Get the wind data from Mavic 2 Zoom?

[Smilingz]

If all you want are the attitude data then why are you bothering with the DAT files. Attitude data are recorded in the regular txt flight log at 10 Hz, and your txt logs were not missing, were they?

Litchi does create its own csv log, but it also creates a regular DJI SDK txt log.

This has just solved everything... Thank you!

do they look correct?

 

[BudWalker]

This has just solved everything... Thank you!
View attachment 95243
do they look correct?

Not quite. I think you want the AC roll, pitch and yaw.

 

[Smilingz]

Not quite. I think you want the AC roll, pitch and yaw.
View attachment 95246

This?

 

[sar104]

View attachment 95248
This?

That is the aircraft pitch, roll and yaw.

 

[BudWalker]

View attachment 95248
This?

Yes. BTW, CsvView will compute the tilt inclination angle and and direction of that tilt in the earth reference

In this case the AC is tilted 3.1° from vertical towards the west.

 

[Smilingz]

That is the aircraft pitch, roll and yaw.

Thank you

 

[Smilingz]

Yes. BTW, CsvView will compute the tilt inclination angle and and direction of that tilt in the earth reference
View attachment 95264
In this case the AC is tilted 3.1° from vertical towards the west.

Ok thank you so much! Is there anything else important to know of what the data represents?
We will basically feed the attitude along with reference wind velocity (speed and direction) to a neural network. After about 200 000 data points we're hoping that if the drone sends "pitch x jaw y roll z" then the neural network will understand that "aha that means wind speed phi and direction omega" becuase of the training.

 

[Smilingz]

Again I can't understate how much I appreciate the help I've gotten. Thank you so much @BudWalker and @sar104 for the assistance!

 

[sar104]

Ok thank you so much! Is there anything else important to know of what the data represents?
We will basically feed the attitude along with reference wind velocity (speed and direction) to a neural network. After about 200 000 data points we're hoping that if the drone sends "pitch x jaw y roll z" then the neural network will understand that "aha that means wind speed phi and direction omega" becuase of the training.

OK - as long as this factors in that pitch, roll and yaw is a function of both airspeed and aircraft acceleration.

 

[BudWalker]

Ok thank you so much! Is there anything else important to know of what the data represents?
We will basically feed the attitude along with reference wind velocity (speed and direction) to a neural network. After about 200 000 data points we're hoping that if the drone sends "pitch x jaw y roll z" then the neural network will understand that "aha that means wind speed phi and direction omega" becuase of the training.

You'll also need velocities North, East and vertical. In the Go .txt conversion those are OSD:xSpeed, OSD:ySpeed and OSD:zSpeed. The neural net will have to know the acceleration in each of the three axes so time is also necessary. Alternatively, the accelerations could be computed external to, and then given to the neural net.

 

[Smilingz]

@sar104 @BudWalker
The drone is going to hover in place, "stationary", if you will. It's only proof of concept and they only want to know that it can be done!

As such we're going to neglect any small movement of the drone, of course we could very well do a much more complex analysis of the data, factor on movements of the drone, air friction coefficient in different tilt angles and so on. But as said only a proof of concept

We're basically hoping that given enough data points and environment differences the neural network can give rough estimates of the wind velocity. Given our hypothesis we will simply use two-sample t-test to evaluate if the project is a sucesss or not.

 

[sar104]

@sar104 @BudWalker
The drone is going to hover in place, "stationary", if you will. It's only proof of concept and they only want to know that it can be done!

As such we're going to neglect any small movement of the drone, of course we could very well do a much more complex analysis of the data, factor on movements of the drone, air friction coefficient in different tilt angles and so on. But as said only a proof of concept

We're basically hoping that given enough data points and environment differences the neural network can give rough estimates of the wind velocity. Given our hypothesis we will simply use two-sample t-test to evaluate if the project is a sucesss or not.

That should work.