Prev: Partial correlation matrix using partialcorr
Next: Increase the precision of an object ( number)
From: ben harper on 26 Dec 2009 09:18
by using "kalman" function i can calculate Kalman estimator gain of my plant with respect to covariance matrixes Qn and Rn.
kalman(sys,Qn,Rn)

I can define covariance amtrix as
Qn = 2.3
Rn = 1

But what if my input disturbance is "white noise",
how can i change the covariance matrix in this situation??
Can i put white noise to the covarience matrix?
From: Michael_RW on 21 Jan 2010 20:32

Ben,

Please consider your audience.

If you are not going to take the time to describe your problem clearly and concisely, why would anyone put forth their effort to respond?


Michael.


---
frmsrcurl: http://compgroups.net/comp.soft-sys.matlab/kalman-filter-white-noise
From: ben harper on 22 Jan 2010 01:25
Hello,

i have a state space model which i call "sys". It has A,B,C,D matrices.
By using Q and R covariance values i use kalman function and calculate static Kalman gain.

Is this gain is for step disturbance effects?



Michael_RW <user(a)compgroups.net/> wrote in message <0amdnRosy8qnncTWnZ2dnUVZ_oWdnZ2d(a)giganews.com>...
>
> Ben,
>
> Please consider your audience.
>
> If you are not going to take the time to describe your problem clearly and concisely, why would anyone put forth their effort to respond?
>
>
> Michael.
>
>
> ---
> frmsrcurl: http://compgroups.net/comp.soft-sys.matlab/kalman-filter-white-noise
From: Michael_RW on 22 Jan 2010 02:03

Ben,

Without details of the source-code you have written, likely in Matlab, I can not comment more specifically about your state-space model or other aspects of your questions.

With respect to "step disturbance effects", I assume you imply your system is operating at a steady-state and it is then "acted-on" by some external force or control-input, yes? With respect to, "http://wapedia.mobi/en/Kalman_filter#4.", the filter model will have a control-input component (matrix Bk in the noted reference).

If you have appropriate models for your case (i.e. state-transition model, control-input model and observation model), and reasonable process noise & observation noise covariances, the gain will be correct.

From your past posts, I assume this is a scalar or one-dimensional Kalman filter application, yes?

Also, keep in mind proper frameworks with respect to underlying Kalman filter assumptions: Gaussian statistics with linear models; non-Gaussian statistics with linear models; non-Gaussian statistics and non-linear models.

Two references come to mind... These involve scalar or one-dimensional Kalman filters with Gaussian statistics and linear models. I can send these to you if you require additional references.

Michael.


---
frmsrcurl: http://compgroups.net/comp.soft-sys.matlab/kalman-filter-white-noise
 | 
Pages: 1
Prev: Partial correlation matrix using partialcorr
Next: Increase the precision of an object ( number)