Litchi mission abruptly ends, yaw error, almost crashed

[DJ Wes]

I was flying one of my drones (P3S) this weekend performing several circular missions around a newly built water tower. Prior to taking off, there were no errors displayed in DJI Go. The drone’s orientation matched that shown in DJI Go (I always check). DJI Go showed 12 satellites on the screen. I flew two missions successfully. The distance between the drone and the water tower was approximately 175 feet at all times.

Two thirds of the way through the third mission, I received an unexpected audible notice that my mission ended. I looked at DJI Go and then back at the drone. I noticed it was flying toward the water tower even though my fingers were not touching the controls! I moved the right stick almost all of the way back to move the drone away from the water tower. Fortunately, the drone responded (although the response felt weird) and I was able to land back near the home point on grass near the tarred parking lot.



After landing I noticed in DJI Go that it said “Compass Error”. I went home and downloaded the log file. Upon viewing the log file I can see that “Yaw Error” is shown from the time the mission abruptly ended and continued as the drone flew toward the water tower. I can also see that the “Yaw Errors” stopped once I took over manual control and brought the drone down. The mode also switched to ATTI mode during this time.

Once I got home I flew the drone again and all appears to be ok. Can anyone tell me what might have happened? What exactly is a “Yaw Error” and why did it all-of-a-sudden happen? The log file is attached.

 

[sar104]

I'd need to see the Litchi mission profile to understand this completely. However, there were a couple of unusual aspects to the flight. There is some evidence in the txt log of compass issues, specifically a 2π periodic variation of magnetic yaw with yaw. The switch to ATTI at 198 seconds was due to a significant discrepancy between the IMU yaw and compass yaw, inferred from the difference between IMU and position-derived velocities shown below:



Looking at the flight path and pitch data:





Why did you apply forward elevator at 120 seconds?

After the switch to ATTI at 198 seconds, the motion towards the tower was due to the SW wind at 20 mph. You correct for that and brought it back in ATTI mode by applying full reverse elevator at 206 seconds - it probably felt weird because you are not used to flying in ATTI mode.

The 2π periodicity in the yaw data usually implies some magnetization on the aircraft that has not been calibrated out. Have you done a compass calibration recently?

 

[DJ Wes]

Why did you apply forward elevator at 120 seconds?

This was a new spiral mission I designed. The purpose was to revolve around the water tower three times while increasing in elevation. I pushed the right stick forward slightly at 120 seconds to increase the speed from my mission default of 10mph to approximately 12-13mph just because I was getting board and I wanted to complete this test a little bit faster.

Have you done a compass calibration recently?

No. I think the last time I calibrated may have been 9 month ago or so. When I do get the compass calibration warning (can't remember exactly what it says), powering the drone off and back on in a slightly different location usually makes it go away.

I always compare the drone orientation in the app to the actual drone orientation and they always match. At least as far as I can tell with my own eyes.

 

[DJ Wes]

I'd need to see the Litchi mission profile to understand this completely.

I've made that mission public. Here is the URL:

Mission Hub - Litchi

Litchi's Mission Hub lets you plan waypoint flights and later execute them using the mobile Litchi app.

flylitchi.com

As I mentioned, the mission is a spiral starting at 50 feet AGL and climbing to 200 feet AGL. Litchi does not have a way to make circular paths. The method I used to create this is a little complicated but I can explain if required.

Seeing your plots is very interesting. I want to do a compass calibration, fly another circular path, and see the resulting plots. Is the creation of these plots available to me or are they from your own custom software?

 

[sar104]

I've made that mission public. Here is the URL:

Mission Hub - Litchi

Litchi's Mission Hub lets you plan waypoint flights and later execute them using the mobile Litchi app.

flylitchi.com

I'll take a look.

As I mentioned, the mission is a spiral starting at 50 feet AGL and climbing to 200 feet AGL. Litchi does not have a way to make circular paths. The method I used to create this is a little complicated but I can explain if required.

That's okay - the method isn't important. Creating circular or spiral Litchi profiles is computationally pretty simple - I have a program that will make both.

Seeing your plots is very interesting. I want to do a compass calibration, fly another circular path, and see the resulting plots. Is the creation of these plots available to me or are they from your own custom software?

Those are produced by custom data processing functions in the data analysis program that I use (Wavemetrics Igor Pro).

 

[DJ Wes]

I calibrated my compass using DJI GO, created and flew a new circular mission using Litchi. Does the log for this new flight still show minor compass errors?

 

[sar104]

I calibrated my compass using DJI GO, created and flew a new circular mission using Litchi. Does the log for this new flight still show minor compass errors?

Better - probably good enough - the offset is mostly ± 0.5 m/s, which is typical. There is still some 2π periodicity but not enough to worry about:

 

[DJ Wes]

So, is there anything to be learned from this? Does this suggest that a periodic compass calibration is a good thing?

 

[sar104]

So, is there anything to be learned from this? Does this suggest that a periodic compass calibration is a good thing?

I think that the magnetic posture of the aircraft is prone to change with time, especially when exposed to significant external magnetic fields. That may be why DJI has implemented the 30-day recalibration requirement with the M2. That said, I don't think that it was the cause of the switch to ATTI that you saw on the previous flight. Something triggered the larger discrepancy, and I don't know what that was.