If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

• TI Thinks Resolved

# CCS/AWR1843BOOST: The selection of process noise covariance matrix Q

Part Number: AWR1843BOOST

Tool/software: Code Composer Studio

Hi,

In the MRR demo, there is a process noise covariance matrix Q  for kalman filtering algorithm. In the demo, the matrix is defined as follows:

QvecList=halfatsq_x * halfatsq_x;

QvecList=halfatsq_y * halfatsq_y;

QvecList=at_x * at_x;

QvecList=at_y * at_y;

in the above array,halfatsq_x=1/2*ax*t^2,halfatsq_y=1/2*ay*t^2,at_x=ax*t,at_y=ay*t.

According to non-homogeneous state equation knowledge,there should be also other two elements, halfatsq_x*at_x and halfatsq_y*at_y for representing the process noise covariance matrix Q. Why  are the two elements discarded?

Thanks,

Regards,

Rata

• Hi,

We are checking with algorithm developer and will get back to you early next week

thank you
Cesar

• In reply to Rata Zhang:

Hi Rata,

We assume that there is no correlation between velocity and acceleration - the process noise matrix contains only diagonal elements.

There are other approaches - where correlations are assumed between velocity, acceleration (in both x, and y), and they have better performance in certain use-cases .

Regards

Anil

• In reply to Anil Mani:

Hi Anil,

Thanks for your reply, if I add the other two elements to the the process noise covariance matrix Q,it is reasonable, is that right?

Regards,

Rata

• In reply to Rata Zhang:

Hi Rata,

I am not sure how the process noise of acceleration relates to velocity.

Also, You would have to overhaul the kalman filter (KF), since the KF assumes a diagonal process noise matrix.

Regards

Anil