Hi, can anybody explain what Kalman filters are for ? Where can I use them ? What is the benefit of a Kalman filter ? Are there any disadvantages ? Thanks in advance Oliver

# What are Kalman filters for ?

OZ wrote:> Hi, > can anybody explain what Kalman filters are for ? > > Where can I use them ? > What is the benefit of a Kalman filter ? > Are there any disadvantages ?http://www.google.com/search?q=Kalman+filter Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������

OZ wrote:> Hi, > can anybody explain what Kalman filters are for ? > > Where can I use them ? > What is the benefit of a Kalman filter ? > Are there any disadvantages ? > > Thanks in advance > Oliver >Kalman Filters are used for many things. I suggest that you get a copy of "Applied Optimal Estimation" by Gelb, MIT Press. The book is very readable and has many applications.

"OZ" <oliver.zind@web.de> wrote in message news:42db9fe3.4338312@news.btx.dtag.de...> Hi, > can anybody explain what Kalman filters are for ? > > Where can I use them ? > What is the benefit of a Kalman filter ? > Are there any disadvantages ? > > Thanks in advance > Oliver >Its an extension of Wieners work on random signals + noise. Given a random signal + noise you can make an estimate of the signal. You get a smaller mean squared error than a Wiener filter for time-varying cases (for which the Wiener filter is not supposed to be applicable but is still used as part of an LMS algorithm or whatever). You will need a state-space description of your signal model. Can also be used for state-estimation and state-feedback control - thing called certainty equivalence. The optimal feedback controller for the noise-free case can be used if you include a KF when there is noise.There are also Extended Kalman filters but they are a bit hit and miss - non-linear.Can be made to work but I wouldn't go near them.In such cases you estimate the states at the same time as the system model! Shytot

Hi shytot, thanks for this answer. But I will need some time to understand it. Oliver>> >Its an extension of Wieners work on random signals + noise. Given a random >signal + noise you can make an estimate of the signal. You get a smaller >mean squared error than a Wiener filter for time-varying cases (for which >the Wiener filter is not supposed to be applicable but is still used as part >of an LMS algorithm or whatever). >You will need a state-space description of your signal model. >Can also be used for state-estimation and state-feedback control - thing >called certainty equivalence. The optimal feedback controller for the >noise-free case can be used if you include a KF when there is noise.There >are also Extended Kalman filters but they are a bit hit and miss - >non-linear.Can be made to work but I wouldn't go near them.In such cases you >estimate the states at the same time as the system model! > >Shytot > >