as a Plan B?
I'm perfectly capable of flying with the joysticks. Again there was certainly no wind at the location... I was hovering for only a few seconds among the trees. The drone can certainly hover properly without a gps/compass input. And the first few moments panning northward indicated no issues

What you are doing when flying one of these is not even remotely "flying with the joysticks". All you are doing is inputing requests to move up/down, left/right, CW/CCW. All the actual flight stability is provided by the FC.

The FC can hover without compass or GPS, but it will drift in position and yaw because the IMU sensors (accelerometers and rate gyros) are all differential measurements and require constant adjustment from the absolute sensors (GPS and compass) in order to eliminate accumulating errors.

The pilot should be in control and the AI features should be assisting, not overriding, so it's very difficult for me to understand that it would just decide to fly away. It seems like as soon as it picked up a GPS signal it just decides that the primary source of control isn't the pilot anymore, fitting the description of "rogue".

I'm surprised, after all the preceding posts, that you still appear to have no clue how these flight control systems work. It didn't "decide to fly away" - it was trying to hold position and/or respond to your stick inputs.

Again, I can see how the RTH location might be off, but it updated correctly upon take off right before the flyaway.

It had nothing to do with the RTH location, as explained previously.

Very frustrating that "such smart tech" could be so... Inferior to human control.

Try flying a quad in full manual mode (you can put a Mavic into manual with a parameter adjustment), and then tell us about how the technology is inferior to your control.

 

[Smithy12]

Here's my very basic explanation:
1. GPS tells the aircraft its position on earth (doesn't control altitude, its barometer does that)
2. Compass tells the aircraft which way is North. They work together to keep the aircraft hovering in one place.
3. See attached diagram; In this example, the aircraft's has initialized thinking "North" is really "East". The aircraft starts in a hover at location #1 and is blown North by the wind to location#2. The aircraft then wants to return to its original position by traveling South. But it thinks South is really West and it moves to position#3. It will continue to spiral (toilet bowl) out of control.

Firstly sorry for original posters loss - I should have said that in my last post!!

Thanks to both Sars and Lonjsnyder for these great explanations ... right up my street, and remarkably similar to the AHRS (attitude heading reference system) that we have on live aircraft. They use Kalman Filters which are very much like the time integration maths that go on in the IMUs to provide the best attitude solution based on a variety of inputs.

So I guess this leads to the next question .... how can you prevent this happening to you?

I noticed the diagnostics so far pointed to a GPS lock occurring after takeoff, which I guess would be considered as the first no-no but are there any other options / things to look for within the DJI go software (other than automated messages I guess) that can be checked before takeoff to confirm no gross errors in heading exist?

As a newbie to drone flying I’d sort of like to get together a checklist of ‘before flight’ actions to minimise this kind of event happening.

Also emergency actions to take if you suspect this kind of event happening?

Once again thanks for the answers so far really great information!,

 

[Lonjsnyder]

Firstly sorry for original posters loss - I should have said that in my last post!!

Thanks to both Sars and Lonjsnyder for these great explanations ... right up my street, and remarkably similar to the AHRS (attitude heading reference system) that we have on live aircraft. They use Kalman Filters which are very much like the time integration maths that go on in the IMUs to provide the best attitude solution based on a variety of inputs.

So I guess this leads to the next question .... how can you prevent this happening to you?

I noticed the diagnostics so far pointed to a GPS lock occurring after takeoff, which I guess would be considered as the first no-no but are there any other options / things to look for within the DJI go software (other than automated messages I guess) that can be checked before takeoff to confirm no gross errors in heading exist?

As a newbie to drone flying I’d sort of like to get together a checklist of ‘before flight’ actions to minimise this kind of event happening.

Also emergency actions to take if you suspect this kind of event happening?

Once again thanks for the answers so far really great information!,

Please see my prior post #38 in this thread

 

[Meta4]

are there any other options / things to look for within the DJI go software (other than automated messages I guess) that can be checked before takeoff to confirm no gross errors in heading exist?
As a newbie to drone flying I’d sort of like to get together a checklist of ‘before flight’ actions to minimise this kind of event happening.

Avoiding launching from steel or reinforced concrete structures and surfaces would prevent more than 90% of these incidents.
DJI should have that warning in the manual.
You can also check the aircraft icon's alignment in the map or radar views matches the actual alignment of the drone before takeoff.

Also emergency actions to take if you suspect this kind of event happening?

By the time you have suspicions, it's probably too late.
If the difference between north and where the compass thinks north is, is very small, you might be able to make an emergency landing.
But in most cases, all you can do is wave good-bye.