UAV toolbox support package for PX4: Calibration slow due to custom flight algorithm

Hello,

I developed my custom flight algorithm(position controller and attitude controller), and uploaded it using default startup script(rcS.txt).

After that, I executed sensor calibration using QGroundControl, and it seems work well but the speed of calibration is way much slow(Take about 1minute for gyroscope calibration).

I am suspecting this symptom is because of my custom flight controller which is too heavy for Pixhawk Cube 2.1 board. But not sure that Controller Computing power can affect to calibration process.

Does development team have experienced this symptom? Or is it possible my controller can affect calibration process due to lack of computing power? Or if it isn't can you give some candidate that I can figure out why calibration is too slow?

Thank you.

Answer 1

[Arun Mathamkode]

To confirm whether the custom app from Simulink (px4_simulink_app) is causing the slowness in calibration, you can try stopping the custom app before trying the calibration. You can use 'px4_simulink_app stop' command in QGC MAVLink Console to stop the custom app.

If you are seeing a performance improvement, may be you are right, the custom flight controller might be very heavy. By default, all the controller logic you design in your Simulink model will be start running as soon as the Simulink custom app starts. You can consider modelling the controllers such that basd on arming status the controller actions are kicking in. You may need to use additional modelling ( Conditionally Executed Subsystems ) for the same.

[종원 이]

Thank you for answering.

Conditionally Executing Subsystem is a great idea. I will check it and comment here whether it worked or not.

I have asked before but still, I am little bit suspecting that computational power can be insufficient to some custom controller algorithm.

Staff Ankur Bose replied to me:

https://www.mathworks.com/matlabcentral/answers/1582249-some-question-about-uav-toolbox-px4-support-package?s_tid=srchtitle

that they had tested many controller and memory was not an issue. If so, what about a problem with computational power? Little bit curious that Pixhawk board can endure MPC or computationally hard contorl algorithm (I will test it soon, but wonder the development team's experience). Although my algorithm was very simple, it seemed to run out of computing power when I went through the calibration process. Therefore I thought that a controller like MPC might not be able to be implemented.

And, as you know, this computational power issue is very important in ensuring that the controller behaves at the desired Hz as designed in Simulink.

Appreciate you again.