Re: [Bayes++] Accuracy of Kalman filter prediction

[Michael S. <ma...@mi...>]

Dear Nithya,

On Monday 07 May 2007 18:12, Nithya Nirmal Vijayakumar wrote:
> Hi all,
>
> I am using Bayes++ Covariance filter for prediction of interger/float
> values in a sequence. From general reading, I found that Kalman filter can
> say how good its prediction is. Can I find out how good my prediciton is
> from the Covariance filter?

The Coveriance filter is the standard implementation of the Kalman filter. 
That is it estimates the state vector (x) and a covariance matrix (X). The 
covariance is the estimate of how good the predicted/estimate state is.

> I currently follow the simpleExample.cpp for 
> predict, update steps. Is there a function which I can invoke to get its
> prediction accuracy?

The simpleExample is 1D position estimator. In this case there is only a 
single element in the covariance matrix, the variance of the position 
estimate. This is printed as  along with the state estimate at each step.

If you take the square root of the variance you have the standard deviation of 
the state estimate. The standard deviation is proportional to the 
uncertainity in the state. Therefore the smaller it is the more certain the 
prediction/estimate is.

Michael

-- 
___________________________________
Michael Stevens Systems Engineering

34128 Kassel, Germany
Phone/Fax: +49 561 5218038

Navigation Systems, Estimation  and
                 Bayesian Filtering
    http://bayesclasses.sf.net
___________________________________