Mavic Air flew off, accelerated and smashed in to a wall at nearly 40mph!

[Quick1]

I have had my Mavic Air since it first came out and have used it many times with very few issues until yesterday. I have not used it for 3 months and charged the battery up fully. Went to my local beach to fly it where there is a lot of wide open space. I placed it down on a big flat concrete wall ready to take off and it asked me to do a compass calibration. I didn't think anything of this as that is normal and it will often have me do the calibration dance prior to take off. Anyway that went well and so I put the drone back down and then I checked the WiFi channels to see if I was clear to use the one I had chosen. Then I checked to see if it had locked on to enough satellites, then I asked the drone to take off and record the home point accurately. All going well so far drone took off and hovered at 6m. I moved forward a few meters and was positioned over some rocks ready to fly out over the beach when I briefly saw an error message come up "IMU heading error, please try flying forwards or backwards or land". As I was over some rocks I slowly moved drone forwards about 20m to see if that would help and the drone few straight with no issues as far as I could see.

Suddenly the drone started moving to the left and away from me so I let go of the sticks but it kept moving away from me, I tried to correct it and bring it back but it did not respond and started accelerating away from me quicker and quicker in an arc and I could see it heading back to the promenade on its own, I tried to get it to respond but it just kept accelerating until it got to nearly 40mph (17.6m/s) when it flew through the metal railings on to promenade and then smashed in to a small brick wall. The front and back legs both broke off (one completely ripped off , the small tab on the side the battery came off, one propeller broke and the others are not great and it has a few minor scratches on the body. To be honest I was expecting a lot worse.

Firstly I am annoyed that such an expensive bit of kit can go just go crazy and out of control like that. Secondly I just relived that it didn't hit anyone of anything else. I am also relived that it did not just fly out to sea or somewhere I would not be able to get it or find it.

So about the loss of control. I noticed from looking back at the flight record that the drone flew forwards about 20m but the DJI GO 4 app flight record shows it is flying backwards. So I am guessing the drone may of tried to correct itself and the more corrections it made the worse the situation got. My big question is why didn't the drone recognise this was happening and stop, not just accelerate away with a total loss of control. This is very dangerous!

Also I now need to get this bird repaired. So do I get a front and back arm on the internet and fit them myself or take the option of getting it repaired by someone like Drone Doctor or BuildYourOwnDrone?

 

[Alex Baxter]

Sorry for your troubles..
Sounds like a classic case of a 'toilet bowl' crash caused by magnetic interference, were you certain there was no steel in the concrete wall or nearby when you did the compass calibration.

 

[skyrex]

+1 On Alex's comment. Without checking the logs I can almost guarantee that concrete had reo-bar (steel) in it affecting the compass calibration.
The calibration done would have then taken in to account being near a large magnetic structure so after take off being in the air and away from the interference it would have immediately put the compass back out of calibration at which point the compass and IMU start fighting.

 

[Davros007]

I agree with you: the issue is dangerous in certain situations. And I bet it is a compass issue. You would think that a confusion between GPS and internal compass could trigger an automatic descent (which may not be straight down as it doesn’t know where it is, but a rapid descent nevertheless). Seems less risk than trying to maintain altitude/position. You’re stuffed if it goes loopy and heads off into a NFZ full tilt.

 

[slup]

I have had my Mavic Air since it first came out and have used it many times with very few issues until yesterday. I have not used it for 3 months and charged the battery up fully. Went to my local beach to fly it where there is a lot of wide open space. I placed it down on a big flat concrete wall ready to take off and it asked me to do a compass calibration. I didn't think anything of this as that is normal and it will often have me do the calibration dance prior to take off. Anyway that went well and so I put the drone back down and then I checked the WiFi channels to see if I was clear to use the one I had chosen. Then I checked to see if it had locked on to enough satellites, then I asked the drone to take off and record the home point accurately. All going well so far drone took off and hovered at 6m. I moved forward a few meters and was positioned over some rocks ready to fly out over the beach when I briefly saw an error message come up "IMU heading error, please try flying forwards or backwards or land". As I was over some rocks I slowly moved drone forwards about 20m to see if that would help and the drone few straight with no issues as far as I could see.

Suddenly the drone started moving to the left and away from me so I let go of the sticks but it kept moving away from me, I tried to correct it and bring it back but it did not respond and started accelerating away from me quicker and quicker in an arc and I could see it heading back to the promenade on its own, I tried to get it to respond but it just kept accelerating until it got to nearly 40mph (17.6m/s) when it flew through the metal railings on to promenade and then smashed in to a small brick wall. The front and back legs both broke off (one completely ripped off , the small tab on the side the battery came off, one propeller broke and the others are not great and it has a few minor scratches on the body. To be honest I was expecting a lot worse.

Firstly I am annoyed that such an expensive bit of kit can go just go crazy and out of control like that. Secondly I just relived that it didn't hit anyone of anything else. I am also relived that it did not just fly out to sea or somewhere I would not be able to get it or find it.

So about the loss of control. I noticed from looking back at the flight record that the drone flew forwards about 20m but the DJI GO 4 app flight record shows it is flying backwards. So I am guessing the drone may of tried to correct itself and the more corrections it made the worse the situation got. My big question is why didn't the drone recognise this was happening and stop, not just accelerate away with a total loss of control. This is very dangerous!

Also I now need to get this bird repaired. So do I get a front and back arm on the internet and fit them myself or take the option of getting it repaired by someone like Drone Doctor or BuildYourOwnDrone?

As all others I said "Yaw error due to magnetic disturbed launch spot" to myself as soon as I read your thread title ...

The DAT log showed the issue pretty clearly ... the IMU Yaw & magYaw angle (Green/Blue graph) are initiated to the same (wrong) value at power on. But once airborne at 0sec & away from the disturbance the magYaw corrects & show the correct heading angle ... which is unknown for the IMU Yaw.



The marker in the chart is placed where the sat picture shows the AC below ... the difference in IMUYaw & magYaw values in the chart above is nearly 180 degrees ... below the angles is shown by the green & blue bars.



You made the fatal mistake not checking the drone icon direction on the map in GO4 & compared if that was equal to reality. Your MA didn't know where it was pointing making it apply thrust to the wrong motors to correct position hold ... leading to an even larger position error ... and again more thrust with wrong motors & the fly away speed picked up.

 

[Alex Baxter]

So I am guessing the drone may of tried to correct itself and the more corrections it made the worse the situation got. My big question is why didn't the drone recognise this was happening and stop, not just accelerate away with a total loss of control. This is very dangerous!

The drone was doing it's best to hover, but was being fed one signal from the IMU and another from the compass, it does not know which is correct so it goes into a death spiral.

 

[Quick1]

Sorry for your troubles..
Sounds like a classic case of a 'toilet bowl' crash caused by magnetic interference, were you certain there was no steel in the concrete wall or nearby when you did the compass calibration.

There was a low steel pipe fence nearby and I am guessing that there is steel rebar in the concrete. I just saw nice flat place away from people to take off and didn't really think about magnetic interference to be honest. I was just excited to be getting out again after 3 months.

 

[Quick1]

So I am guessing the drone may of tried to correct itself and the more corrections it made the worse the situation got. My big question is why didn't the drone recognise this was happening and stop, not just accelerate away with a total loss of control. This is very dangerous!

The drone was doing it's best to hover, but was being fed one signal from the IMU and another from the compass, it does not know which is correct so it goes into a death spiral.

From having control and flying out a few meters in a straight line all seamed fine with the drone. From the moment I realised that something was up and I had no controls and the drone was shooting off until it crashed was literally seconds. Hardly any time to think things through or react any differently to what I did. This to me is a safety issue with the drone. Imagine getting hit by that at nearly 40mph and how fast would it of gone if the wall had not stopped it when it did. Would it just of kept going round and round?

 

[Alex Baxter]

I don't wish to be contentious, but I regret to have to say in my opinion it's your responsibility, not the drones. It was in an unsafe condition when it took off.

 

[Quick1]

I don't wish to be contentious, but I regret to have to say in my opinion it's your responsibility, not the drones. It was in an unsafe condition when it took off.

You do all the checks you think you need to do at the time but sadly things like magnetic interference from a concrete take off point did not even cross my mind. I certainly will in future! Just need to get it repaired now.

 

[Starshots]

Out of interest do DJI have anything in writing to advise against flying next to metal objects ? Is it a design flaw that the electronics can't cope with ?

 

[BossBob]

Copied from the Mavic Pro manual.

DO NOT calibrate your compass where there is a chance of strong magnetic interference, such as magnetite, parking structures, and steel reinforcements underground.
DO NOT carry ferromagnetic materials with you during calibration such as cellular phones.
The DJI GO 4 app will prompt you to resolve the compass issue if the compass is affected by strong interference after calibration is complete. Follow the prompted instructions to resolve the compass issue.

I’d be surprised if similar wasn’t included in the manuals of subsequent models.

 

[Alex Baxter]

Out of interest do DJI have anything in writing to advise against flying next to metal objects ? Is it a design flaw that the electronics can't cope with ?

It's a design flaw inasmuch as a magnetic compass is flawed when you do orienteering and get lost due to magnetic interference with the steel box you happen to keep the compass in.

 

[slup]

So I am guessing the drone may of tried to correct itself and the more corrections it made the worse the situation got. My big question is why didn't the drone recognise this was happening and stop, not just accelerate away with a total loss of control. This is very dangerous! ...

That's correct ... like a rolling snowball. The flight controller will eventually give up & go to ATTI, but in most cases when it comes to yaw errors most AC will have crashed by then ...

 

[JAW]

Air vs. Wall at nearly 40mph ...... I’m surprised there’s anything left to repair.

 

[BadWolf1]

Copied from the Mavic Pro manual.

DO NOT calibrate your compass ...

I’d be surprised if similar wasn’t included in the manuals of subsequent models.

Calibration isn't the same

Every bit of metal in the drone affects the reading of the earth's magnetic field. It's a constant, so the "rotate in two dimensions" thing is to work out what needs to be added to / subtracted from magnetic field readings.

If the system is correctly calibrated but the magnetic field at take take off is distorted (or if the aircraft flies into a strongly distorted field, which is less likely up in the air) then the inertial system will say the drone is pointing here, and the magnetic system will say it is pointing there.

I think the magMod value in the log is the total field strength and it changes massively before take off and in the first few seconds of the flight, which just reinforces what we already know - the pre-flight readings were wrong. If I have that right the software has two indicators to say the compass is wrong but it doesn't use them.

 

[SkyeHigh]

Calibration isn't the same

Every bit of metal in the drone affects the reading of the earth's magnetic field. It's a constant, so the "rotate in two dimensions" thing is to work out what needs to be added to / subtracted from magnetic field readings.

If the system is correctly calibrated but the magnetic field at take take off is distorted (or if the aircraft flies into a strongly distorted field, which is less likely up in the air) then the inertial system will say the drone is pointing here, and the magnetic system will say it is pointing there.

I think the magMod value in the log is the total field strength and it changes massively before take off and in the first few seconds of the flight, which just reinforces what we already know - the pre-flight readings were wrong. If I have that right the software has two indicators to say the compass is wrong but it doesn't use them.

Only metal containing iron.

 

[Meta4]

I placed it down on a big flat concrete wall ready to take off and it asked me to do a compass calibration.
I didn't think anything of this as that is normal and it will often have me do the calibration dance prior to take off.

You'll find an explanation of the cause of your incident in this post I made on another forum:

Avoiding the rebar effect

I would like to make my takeoff and home point on a concrete portion of my driveway. Thinking of using a wooden table to distance my takeoff point from the rebar underneath. How tall should the table be? .... or is this just a wish that should not be pursued at all .... and why?

phantompilots.com

Out of interest do DJI have anything in writing to advise against flying next to metal objects ? Is it a design flaw that the electronics can't cope with ?

DJI manuals have always been deficient in not mentioning this.
It's a common cause of crashed drones.

Copied from the Mavic Pro manual.
DO NOT calibrate your compass where there is a chance of strong magnetic interference, such as magnetite, parking structures, and steel reinforcements underground.

The issue had nothing to do with compass calibration.
It was all about powering the drone on in a spot where the earth's magnetic field was distorted by the magnetic field associated with the steel reinforcing inside the concrete.
Read the post I linked to above for details.

 

[sar104]

It's too easy to look at these events and just assume that it was magnetic interference at startup. This event was more complex. The clues are all there, but you have to dig.

Firstly, the position mismatch is characteristic of simple compass interference:



The error does look like nearly 180° (it actually starts off at about 140°), but that's the first clue - 180° yaw errors do not produce curved flight paths - they produce linear uncontrolled flight. So what is going on here?

Firstly, the compass was calibrated after power up, so look there, because it is very telling:



Okay - that's very interesting. Two CCW 360° rotations, as expected, and as detected by the IMU yaw values (green), the VPS yaw (red) and most significantly the pure inertial yaw computed from just the rate gyros (orange). The magnetometer data are clearly wrong though - the first rotation is correct, but when the aircraft is switched to the second requested orientation prior to the second rotation, the magnetic yaw changes by 180° with no actual rotation, and then it doesn't change much at all for the subsequent 360° rotation. That's weird, and a big red flag.

So next look at the aircraft attitude for those two calibration rotations:



First rotation - pitch and roll approximately zero with the aircraft level. Then the aircraft is rolled onto its right side (pitch ≈ 0°, roll ≈ 90°). That causes the 180° change in magnetic yaw when there was actually no change in yaw. That's not the calibration procedure specified in the manual, although that sometimes differs from the on-screen instructions - I don't have an MA so I cannot verify:



The direction of rotation doesn't matter for calibration, but using an unexpected axis may, or may not matter.

Anyway - the compass was apparently significantly out of calibration based on those data, so the question is - did the calibration fix the issue. Looking at takeoff makes it clear that it did not:



The magnetic yaw changes with no physical rotation as the aircraft lifts off - indicating a local external field at the takeoff location. Then, at around 3.8 seconds, the FC rotates the aircraft CCW, likely to attempt to match the new magnetic yaw. Normally that is done while holding the IMU yaw constant; in this case the active IMU1 yaw deviates and is then returned to its original value (-135°), but the magnetic yaw doesn't see the rotation, and so the attempt at convergence fails. At this point it is hard to say what the real yaw value is since we clearly can't trust the compass even after takeoff.

So looking at the remainder of the flight:



After the FC-commanded rotation stops, we have a magnetic yaw of around 10° (N) and an active IMU yaw of around -135° (SW). At 22 seconds the pilot rotates the aircraft further CCW by 162° back to its original takeoff heading, but the compass only sees a 40° CCW turn. IMU1 goes with the rate gyro data and the IMU yaw changes to 70°.

At 26 seconds we finally get to find out the real aircraft yaw, when the pilot applies full forward elevator. Note that the IMU velocities completely disagree with the GPS velocities, which are correct. Computing the actual track from the GPS velocity data shows the following:



The aircraft heads southwest on a track of -135°, and so the original takeoff yaw was actually correct, but IMU1 is now on 70°. The other significant feature is at 28.5 seconds, when the compass records a nearly 180° CCW turn in 0.2 seconds, followed by the reverse turn at 30.5 seconds, also in 0.2 seconds. That's impossible, of course, and another red flag with a very obvious cause - here's the x-axis magnetic field strength plotted with pitch:



That is either a broken compass or a bad calibration, because the magnetic yaw is changing by 180° with a 30° pitch change. Yaw varying with pitch was the main trigger of the flight problems. Whether it was caused by, or made worse by the compass calibration procedure used is hard to say, but the calibration didn't fix it, that much is clear. And the change in yaw on takeoff - normally indicative of magnetic interference, was actually just due to a pitch change on takeoff, followed by a failed reconciliation attempt by the FC. There likely was no interference.

 

[KeviP]

Nothing to do with the crash itself, but I am confused about the compass calibration procedure.
Don't hesitate to tell me I am dumb (or more politely, mis-interpretating things), but for me, the nose is the end where the gimbal and antenna legs are.
The instructions say nose pointing downward but the image shows nose upward!
It probably doesn't make a difference, just being picky.

(Maybe just a comment or two, then we can get back to the subject)