Mavic started to flyaway but recovered

[dcubrakovic]

Hi pilots,
I had a strange situation on my last flight, and although I have tried to analyze what was going wrong, I could not see any problem. I would appreciate if someone could take a look at the log?
The problem: mavic air started to drift and did not react to commands. Eventually, it recovered and started to react on my commands.
Drone started just fine, Home point recorded, enough GPS satellites, some wind yes but not very strong (later on I flew without any problems). Connection was good all the time. About time 1:05 mavic starts to move without any command from my side, with the speed of about 4.5m/s . I think he just drifted with the wind, not trying to hold it's position at all. The "out of control" lasted from 1:05 to 1:26, when it started to follow my commands again and I brought it back. For example, at 1:11 I tried throttle full back, and from 1:14 to 1:17 full forward, but mavic was not moving in wished direction, just drifted with the wind. At about 1:26 he followed my command to turn and then he followed full forward and was flying back (to the wind) with about 5m/s.
Later on, the fly was pretty much as expected.
-Can anyone tell why he flew away and not stayed at his position? I lost confidence in it

The log can be found here:

Thanks

 

[Drone senior pilot]

I had the same happen to my mavic air. I don’t know why but it finally came back. This happened on the fourth flight it was ever flown

 

[sar104]

I don't see any evidence of a flyaway - the aircraft responded appropriately to commands at all times. If you compare the velocity of the aircraft forwards with the elevator input you will see that it moves as commanded at all times, except during one period of RTH and one period of ActiveTrack when it is under autonomous FC control.

 

[Meta4]

I had a strange situation on my last flight, and although I have tried to analyze what was going wrong, I could not see any problem. I would appreciate if someone could take a look at the log?
The problem: mavic air started to drift and did not react to commands. Eventually, it recovered and started to react on my commands.
Drone started just fine, Home point recorded, enough GPS satellites, some wind yes but not very strong (later on I flew without any problems). Connection was good all the time. About time 1:05 mavic starts to move without any command from my side, with the speed of about 4.5m/s . I think he just drifted with the wind, not trying to hold it's position at all. The "out of control" lasted from 1:05 to 1:26, when it started to follow my commands again and I brought it back. For example, at 1:11 I tried throttle full back, and from 1:14 to 1:17 full forward, but mavic was not moving in wished direction, just drifted with the wind. At about 1:26 he followed my command to turn and then he followed full forward and was flying back (to the wind) with about 5m/s.

Here's your flight data: DJI Flight Log Viewer - PhantomHelp.com
Like Sar104, I can't find anywhere that your Mavic didn't follow the joystick input you gave it.
There's certainly no point where it "flew away".

What I do see though is that Until 1:11.1 you only pushed the left stick forward and went straight up.
When you got above 25m, the Mavic starts to show some horizontal movement which increases steadily as the drone goes higher.
You don't use the right stick until the drone is at 50 metres, by which time the horizontal movement is 4.5 m/s.
This looks very much like a strong wind that is making it difficult for the Mavic to hold position.
It must have been a strong wind to do that.
At 1:13.9 you give it full right stick but this hardly changes the horizontal speed.
At 1:27.5 you give it a longer burst of full right stick and also bring it down to 40m but the speed never exceeds 5 m/s on a bearing of 72 degrees.
That suggests it was quite a gusty, strong wind affecting your flight.

 

[Rickolas]

Wind speed at height will be different to wind speed on the ground, so Meta4's comment is valid.

 

[sar104]

Here's your flight data: DJI Flight Log Viewer - PhantomHelp.com
Like Sar104, I can't find anywhere that your Mavic didn't follow the joystick input you gave it.
There's certainly no point where it "flew away".

What I do see though is that Until 1:11.1 you only pushed the left stick forward and went straight up.
When you got above 25m, the Mavic starts to show some horizontal movement which increases steadily as the drone goes higher.
You don't use the right stick until the drone is at 50 metres, by which time the horizontal movement is 4.5 m/s.
This looks very much like a strong wind that is making it difficult for the Mavic to hold position.
It must have been a strong wind to do that.
At 1:13.9 you give it full right stick but this hardly changes the horizontal speed.
At 1:27.5 you give it a longer burst of full right stick and also bring it down to 40m but the speed never exceeds 5 m/s on a bearing of 72 degrees.
That suggests it was quite a gusty, strong wind affecting your flight.

That's what happens when one is in too much of a hurry and only looks at one direction (elevator). I should have checked aileron response too, which does show significant drift from 55 to 87 seconds:



That shows around 70 m of drift to the SW in around 30 seconds. That's consistent with the wind direction, but the wind speed estimate is only around 20 mph and the aircraft attitude did not max out, so I'm not sure why it did that.

 

[dcubrakovic]

Thank you for such an analysis. Something was definitely wrong in time-slot 55-87s. There was wind, but not too strong, I am sure mavic should handle it. I was flying further, and during next 10 minutes mavic was flying up to 100m attitude (where wind is probably even stronger as in 50m where misbehavior happened), without any problems. I am not completely inexperienced, I have about 180 hours of flying and I felt something is wrong. I was flying (by purpose) in strong winds and I know how mavic is behaving. Mavic can hold it's position in ~30 mph (13m/s). If it was here drifting with 5m/s, it means wind was about 18 m/s? This lasts about 20-25 seconds and after that he is able to fly with 5m/s in headwind? And after that 10 minutes in all directions and up to 100m attitude without problem? Not likely. This is the first time that I experienced strange behavior.
My rough guess is that something in GPS-compass was temporarily wrong, and that mavic was thinking that it was moving and he was trying to compensate. But maybe I am wrong.
Although the reason is not known, I want to thank you all, all this was very useful for me. I learned further things, and also learned to be more careful. And from this and similar cases I learned to feel bad because we do not have "switch to atti" command

 

[sar104]

Thank you for such an analysis. Something was definitely wrong in time-slot 55-87s. There was wind, but not too strong, I am sure mavic should handle it. I was flying further, and during next 10 minutes mavic was flying up to 100m attitude (where wind is probably even stronger as in 50m where misbehavior happened), without any problems. I am not completely inexperienced, I have about 180 hours of flying and I felt something is wrong. I was flying (by purpose) in strong winds and I know how mavic is behaving. Mavic can hold it's position in ~30 mph (13m/s). If it was here drifting with 5m/s, it means wind was about 18 m/s? This lasts about 20-25 seconds and after that he is able to fly with 5m/s in headwind? And after that 10 minutes in all directions and up to 100m attitude without problem? Not likely. This is the first time that I experienced strange behavior.
My rough guess is that something in GPS-compass was temporarily wrong, and that mavic was thinking that it was moving and he was trying to compensate. But maybe I am wrong.
Although the reason is not known, I want to thank you all, all this was very useful for me. I learned further things, and also learned to be more careful. And from this and similar cases I learned to feel bad because we do not have "switch to atti" command

The FC knew that it was moving - that's the puzzle here. And GPS health was already at level 5 when it started to move. @BudWalker?

 

[CanadaDrone]

Not enough satellites and numerous wind / attitude warnings looks like the culprit. Wind is dramatically different in the air than it is on the ground, and your drone was repeatedly reporting that it was too strong. That would be my diagnosis.

 

[sar104]

Not enough satellites and numerous wind / attitude warnings looks like the culprit. Wind is dramatically different in the air than it is on the ground, and your drone was repeatedly reporting that it was too strong. That would be my diagnosis.

I don't think either of those was the problem though. GPS was fine, and the IMU data recorded the drift, so it knew it was drifting. If the wind had been the problem then the aircraft would have maxed out on attitude (pitch and/or roll) but it didn't. What is also apparent is that the FC is largely ignoring the stick inputs during this early part of the flight.

 

[BudWalker]

The FC knew that it was moving - that's the puzzle here. And GPS health was already at level 5 when it started to move. @BudWalker?

Yeah, the FC had to have known the MA was moving but chose not to increase the roll to compensate.

 

[BudWalker]

The FC knew that it was moving - that's the puzzle here. And GPS health was already at level 5 when it started to move. @BudWalker?

Maybe the eventLog stream will have something interesting. @dcubrakovic if you could provide the .DAT log file from the tablet we could take a look. Look here to see how to retrieve the .DAT. The file name will be something like XXXXFLY040.DAT.

 

[dcubrakovic]

Maybe the eventLog stream will have something interesting. @dcubrakovic if you could provide the .DAT log file from the tablet we could take a look. Look here to see how to retrieve the .DAT. The file name will be something like XXXXFLY040.DAT.

Here it is, attached. I have just realized from your signature where these graphs on this forum are coming from I have installed CsvView (really great tool), but I am not familiar with all the signals and what they mean.

Beside signals, in the same graph, how can I see the commands on Remote Controller for example?

 

[BudWalker]

Here it is, attached. I have just realized from your signature where these graphs on this forum are coming from I have installed CsvView (really great tool), but I am not familiar with all the signals and what they mean.

Beside signals, in the same graph, how can I see the commands on Remote Controller for example?

Actually, the graphs that you see in this thread are created by @sar104 using Igor Pro. It is many more capabilities than CsvView has. CsvView was my answer to difficulties encountered when using the Excel plot package.

For some info about the .DAT signals take a look at this
V3 .CSV column descriptions
The RC section has some stuff about the RC commands

I don't know of any tabulated info about the .txt signals. But, it has considerably more RC related info.

Regarding your .DAT there was some data in the eventLog stream that could be relevant
68.183 : [L-FDI][CTRL]: fault on , horiz_ctrl_fail
69.638 : [L-BATTERY]get_cell_voltage_callback_ack failed!||
71.195 : [L-FDI][CTRL]: fault off, horiz_ctrl_fail

Doesn't add much as to why this occurred. I'm guessing that the FC algorithms are pretty complicated and are always being developed, modified, tweaked, etc. Maybe what we're seeing here is that an error was detected followed by a recovery. Personally, if it were my MP, I'd assume this is a one-off.

 

[sar104]

Actually, the graphs that you see in this thread are created by @sar104 using Igor Pro. It is many more capabilities than CsvView has. CsvView was my answer to difficulties encountered when using the Excel plot package.

For some info about the .DAT signals take a look at this
V3 .CSV column descriptions
The RC section has some stuff about the RC commands

I don't know of any tabulated info about the .txt signals. But, it has considerably more RC related info.

Regarding your .DAT there was some data in the eventLog stream that could be relevant
68.183 : [L-FDI][CTRL]: fault on , horiz_ctrl_fail
69.638 : [L-BATTERY]get_cell_voltage_callback_ack failed!||
71.195 : [L-FDI][CTRL]: fault off, horiz_ctrl_fail

Doesn't add much as to why this occurred. I'm guessing that the FC algorithms are pretty complicated and are always being developed, modified, tweaked, etc. Maybe what we're seeing here is that an error was detected followed by a recovery. Personally, if it were my MP, I'd assume this is a one-off.

The horizontal control fault appeared at the same time that the drift started, and while it went off 3 seconds later but the drift continued, I'd suspect they are related. Another thing that I notice about the DAT log is that the magnetic yaw data are very noisy, and show significant deviation from the IMU yaw in a number of places.





That noise looks excessive me and appears to be being induced by the motors. I'd take a look at the compass interference value in the GO app, and do a compass calibration, but there may be more going on.

 

[dcubrakovic]

Regarding your .DAT there was some data in the eventLog stream that could be relevant
68.183 : [L-FDI][CTRL]: fault on , horiz_ctrl_fail
69.638 : [L-BATTERY]get_cell_voltage_callback_ack failed!||
71.195 : [L-FDI][CTRL]: fault off, horiz_ctrl_fail

Doesn't add much as to why this occurred. I'm guessing that the FC algorithms are pretty complicated and are always being developed, modified, tweaked, etc. Maybe what we're seeing here is that an error was detected followed by a recovery. Personally, if it were my MP, I'd assume this is a one-off.

Looks exactly so. Googling for this error, I found some results that were in all cases connected with crashes/unexpected behavior.
I will take it as one-off, I will continue to fly carefully, and I will analyze my logs more often.
Thanks!

 

[BudWalker]

The horizontal control fault appeared at the same time that the drift started, and while it went off 3 seconds later but the drift continued, I'd suspect they are related. Another thing that I notice about the DAT log is that the magnetic yaw data are very noisy, and show significant deviation from the IMU yaw in a number of places.

View attachment 53862

View attachment 53864

That noise looks excessive me and appears to be being induced by the motors. I'd take a look at the compass interference value in the GO app, and do a compass calibration, but there may be more going on.

Good call! I plotted magMod with a motor current which illustrates your point. But, notice the funny behavior prior to motorStart when the MA is sitting on the ground not moving. MagMod should have just a small amount of noise. That's a .15 HZ signal.

 

[BudWalker]

@dcubrakovic what is the circular feature surrounding the launch site?


Possibly a particle accelerator facility? I noticed this is at the Karlsruhe Institute of Technology

 

[sar104]

@dcubrakovic what is the circular feature surrounding the launch site?
View attachment 53870

Possibly a particle accelerator facility? I noticed this is at the Karlsruhe Institute of Technology

it's just a park.

 

[dcubrakovic]

@dcubrakovic what is the circular feature surrounding the launch site?

Good try, but no, it is just the very special city. City founder built his castle some 300 years ago as the absolute center, and then he planned 32 "sun rays" going from the center radially.. And the circular features are just streets (outer circle) and paths (inner circle). There is a football stadion in the right-top edge of your map, it is no-fly zone I think.

Here is the picture that I took that day (after the drone recovered), looking to the castle in the center: