Covariance Matrix not Generated.

[adityasivaraj]

After calibrating either using RTQF or Kalman, covariance matrix is not generated. I have tried calibration both through RTIMULibCal and RTIMUDemoGL, and the RTIMULib.ini generated does not contain the covariance matrix for linear acceleration, orientation and angular velocity. Does this library generate the covariance matrices or should it be filled in through some other method?

[HongshiTan]

Yes, the current calibration does not generate the covariance matrix. For Kalman fusion, it will be estimated online.

[adityasivaraj]

Hey @HongshiTan Thanks for your time! Could you explain what you mean by "estimated online"? Do you mean it will be generated while using the library in general using the calibration file?

[HongshiTan]

Hi, sorry for the delay, the calibration file does not contain the covariance matrix. Maybe you can send the detailed algorithm that you refer to so that I can identify the difference for you.

[adityasivaraj]

Hi @HongshiTan. I'm sorry, I think there has been a miscommunication. What I meant to say/ask was, in your first comment you mentioned that for Kalman Fusion, the calibration will be "estimated online". I wanted to know what exactly you meant by "estimated online", and can I use that covariance matrix? Because, in my usage the covariance matrix is always shown to be empty, and I cannot see online generated covariance matrix anywhere. I am not referring to any detailed algorithm.

Thanks in Advance!

[SivertHavso]

Hi @adityasivaraj. I believe you are looking for this matrix:

RTIMULib2/RTIMULib/RTFusionKalman4.h

Line 66 in e005545

RTMatrix4x4 m_Pkk; // the updated estimated covariance matrix

Which gets updated in this function:

RTIMULib2/RTIMULib/RTFusionKalman4.cpp

Line 132 in e005545

void RTFusionKalman4::update()

However, I don't think there's any way of accessing this matrix without modifying the code.

[brightproject]

Hello @HongshiTan, @adityasivaraj and @SivertHavso.
I'm using library code RTIMULib for Arduino.
But there is no RTFusionKalman4.h file in the Arduino repository
There is a simplified (according to the author) filter RTFusionRTQF.h
I am also interested in the question about the covariance matrix.
As I understand it, the library RTIMULib was written almost 8 years ago and then microcontrollers were not very powerful.
Now it is possible to run code with double-precision floating point.
Therefore, I would like to try a full-fledged Kalman filter on the Arduino platform.