Hello, I already started some discussions dealing with kalman filter and Hinfinity Filter. But to get the point, again a question: According to my practical experiences so far with implementing both filters, I don't see any difference between them. my asumptions are: covariance matrices of the kalman filter = weighting matrices for the hinfinity filter. can anyone please point out the practical benefits of the hinfinity filter? it was often mentioned in this group to have advantages in some cases over the kalman filter regards, Kai
practical difference between Kalman Filter and Hinfinity Filter
Started by ●September 17, 2008
Reply by ●September 17, 20082008-09-17
On Sep 17, 10:49 pm, Kai <inva...@invalid.invalid> wrote:> Hello, > > I already started some discussions dealing with kalman filter and Hinfinity > Filter. But to get the point, again a question: > > According to my practical experiences so far with implementing both filters, > I don't see any difference between them. my asumptions are: covariance > matrices of the kalman filter = weighting matrices for the hinfinity > filter. > > can anyone please point out the practical benefits of the hinfinity filter? > it was often mentioned in this group to have advantages in some cases over > the kalman filter > > regards, > KaiH infinity can work with deterministic systems and is robust to changes in the model. It minimizes the maximum value of error - not the mean-square error. If you compare mean-square errors you will find not much difference. Try varying the model of the signal generating process and not that of the filter and compare again. The idea arose from robust control where the plant model is not accurately known. Hardy
Reply by ●September 19, 20082008-09-19
On Sep 17, 3:49 pm, Kai <inva...@invalid.invalid> wrote:> Hello, > > I already started some discussions dealing with kalman filter and Hinfinity > Filter. But to get the point, again a question: > > According to my practical experiences so far with implementing both filters, > I don't see any difference between them. my asumptions are: covariance > matrices of the kalman filter = weighting matrices for the hinfinity > filter. > > can anyone please point out the practical benefits of the hinfinity filter? > it was often mentioned in this group to have advantages in some cases over > the kalman filter > > regards, > Kaihii sir i am doing my project related to Kalman and H-infinity filter. i want matlab codes of both and practical comparisions between them urgently. please send if u have any related to that.
Reply by ●September 19, 20082008-09-19
On Sep 17, 11:55 pm, HardySpicer <gyansor...@gmail.com> wrote:> On Sep 17, 10:49 pm, Kai <inva...@invalid.invalid> wrote: > > > > > Hello, > > > I already started some discussions dealing with kalman filter and Hinfinity > > Filter. But to get the point, again a question: > > > According to my practical experiences so far with implementing both filters, > > I don't see any difference between them. my asumptions are: covariance > > matrices of the kalman filter = weighting matrices for the hinfinity > > filter. > > > can anyone please point out the practical benefits of the hinfinity filter? > > it was often mentioned in this group to have advantages in some cases over > > the kalman filter > > > regards, > > Kai > > H infinity can work with deterministic systems and is robust to > changes in the model. It minimizes the maximum value of error - not > the mean-square error. If you compare mean-square errors you will find > not much difference. Try varying the model of the signal generating > process and not that of the filter and compare again. The idea arose > from robust control where the plant model is not accurately known. > > Hardyhii sir i am doing my project related to Kalman and H-infinity filter. i want matlab codes of both and practical comparisions between them urgently. please send if u have any related to that.
Reply by ●September 19, 20082008-09-19
hii sir i am doing my project related to Kalman and H-infinity filter. i want matlab codes of both and practical comparisions between them urgently. please send if u have any related to that. my id krantibalaga@gmail.com Kai wrote:> Hello, > > I already started some discussions dealing with kalman filter and Hinfinity > Filter. But to get the point, again a question: > > According to my practical experiences so far with implementing both filters, > I don't see any difference between them. my asumptions are: covariance > matrices of the kalman filter = weighting matrices for the hinfinity > filter. > > can anyone please point out the practical benefits of the hinfinity filter? > it was often mentioned in this group to have advantages in some cases over > the kalman filter > > regards, > Kai
Reply by ●September 19, 20082008-09-19
On Sep 17, 11:55 pm, HardySpicer <gyansor...@gmail.com> wrote:> On Sep 17, 10:49 pm, Kai <inva...@invalid.invalid> wrote: > > > > > Hello, > > > I already started some discussions dealing with kalman filter and Hinfinity > > Filter. But to get the point, again a question: > > > According to my practical experiences so far with implementing both filters, > > I don't see any difference between them. my asumptions are: covariance > > matrices of the kalman filter = weighting matrices for the hinfinity > > filter. > > > can anyone please point out the practical benefits of the hinfinity filter? > > it was often mentioned in this group to have advantages in some cases over > > the kalman filter > > > regards, > > Kai > > H infinity can work with deterministic systems and is robust to > changes in the model. It minimizes the maximum value of error - not > the mean-square error. If you compare mean-square errors you will find > not much difference. Try varying the model of the signal generating > process and not that of the filter and compare again. The idea arose > from robust control where the plant model is not accurately known. > > Hardyhii sir i am doing my project related to Kalman and H-infinity filter. i want matlab codes of both and practical comparisions between them urgently. please send if u have any related to that. my id krantibalaga@gmail.com
Reply by ●September 19, 20082008-09-19
On Sep 19, 2:52�am, kranti <krantibal...@gmail.com> wrote:> hii �sir �i am doing my project �related to Kalman and H-infinity > filter. i want matlab codes of both and practical comparisions between > them urgently. please send if u have any related to that. my id > krantibal...@gmail.com > > Kai wrote:I want, gimme gimme, I want it now. I don't want to understand. I just want the comparison with out understanding, Waahhhhh! sniff, snort... Waaaaahhhhhhhhhhh! I thought HarderSpicer provided a good answer for when the model is changing. If the model does change one can use steady state Kalman filters and use the steady state H-infinity gains during run time. I have found that under there conditions the gains are similar but the H- infinity gains are just a little larger which just means the error will be a little less at the expense of smoothing. Peter Nachtwey
Reply by ●September 20, 20082008-09-20
On Sep 19, 8:14 pm, pnachtwey <pnacht...@gmail.com> wrote:> On Sep 19, 2:52 am, kranti <krantibal...@gmail.com> wrote:> hii sir i am doing my project related to Kalman and H-infinity > > filter. i want matlab codes of both and practical comparisions between > > them urgently. please send if u have any related to that. my id > > krantibal...@gmail.com > > > Kai wrote: > > I want, gimme gimme, I want it now. I don't want to understand. I > just want the comparison with out understanding, Waahhhhh! sniff, > snort... Waaaaahhhhhhhhhhh! > > I thought HarderSpicer provided a good answer for when the model is > changing. If the model does change one can use steady state Kalman > filters and use the steady state H-infinity gains during run time. I > have found that under there conditions the gains are similar but the H- > infinity gains are just a little larger which just means the error > will be a little less at the expense of smoothing. > > Peter Nachtweysir actually my doubt is when i wrote a matlab code for state estimation using H-infinity filtering we know that the the noise characteristics are known a priori ,sir we are selecting the weighting matrices? and then how we get the both noises inthe matlab code? please consider my doubt.
Reply by ●September 20, 20082008-09-20
On Sep 20, 10:38 am, kranti <krantibal...@gmail.com> wrote:> On Sep 19, 8:14 pm, pnachtwey <pnacht...@gmail.com> wrote: > > > > > On Sep 19, 2:52 am, kranti <krantibal...@gmail.com> wrote:> hii sir i am doing my project related to Kalman and H-infinity > > > filter. i want matlab codes of both and practical comparisions between > > > them urgently. please send if u have any related to that. my id > > > krantibal...@gmail.com > > > > Kai wrote: > > > I want, gimme gimme, I want it now. I don't want to understand. I > > just want the comparison with out understanding, Waahhhhh! sniff, > > snort... Waaaaahhhhhhhhhhh! > > > I thought HarderSpicer provided a good answer for when the model is > > changing. If the model does change one can use steady state Kalman > > filters and use the steady state H-infinity gains during run time. I > > have found that under there conditions the gains are similar but the H- > > infinity gains are just a little larger which just means the error > > will be a little less at the expense of smoothing. > > > Peter Nachtwey > > sir actually my doubt is when i wrote a matlab code for state > estimation using H-infinity filtering we know that the the noise > characteristics are not known a priori ,sir we are selecting the > weighting matrices? and then how we get the both noises inthe matlab > code? please consider my doubt.
Reply by ●September 20, 20082008-09-20
On Sep 20, 10:56 am, kranti <krantibal...@gmail.com> wrote:> On Sep 20, 10:38 am, kranti <krantibal...@gmail.com> wrote: > > > > > > > On Sep 19, 8:14 pm, pnachtwey <pnacht...@gmail.com> wrote: > > > > On Sep 19, 2:52 am, kranti <krantibal...@gmail.com> wrote:> hii sir i am doing my project related to Kalman and H-infinity > > > > filter. i want matlab codes of both and practical comparisions between > > > > them urgently. please send if u have any related to that. my id > > > > krantibal...@gmail.com > > > > > Kai wrote: > > > > I want, gimme gimme, I want it now. I don't want to understand. I > > > just want the comparison with out understanding, Waahhhhh! sniff, > > > snort... Waaaaahhhhhhhhhhh! > > > > I thought HarderSpicer provided a good answer for when the model is > > > changing. If the model does change one can use steady state Kalman > > > filters and use the steady state H-infinity gains during run time. I > > > have found that under there conditions the gains are similar but the H- > > > infinity gains are just a little larger which just means the error > > > will be a little less at the expense of smoothing. > > > > Peter Nachtwey > > > sir actually my doubt is when i wrote a matlab code for state > > estimation using H-infinity filtering we know that the the noise > > characteristics are not known a priori ,sir we are selecting the > > weighting matrices? and then how we get the both noises inthe matlab > > code? please consider my doubt.sir actually my doubt is when i wrote a matlab code for state estimation using H-infinity filtering we know that the the noise characteristics are not known a priori ,sir we are selecting the weighting matrices? and then how we get the both noises inthe matlab code? please consider my doubt.
