Mavic Air seemingly flew away

[Irish-apple]

As @msinger mentioned above, some clarification might help here. You said that the aircraft took off on a heading of SW or so, and then spun around to face you. Looking first at the IMU yaw data, that does appear to be the case:

View attachment 38676

The IMU yaw initialized at -125° (approximately SW) and then the data indicate that it yawed CW to -16° starting at the 5 second mark. But that yaw maneuver was caused by full CW rudder applied for 2 seconds. Is that the yaw maneuver that you were referring to?

Jesus Christ what kind of machine are you lol you remind me of something from the terminator or something pal. we see lines on a page and want to colour them in you see a map or how it happened

 

[ddublu]

Jesus Christ what kind of machine are you lol you remind me of something from the terminator or something pal. we see lines on a page and want to colour them in you see a map or how it happened

Hahaha..

 

[msinger]

I am confident I took off, waited a sec or two and then tried to fly straight away from me. That's what I do every time I take off.

So, you did not yaw the aircraft and point it toward yourself before flying away from the takeoff location (like your video shows)?

 

[sar104]

Glad to...like some have said, this dissection is pretty fascinating to me (just sucks that it happened to me this time). I don't recall actual stick movements but I am confident I took off, waited a sec or two and then tried to fly straight away from me. That's what I do every time I take off. I can remember that sinking feeling when things left my control so my first reaction was to increase altitude (I have had plenty of situations where I lose contact and raising altitude almost always reconnects) but then I quickly thought that going up may not make the most sense given my proximity to Lake Shore Drive so I stopped doing that and may have actually started to decrease altitude. I also remember giving hard stick left as I saw it heading right. I actually did hit RTH but it was too late and the connection had already been lost.

Your recollection is not correct. Immediately after takeoff you applied full backwards elevator, and it reacted accordingly. After the first yaw maneuver you applied forward elevator, and it again reacted correctly. Those observations aside, something did go wrong at 13.5 seconds. Still working on that.

 

[sar104]

The comparison of location computed from IMU x and y velocities with the fusion location using GPS is fascinating and alarming. The markers are at one second intervals, with the 15 s locations indicated.



If you look at the pitch, roll and yaw data the issue becomes clearer:



Note that after the start of significant motion around 8 seconds, and especially after 15 seconds, the pitch is negative and the roll is small. The indicated yaw is approximately north before turning CCW towards NW. That combination should cause the aircraft to move forwards, towards the north, but it is actually moving south. That means that the yaw is almost 180° off. A 180° yaw error will cause exactly this kind of uncontrolled flight in P-GPS as the FC attempts to correct the motion but with exactly the opposite pitch and roll than is actually needed.

The next question is how can the yaw be 180° off if it was initialized correctly, as "appeared" to be the case. The only suggestion I can offer there is that the OP actually placed the aircraft facing the NE and incorrectly thought that it was facing SW. That's only a hypothesis, however.

If that were correct, however, then the explanation is magnetic distortion at the takeoff location, causing the yaw to initialize 180° off the correct value.

 

[msinger]

That means that the yaw is almost 180° off.

That's what the data seems to indicate. The aircraft moved away from the OP as the elevator was in the full up position and the aircraft was pointing toward him (an impossible combination).

 

[sar104]

That's what the data seems to indicate. The aircraft moved away from the OP as the elevator was in the full up position and the aircraft was pointing toward him (an impossible combination).

My suspicion is that perhaps the aircraft did not spin towards the OP on takeoff, but that it was actually already pointing that way at takeoff, so the OP assumed that it must have spun. The yaw maneuver at 5 seconds was to make it face away from the OP in the direction that he originally intended. That explains how the initialized yaw really was 180° out.

 

[msinger]

That explains how the initialized yaw really was 180° out.

... and why we don't take off from sidewalks

 

[ddublu]

Amazing amounts of information. Thank you for the analysis. I don’t typically fly from sidewalks but when I do, it’s Dos Equis. Kidding...when I do, I’ve encountered compass errors and immediately changed location. That didn’t happen here but I’m sure the sidewalk didn’t help things.

 

[sar104]

Amazing amounts of information. Thank you for the analysis. I don’t typically fly from sidewalks but when I do, it’s Dos Equis. Kidding...when I do, I’ve encountered compass errors and immediately changed location. That didn’t happen here but I’m sure the sidewalk didn’t help things.

So is it possible that the aircraft was actually facing you at at takeoff, rather than spinning around after takeoff?

 

[ddublu]

So is it possible that the aircraft was actually facing you at at takeoff, rather than spinning around after takeoff?

Possible but very unlikely. It’s one of the most common reasons that new pilots crash and probably the one thing I say the most when talking to new pilots.

 

[sar104]

Possible but very unlikely. It’s one of the most common reasons that new pilots crash and probably the one thing I say the most when talking to new pilots.

OK - well it's the only consistent explanation for the log data and the event.

 

[MavicCF]

The comparison of location computed from IMU x and y velocities with the fusion location using GPS is fascinating and alarming. The markers are at one second intervals, with the 15 s locations indicated.

View attachment 38686

If you look at the pitch, roll and yaw data the issue becomes clearer:

View attachment 38687

Note that after the start of significant motion around 8 seconds, and especially after 15 seconds, the pitch is negative and the roll is small. The indicated yaw is approximately north before turning CCW towards NW. That combination should cause the aircraft to move forwards, towards the north, but it is actually moving south. That means that the yaw is almost 180° off. A 180° yaw error will cause exactly this kind of uncontrolled flight in P-GPS as the FC attempts to correct the motion but with exactly the opposite pitch and roll than is actually needed.

The next question is how can the yaw be 180° off if it was initialized correctly, as "appeared" to be the case. The only suggestion I can offer there is that the OP actually placed the aircraft facing the NE and incorrectly thought that it was facing SW. That's only a hypothesis, however.

If that were correct, however, then the explanation is magnetic distortion at the takeoff location, causing the yaw to initialize 180° off the correct value.

I thought that was what I saw, I tried your suggestion from a previous thread of comparing the flight paths of X and Y speeds and GPS. Came to the same conclusion of about 180* discrepancy. It is a really cool and useful trick to have!

 

[sar104]

I thought that was what I saw, I tried your suggestion from a previous thread of comparing the flight paths of X and Y speeds and GPS. Came to the same conclusion of about 180* discrepancy. It is a really cool and useful trick to have!

Right - it's simple to integrate the x and y velocities with respect to time and then compare with the lat/long. Ideally you want to convert those to distance north and east for a direct comparison. You can do that with a simple procedure if you are using Igor. This one also fills in any blanks:

Function LatlongToMeters(LatWave, LongWave)
Wave/D LatWave, LongWave
Duplicate/O/D LatWave, Nm, Em
variable/g radius = 6371e3 // Mean earth radius in meters
int pt, firstdatarow
Nm[0] = 0
Em[0] = 0
pt = -1
// Find first row of lat/long data
do
pt += 1
while (numtype(LatWave[pt]) != 0)
firstdatarow = pt
pt = 0
// Replace blank data with first lat/long data
do
LatWave[Pt] = LatWave[firstdatarow]
LongWave[Pt] = LongWave[firstdatarow]
pt += 1
while (numtype(LatWave[pt]) != 0)
WaveStats /Q LatWave
pt = 1
// Fill in blank lat/long entries with previous valid entry and convert to meters offset north and east from first location
do
if (numtype(LatWave[Pt]) != 0)
LatWave[Pt] = LatWave[Pt-1]
LongWave[Pt] = LongWave[Pt-1]
endif
Nm[pt] = Nm[pt-1] + radius*(LatWave[pt] - LatWave[pt-1])*pi/180
Em[pt] = Em[pt-1] + radius*cos(LatWave[pt]*pi/180)*(LongWave[pt] - LongWave[pt-1])*pi/180
pt += 1
while (pt <= V_endRow)
print "V_endRow = ", V_endRow, pt
End​

 

[dirkclod]

Dang .

 

[ddublu]

View attachment 38694 Dang .

Haha

 

[MavicCF]

View attachment 38694 Dang .

It looks like I need to shave

 

[Simmo]

Classic...

 

[M3d4r]

You guys never cease to amaze me.

 

[Irish-apple]

my head after reading this fair play guys