validation_pendulum

intro

The idea is to validate the angle logging using a pendulum and overlaying the animation from the logger and the real video.

Protocol

The sensor is attach to a pendulum. Then we recontruct the rotation using the "pitch" mesurement of the device. and overlay the reconstruction with the reality.

Sometime the bar disapear and the data are 'Nan', it is due to some drop-out of th SD card.

First attempts 19/09/2019:

Hypothese:

• Pendulum 1m is simulation.
• Reconstruct the pendulum from an angle assuming it is a pendulum.

Logger parameters:

• SDLOG_PROFILE = 19 (default + estimator replay+ high rate )

The data logger at 250hz are reample using Pandas resample: with interpolate(method='linear').

• resample at 25 Hz
• interpolate(method='linear')

I use kdenlive for the overlay

Results:

• Video

Comments

• Camera not exactly normal to the pendulum axis.
• Resampling not working well
• Missing the pendulum attachement on the video.

Second attempts 25/09/2019:

Based on the data and video than the first one.

code:

• resample(time_gap).last() , last instead of interpolate

result: Better sync of the two, we can observe jitter around the perfect point. We can also detect some roll movement , especially at the start.

• Video

Note by displaying the points, I noticed that some part are missing in the original log ( this why interpolate was not working)

Third attempts 29/09/2019:

With a new pendulum, better quality .

Comments

• SDLOG_PROFILE= 19
• log at 250 hz , downsample at 25Hz

Results:

• Video

** The Pitch Angle mesurement looks good enough to continue**

Ressources:

• PX4 review link for the log of attempt 1 and 2 : https://review.px4.io/plot_app?log=115d59c2-5ffb-4938-895c-ed789f14170c

• PX44 review link for the log of attempt 3 : https://review.px4.io/plot_app?log=3487c0a5-0572-4742-be51-a29b747fb739

To-Do:

• understand the missing point
• trying with lower log rate, or other log mode.