Get more from your Mavic
The largest Mavic community in the world
Join Us Now

Mavic 2 Enterprise Dual IMU Attitude Error followed by immediate loss of control and crash.

jaja6009

Well-Known Member
Joined
Jul 14, 2019
Messages
46
Reactions
39
Hello,

I never wanted to have to post to this section of the forum, but my turn has finally come.

I was flying for my job today in Public Safety and our Mavic 2 ED flashed an IMU Attitude error before immediately flying backwards into a concrete bridge support, falling onto its base and then bouncing into the bay. I put on some waders and recovered it, but it is a complete loss. I was hovering at the time under the bridge which brought me down from 17 to 9 satellites but the aircraft was still hovering fine. I had just taken a picture when this happened and was looking at the FPV to decide if I wanted to get another picture. The drone shot backwards after an error flashed on the screen. My fingers were not on the sticks.

In DJI Pilot I watched the replay available and it shows that the control sticks were not touched.

I am hoping for some clarity on what happened and I will add any insight from here to my accident report along with a video of the DJI Pilot flight replay.

We use a Smart Controller and I attached the Flight Records from the SD Card that was in the Smart Controller. I believe these are the correct files?

Thank you in advance.
 

Attachments

  • DJIFlightRecord_2021-03-03_[10-37-11].txt
    1.6 MB · Views: 11
  • 21-03-03-10-36-11_FLY021.DAT
    9.5 MB · Views: 5

Aerial Edge

Active Member
Joined
Sep 25, 2020
Messages
41
Reactions
41
Age
39
Location
Smithfield VA
Im not a log expert so not much help, but anytime you see your sats droo below 10 its is time to take corrective measures. Because flawed data can be worse then no data, when sats drop like that the drones can have erratic movement and can start doing what is called a toilet bowl maneuvers. We train to do ATTI mode recoveries for this and switch to ATTI mode and determine yaw and fly back, hover and land in ATTI mode. This is really dangerous around things like bridges, or any tight quarters.

I am going to guess the IMU error was cause by loss of satellites. The compass was affected by the Bridge and the drone position thought is changed since the satellite lock had dropped. Drone attempted to correct hover and went into the bridge. There is a reason DJI will not allow for takeoff with less the 8-9 satellites.

GPS mode is not suppose to be used to fly under structures, Vision modes and barometer hold modes are much better for the drones that are equipped with them. Someone will be able to tell you soon from the logs.

Dave
 
  • Like
Reactions: jaja6009

boblui

Well-Known Member
Joined
May 11, 2020
Messages
1,465
Reactions
1,207
Location
Hong Kong
Seems that the propeller on the right-front motor suddenly came off because the speed of that motor maxed out but the current drawn dropped to almost zero indicating that there was no load.

1614832911132.png
 
Last edited:
  • Like
Reactions: jaja6009

jaja6009

Well-Known Member
Joined
Jul 14, 2019
Messages
46
Reactions
39
Im not a log expert so not much help, but anytime you see your sats droo below 10 its is time to take corrective measures. Because flawed data can be worse then no data, when sats drop like that the drones can have erratic movement and can start doing what is called a toilet bowl maneuvers. We train to do ATTI mode recoveries for this and switch to ATTI mode and determine yaw and fly back, hover and land in ATTI mode. This is really dangerous around things like bridges, or any tight quarters.

I am going to guess the IMU error was cause by loss of satellites. The compass was affected by the Bridge and the drone position thought is changed since the satellite lock had dropped. Drone attempted to correct hover and went into the bridge. There is a reason DJI will not allow for takeoff with less the 8-9 satellites.

GPS mode is not suppose to be used to fly under structures, Vision modes and barometer hold modes are much better for the drones that are equipped with them. Someone will be able to tell you soon from the logs.

Dave
Thank you for your response.

Before I went under the bridge I did think about satellite loss and the down VPS. I figured that since I was right in the middle between the underside of the bridge and the water I should of been okay. I was also around 100-150 ft away unobstructed VLOS, so that factored into my decision too as I thought the controller signal would be strong. I thought that if the satellites did drop, the drone would go into Attitude mode which I have used before to train myself how to fly in the event of GPS loss. If it did go into Attitude mode I had the drone facing me and was prepared to bank left the 20-25 feet out from under the bridge.

IMG_20210303_113252.jpg

This was about my position I was standing, but I was probably back another 30-50 feet.

I was not trying to fly under all the way and out the other side, I was positioned in the middle in front of the middle column. I was trying to get pictures where the bridge meets the ground as work was being done there. The drone went full reverse and hit the middle column of the pier, battery popped out and stayed on the pier and drone almost stayed on the pier but then fell in.
 

jaja6009

Well-Known Member
Joined
Jul 14, 2019
Messages
46
Reactions
39
Seems that the propeller on the right-front motor suddenly came off because the speed of that motor maxed out but the current drawn dropped to almost zero indicating that there was no load.

View attachment 125021

Thank you for the reply.

From what I observed, the drone did not spin around as indicated on the DJI Pilot Flight Log replay. It was facing me and went backwards at high speed into the column.


The damage the drone received was:

Right Rear Motor was broken off and hanging by wiring
Right Rear Propellers were missing
Right Front Propellers were missing
Body damage in the form of it trying to separate
Salt Water Damage

DJI says it is out of warranty by 4 days!! o_O
 

jaja6009

Well-Known Member
Joined
Jul 14, 2019
Messages
46
Reactions
39
I just got done viewing the txt log in phantomhelp and it shows the drone to appear to be tumbling, though that is not what I saw. But I must admit I was looking at the FPV screen after taking a phot and looked up right as I heard the motors race and the drone shoot in reverse, so it could have been tumbling, it happened really quick. Also it didn't help that it was bright and sunny where I was standing and under the bridge was darker and contrasting. Also all logs show the drone crash happening over the water while the drone went backwards 15 - 20 feet into the columns.

1615151102615.png

Also could the Right Front Motor high RPM and no current be from the crash? Watching the flight replay on Pilot and remembering what I saw makes me think it is possible that the last 3 seconds of flight where the drone going in reverse, hitting the column, falling on the base and dislodging the battery and then bouncing into the bay water.

I am currently trying to write up an after action report with all findings about contributing factors to the accident and crash. I have already documented bad pilot decisions on going under the bridge, not aborting when satellites dropped in number, knowing that the downward positioning sensor may be affected by the water below if the aircraft went to OPTI mode. In the end I would like this to be a learning and training experience for all of our pilots.
 

Meta4

Well-Known Member
Joined
Oct 30, 2016
Messages
8,430
Solutions
1
Reactions
10,074
Age
65
I just got done viewing the txt log in phantomhelp and it shows the drone to appear to be tumbling, though that is not what I saw. But I must admit I was looking at the FPV screen after taking a phot and looked up right as I heard the motors race and the drone shoot in reverse.
How close to the bridge column do you think the drone was immediately before this happened?
Also could the Right Front Motor high RPM and no current be from the crash?
It will be as a result of propeller damage from the collision.
I am currently trying to write up an after action report with all findings about contributing factors to the accident and crash. I have already documented bad pilot decisions on going under the bridge, not aborting when satellites dropped in number
You had 5 seconds of flight (13:09.3 - 13:14.4) where the bridge blocked enough of the skyview to affect satellite reception significantly.
The flight controller was not confident in the location data for that period and the drone would have effectively been in atti mode for that period.
That is not necessarily a problem and just requires additional caution if flying close to obstacles.
The drone was safe and stable during that 5 seconds of the flight.

But after satellite numbers returned the data shows a discrepancy between indicated speed and GPS speed.
It's a bit hard to explain but it looks like the drone may have "jumped" to catch up with where it thought it was, which would account for the sudden unexpected movement.

The drone collided with something at 13:17.1.
Any error or warnings after that are a result of the collision, not indicating a cause.
In 0.3 seconds the pitch angle went from +65° to -77° while the drone rolled over 169°.
that was the cause of the IMU warning.

 
Last edited:
  • Like
Reactions: jaja6009

Meta4

Well-Known Member
Joined
Oct 30, 2016
Messages
8,430
Solutions
1
Reactions
10,074
Age
65
when sats drop like that the drones can have erratic movement and can start doing what is called a toilet bowl maneuvers
Losing GPS doesn't cause erratic movement or toilet bowl effect (that's a compass thing).
Losing GPS just the drone loses the ability to hold horizontal position.
I am going to guess the IMU error was cause by loss of satellites.
Losing GPS has no effect on the IMU.
The IMU warning was a result of the drone tumbling after a collision.
 
  • Like
Reactions: jaja6009

jaja6009

Well-Known Member
Joined
Jul 14, 2019
Messages
46
Reactions
39
Thank you very much for taking your time to look at this. I want to add independent analysis to the accident report and after action report that I must fill out.

How close to the bridge column do you think the drone was immediately before this happened?

Pretty much the distance shown on the above screenshot, but more under the bridge. You can see where the columns are behind it. It hit the middle one.



Would a fair summary be:

Pilot took large risk flying under bridge
Pilot did not recognize drone changed to ATTI mode
Aircraft could of drifted into column due to ATTI mode and pilot looking at FPV after photograph
Aircraft could of drifted from ATTI and had jump and hit column due to GPS correction

Lesson Learned:
Do not fly under bridges or other structures that could block GPS and induce ATTI
Fly with a VO whenever possible to spot any drift
Train more in ATTI mode, especially in winds to appreciate the drifting aspect

My memory just tells me though that I heard the motors rev and the sound of it hitting the column was a loud sound from a hard impact (Followed of course by the sound of it falling and hitting the concrete base below).

Also the wind that day was from the West at 13.5 mph, so the drifting from ATTI mode makes sense.

Again thank you.

Sad thing is out of all of my own drones and my real jobs drones, I had the second most hours on this one and even with its thermal sensor being kind of a toy, this was my workhorse, especially after they added the digital zoom.
 

boblui

Well-Known Member
Joined
May 11, 2020
Messages
1,465
Reactions
1,207
Location
Hong Kong
Also could the Right Front Motor high RPM and no current be from the crash?

You are right. I was rushing through the flight log data in my previous post and after taking a second more careful look, It can be confirmed that the prop on the right-front motor came off immediately after the collision. The time was so close that I misidentified the cause and result in my previous anlaysis.

1615176249906.png

The data is also in line with your report that the caft accelerated backward by itself because with no stick inputs, the pitch angle went up to 18 degrees ( nose up ) one second before the collision. The craft did reduce the pitch angle after that but it was way too late.

The question now is why the craft did that ? I am sure it was not due to the craft entering ATTI mode because in this mode, the pitch angle will remain zero when there is no elevator stick input although the craft will drift with the wind. From the flight log, there was very little wind as can be seen at this moment. There was no stick inputs, the speed was zero and craft was perfectly level. If there was wind, there should have been some noticeable tilt :

1615177333919.png


Unfortunately the .TXT file you have provided is corrupted in some ways because only 80 seconds of flight data can be seen with the tool I use. Can @BudWalker lend some help pls ?
 
Last edited:
  • Like
Reactions: jaja6009

Meta4

Well-Known Member
Joined
Oct 30, 2016
Messages
8,430
Solutions
1
Reactions
10,074
Age
65
The question now is why the craft did that ? I am sure it was not due to the craft entering ATTI mode
Entering Atti Mode wasn't the problem.
It was coming back to P-GPS mode that caused the movement with the drone jumping to "catch up".
Atti Mode isn't a big problem out in the open, but where the drone can drift in and out of Atti Mode and there are nearby obstacles, it can be.
Unfortunately the .TXT file you have provided is corrupted in some ways because only 80 seconds of flight data can be seen
The txt file looks fine.
Perhaps you mean the .dat file?
 

boblui

Well-Known Member
Joined
May 11, 2020
Messages
1,465
Reactions
1,207
Location
Hong Kong
The txt file looks fine.
Perhaps you mean the .dat file?
The TXT file looks fine on the PhantomHelp interface but not in CsvView. The same amount of data loss is seen in the "verbose csv" file downloaded from PhantomHelp. The csv file is OK but it contains too little information.
 
  • Like
Reactions: Meta4

BudWalker

Well-Known Member
Premium Pilot
Joined
Oct 24, 2016
Messages
1,784
Reactions
1,238
Age
72
Location
Arroyo Grande, California
The TXT file looks fine on the PhantomHelp interface but not in CsvView. The same amount of data loss is seen in the "verbose csv" file downloaded from PhantomHelp. The csv file is OK but it contains too little information.
I think you must have answered your question. CsvView uses TXTlogToCSVtool to convert the .txt to a .csv. PhantomHelp also uses TXTlogToCSVtool to produce the Verbose CSV. For this particular .txt TXTlogToCSVtool chokes and quits before producing the entire .csv
 

jaja6009

Well-Known Member
Joined
Jul 14, 2019
Messages
46
Reactions
39
Thank you to all for your input. If its okay I am going to use this information in my After Action Report in the possible causes section if that is okay.

So my summary:

Cause:
Satellites loss caused aircraft to most likely go into ATTI mode
Aircraft regained GPS mode and possibly "jumped" causing it to strike column

Decisions and Piloting
Bad pilot decision to go under bridge in non emergency mission especially in the section I was under (not much maneuver room). In the middle of the bridge there was much more height and spacing between columns
Pilot should of aborted being under bridge at signs of significant satellite loss
Pilot did not possibly recognize ATTI mode from GPS
If pictures were needed, pilot should of taken angled pictures from sides with full awareness of GPS Satellite numbers


Lessons Learned and Prevention
As mentioned only fly under bridge in emergency and no other course of action
If flying under bridge abort at loss of satellites
Learn to recognize ATTI mode
Training for all pilots using ATTI mode in safe area

One last thing, would it of been better if I just placed the aircraft in ATTI mode and then then flew under it in this same situation?


Again I thank everyone very much! You guys are the NTSB of drones and your info is always enlightening.

Just in case it was something on my end here is the Flight Record again.
 

Attachments

  • DJIFlightRecord_2021-03-03_[10-37-11].txt
    1.6 MB · Views: 0
Lycus Tech Mavic Air 2 Case

New Threads

Forum statistics

Threads
104,378
Messages
1,208,475
Members
133,048
Latest member
nlorntson