Lu Zhaojin
Guest
|
 Posted: Fri Nov 07, 2008 2:43 pm Post subject: Topics on particle filtering |
 |
|
now I>m working on particle filtering, but I have some questions
still confused.
what is the advantage of particle filtering compare with the kalman
filtering, from the paper I know that particle filtering can works
well in non-linear/non-Gaussian. but in the real case, what kind of
application is non-linear/non-Gaussian?
and what is the advantage and disadvantage of kalman filtering in real
application.
maybe I can understand roughly from journal or books. but still so
abstract. so I want to get deep understanding of this one, is there
anyone can give me some suggestions or recommendations?
I need solid example and demo.
Thank you very much. |
|
Chris Maryan
Guest
|
| Posted: Fri Nov 07, 2008 2:51 pm Post subject: Re: Topics on particle filtering |
|
|
On Nov 7, 9:43 am, Lu Zhaojin <luzhao...@gmail.com> wrote:
[quote] now I>m working on particle filtering, but I have some questions
still confused.
what is the advantage of particle filtering compare with the kalman
filtering, from the paper I know that particle filtering can works
well in non-linear/non-Gaussian. but in the real case, what kind of
application is non-linear/non-Gaussian?
and what is the advantage and disadvantage of kalman filtering in real
application.
maybe I can understand roughly from journal or books. but still so
abstract. so I want to get deep understanding of this one, is there
anyone can give me some suggestions or recommendations?
I need solid example and demo.
Thank you very much.
[/quote]
Regarding your first question, write out the model you are trying to
estimate, if it>s anything other than linear/gaussian, you probably
need something other than a Kalman filter. The biggest users of it
that I know use it for radar tracking and navigation (terrain
following - there are some good examples on the web, especially when
you look up "marginalized particle filter", which is a combination of
Kalman filter and PF)
The downside is massively higher computational complexity and very
poor scalability with the number of variables you are trying to
estimate.
Chris |
|
