Mavic 2 crashed; trying to learn/figure out why

[j2pilot]

New Mavic pilot with a sad story of a crash. I'm very much looking to learn from this, as well as figure out why the drone was seemingly flying on its own, and then fell from the sky. My flight log is attached.

I began the flight with GPS, 8 satellites, and a home point set. The flight proceeded without incident and I took some photos & videos of a beautiful river valley.

At 19m flight time exactly, I had the Mavic hovering a good 50 feet above a river. I was looking at the Smart Controller screen when I noticed the drone began flying on it's own, north, towards a cluster of trees.

Between 19m 0s and 19m 10.7s you can see the drone moving north without any input from the right stick (mode 2). I was shocked to look up and see the drone careening into the trees without having been controlling it. At 19m 10.9s I panicked and tried to make a corrective action. It seems I didn't control in the right direction and a second or two later the Mavic flew into some trees.

At 19m 20s, the drone came out the other side of the trees and to my greatest fear stopped about 12 feet above a group of people who were having their photo taken. Given what had just happened my first instinct was to quickly ask the people to move from underneath the drone, and they quickly did.

A few seconds later the drone just dropped hard from the sky, about 12 feet to the ground, crashing and really damaging the Mavic. I was really glad the group of people were able to move and none of them were hurt. Checking the logs it appears the drone began auto landing... however, instead of coming the full 20 feet to the ground, it somehow thought it was only a few feet off the ground.

I've got a lot to learn obviously, but I'm really hoping to understand what happened. At 19m, why did the drone start moving on its own? And at the end, why did it begin autolanding and then just fall from the sky? I've got Care Refresh, luckily, I'm not feeling very confident in being able to trust my abilities or the Mavic itself. I appreciate any advice or guidance.

 

[Seattle Drone Repair]

Sounds like a pretty unusual issue. Usually the Mavics do a pretty good job of flying themselves. First of all I would recommend reaching out to DJI and see if they will cover it under warranty because it looks like you were not a fault for the crash. I took a look at your flight logs and it looks like it wasn't throwing any crazy errors but the battery was getting low towards the end of the flight which could trigger autoland. There are some settings in the DJI Go app you can adjust to change the way the drone behaves in low battery mode. I suppose it is possible that the ultrasonic sensor on the bottom was detecting a tree instead of the ground but that doesn't explain the random movements. Did you calibrate the compass or IMU ahead of the flight? Sometimes a failure of those components can cause issues as well.

 

[j2pilot]

Thanks, much apperciated. Great idea on contact DJI regarding a warranty replacement.

Did you calibrate the compass or IMU ahead of the flight? Sometimes a failure of those components can cause issues as well.

I did not calibrate either of those things; I made the false assumption that if calibration was needed the app/controller would have prompted me to do so. I'll definitely make a habit of calibrating before flying in the future though.

I'm still perplexed as to why the random movement north even started at 19m... could have avoided the whole crash!

 

[Seattle Drone Repair]

I did not calibrate either of those things; I made the false assumption that if calibration was needed the app/controller would have prompted me to do so. I'll definitely make a habit of calibrating before flying in the future though.

Sounds like a good plan. Might help you avoid some issues in the future.

Best of luck getting your issue resolved.

-Kale

 

[BD0G]

Lets see if @sar104 wants to take a look at your flight log and determine what happened.

 

[Meta4]

Did you calibrate the compass or IMU ahead of the flight? Sometimes a failure of those components can cause issues as well.

I did not calibrate either of those things; I made the false assumption that if calibration was needed the app/controller would have prompted me to do so. I'll definitely make a habit of calibrating before flying in the future though.

There's no need to recalibrate either the IMU or the compass before a flight and doing it would have made no difference and wouldn't make the flight any safer.
There's no hint that there were any issues at all related to the compass or IMU in the flight.

Here's what the flight data looks like: DJI Flight Log Viewer - PhantomHelp.com
Your launch point was very poor for GPS reception, being at the bottom of a 1000 ft deep narrow valley, the terrain blocked a significant part of the sky.
That reduced the number of satellites your drone could acquire.
GPS health was 1/5 (terrible) for the first 40 seconds of the flight and did not reach 5/5 until you climbed to 400 ft 5 minutes later.

At the time you were concerned about, you were back near the valley floor again and GPS health dropped to 1/5 again for 5 seconds.
Your drone lost horizontal position holding ability without a good GPS signal and was subject to drifting on the breeze.
That's the movement that alarmed you, and it was quite gentle and slow.
You still had full control but probably no braking ability.

The landing issue isn't clear to me.
Autolanding commences at 19:37.3 (did you push the autoland button?) with VPS indicating the drone was 16 feet above something.
At 19:39.4 the descent goes wrong 12 feet above the ground.
The drone rolls sideways and gets turned 90 degrees clockwise as it comes down.

Was it in an open area or is it possible that it clipped a tree or other obstacle?

 

[j2pilot]

@Meta4, super, super helpful. Thank you. I foolishly and mistakenly assumed that when the GPS bar turned green, I was good to go. I'll learn to keep a better eye on the number of satellites and the GPS health. And very good to know that low GPS signal can cause the drone to drift unexpectedly.

Autolanding commences at 19:37.3 (did you push the autoland button?) with VPS indicating the drone was 16 feet above something.
At 19:39.4 the descent goes wrong 12 feet above the ground.
The drone rolls sideways and gets turned 90 degrees clockwise as it comes down.

Was it in an open area or is it possible that it clipped a tree or other obstacle?

I had not yet commenced auto-landing. The drone, a good 10 seconds earlier had flown through some trees and after the crash I noticed that one of the props had been slightly damaged-- its ends were missing a centimeter or so. Perhaps the damaged blade caused it to roll as it was attempting to land.

 

[Mav1c]

If it loses GPS signal it will enter Attitude (atti) mode. It will hold it's elevation but it will drift with the wind and will no have active braking when you let off the sticks. The drone will act like a hockey puck on ice. If you aren't used to it, it can be quite a surprise and invoke some panic as you found out.

There's really no way to prepare for this as it is not a selectable flight mode on the Mavic series unless you hack it with the assistant software.

 

[j2pilot]

If it loses GPS signal it will enter Attitude (atti) mode. It will hold it's elevation but it will drift with the wind and will no have active braking when you let off the sticks. The drone will act like a hockey puck on ice. If you aren't used to it, it can be quite a surprise and invoke some panic as you found out.

There's really no way to prepare for this as it is not a selectable flight mode on the Mavic series unless you hack it with the assistant software.

Looking at the logs, though, it doesn't seem like it ever entered ATTI mode. It was in P-GPS the whole time, and particularly at the time (19m 0s) when the drifting and crash started.

 

[Mav1c]

Looking at the logs, though, it doesn't seem like it ever entered ATTI mode. It was in P-GPS the whole time, and particularly at the time (19m 0s) when the drifting and crash started.

Yeah that part is above my knowledge. I'm interested to see what @sar104 has to say about this..

 

[runeOfRune]

The initial reason for position uncertainty was weak GPS (health 3 minimum with 7 sats). The drone lost precise positioning and started drifting. It flew through some trees and hit some branches (pitch/roll values, upwards obstacle avoided, downwards vision 0.3, 0, 0.7m). Finally, AC lost front right prop, rolled and collided with terrain.

If it loses GPS signal it will enter Attitude (atti) mode. It will hold it's elevation but it will drift with the wind

Downwards ultrasonic had reading of 6m throughout the emergency (minus the part where it hit some branches), so the AC would have switched to OPTI. For some reason, it didn't.

 

[Meta4]

Looking at the logs, though, it doesn't seem like it ever entered ATTI mode. It was in P-GPS the whole time, and particularly at the time (19m 0s) when the drifting and crash started.

You need to look at the actual data rather than a summary to see the GPS health.
Click in the Download CSV on the Phantomhelp summary and go through the spreadsheet of data, looking at Column T "GPS Level"
That shows the drone having GPS Level1/5 for the first 40 seconds and dropping in and out of 1/5 many times during the flight.
It only had good GPS for the period 5:23-8:51 when it was above 400 ft and seeing more sky than it could down in the valley floor.
The period 19:16.6-19:20.7 is the one of concern.
Although the GPS receiver was getting signal from 8 sats but the flight controller had no confidence in the position data due to a poor satellite spread, multi-path issues or other factors.
Although it had not officially dropped into atti mode, the flight controller was aware of errors in the GPS data and had the lowest confidence level in the data, as it had for several other parts of the flight.

Flying in a poor GPS reception area as you were, it's a good idea to look at the number of bars shown next to the satellite icon on your display for an idea of how reliable GPS is.

 

[sar104]

This looks to me simply like a case of poor GPS reception, as suggested by others. There was uncommanded drift:





If you want to explore the cause further then post the mobile device DAT file (...FLY015.DAT).

Mobile device DAT file: How to retrieve a V3.DAT from the tablet

 

[j2pilot]

@Meta4 @runeOfRune @sar104 Thanks very much for your further analysis. It's been very enlightening to learn about how low GPS signal can cause uncontrolled drifts. Certainly didn't learn that from all of the flying tutorial YouTubes I've watched.

I'll be repairing/replacing my Mavic, and taking extra care with future flights. Is keeping an eye on the GPS # of satellites and signal strength enough while flying? Any other tips to avoid this type of crash in the future?

I'm terrified of future uncontrolled drifts.

 

[Meta4]

Is keeping an eye on the GPS # of satellites and signal strength enough while flying?

If you are flying where your Mavic gets a good view of most of the sky rather than at the bottom of a deep valley or somewhere else where a lot of the sky is blocked, you should always get more satellites than you need.

 

[j2pilot]

Finally was able to get the DAT files off the controller. I welcome and appreciate any further analysis, @sar104 et. al. I wasn't sure which if these two files it was as I had two flights that day:

DAT file 1: 19-05-26-05-21-51_FLY015.DAT
DAT file 2: 19-05-26-07-34-49_FLY017.DAT

The thing that is surprising to me is that when the first uncommanded drift started at 19m 0s sharp, that GPS strength was 4 with 9 satellites... not great it seems, but certainly seems like it should have held it's position, right? It was that first drift that cause the drone to fly into the trees, and ultimately the final crash.

 

[sar104]

Finally was able to get the DAT files off the controller. I welcome and appreciate any further analysis, @sar104 et. al. I wasn't sure which if these two files it was as I had two flights that day:

DAT file 1: 19-05-26-05-21-51_FLY015.DAT
DAT file 2: 19-05-26-07-34-49_FLY017.DAT

The thing that is surprising to me is that when the first uncommanded drift started at 19m 0s sharp, that GPS strength was 4 with 9 satellites... not great it seems, but certainly seems like it should have held it's position, right? It was that first drift that cause the drone to fly into the trees, and ultimately the final crash.

Your compass, and maybe IMU needs calibrating - you have some 2π periodic magnetic artifacts in the magnetometer data:

 

[Deleted member 93449]

Your compass, and maybe IMU needs calibrating - you have some 2π periodic magnetic artifacts in the magnetometer data:

View attachment 74467

How does calibrating the compass and IMU get rid of magnetic artifacts? Aren't the artifacts the product of something in the surrounding environment?

 

[sar104]

How does calibrating the compass and IMU get rid of magnetic artifacts? Aren't the artifacts the product of something in the surrounding environment?

Quite the opposite - the artifacts are caused by the magnetization state of the aircraft itself, and those need to be calibrated out.

 

[Deleted member 93449]

Quite the opposite - the artifacts are caused by the magnetization state of the aircraft itself, and those need to be calibrated out.

Does this mean every time the aircraft is flown the compass and IMU have to be calibrated? Does the magnetization state of the air ever change or is it constant? If it changes, what cause the change?