H-inf and Kalman references for a controls person

-Hello all,

Does anyone have some good references on H-inf filter and Kalman
filters for a controls person? I am pretty familiar with how these
work (and the associated algorithms) in a controls context and I am
trying to broaden my horizons with their DSP-sister uses. Thanks for
your time.

AM

On 20 Mai, 20:12, AM <quarkthed...@gmail.com> wrote:
> -Hello all,
>
> Does anyone have some good references on H-inf filter and Kalman
> filters for a controls person? I am pretty familiar with how these
> work (and the associated algorithms) in a controls context and I am
> trying to broaden my horizons with their DSP-sister uses.

The best prespective text I have found on Kalman filters, is the
Durbin & Koopman book on state space methods for time series
analyses. The subject of the book is econometrics, not DSP,
but it contains a number of strategies and approaches I have not
seen in DSP texts.

Rune

On May 21, 6:12&#4294967295;am, AM <quarkthed...@gmail.com> wrote:
> -Hello all,
>
> Does anyone have some good references on H-inf filter and Kalman
> filters for a controls person? I am pretty familiar with how these
> work (and the associated algorithms) in a controls context and I am
> trying to broaden my horizons with their DSP-sister uses. Thanks for
> your time.
>
> AM

For KF an old book by Anderson and Moore..

AM wrote:
> -Hello all,
> 
> Does anyone have some good references on H-inf filter and Kalman
> filters for a controls person? I am pretty familiar with how these
> work (and the associated algorithms) in a controls context and I am
> trying to broaden my horizons with their DSP-sister uses. Thanks for
> your time.

I don't see much difference between Kalman filters for use in 'control 
problems' and those for use in 'DSP-sister' uses.  There may be certain 
applications that lean one way or another (i.e. for adaptive control 
your Kalman will almost always be time-varying, while if you're doing an 
observer it'll be whatever the plant is, if you're demodulating AM then 
it'll be time varying again, etc.)

But I'm ready to have my horizons broadened.

-- 
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com

@hardy and @rune, thanks for the books, i will take a look at them.

@Tim

I am probably splitting hairs with the DSP v. controls distinction. My
experience with these methods is entirely from a controls context. So
I figured that learning about them from a different viewpoint would
ensure that i really understand them (as opposed to being able to get
them to work for a controller only).

-Andrew

On May 20, 4:14&#4294967295;pm, Tim Wescott <t...@seemywebsite.now> wrote:
> AM wrote:
> > -Hello all,
>
> > Does anyone have some good references on H-inf filter and Kalman
> > filters for a controls person? I am pretty familiar with how these
> > work (and the associated algorithms) in a controls context and I am
> > trying to broaden my horizons with their DSP-sister uses. Thanks for
> > your time.
>
> I don't see much difference between Kalman filters for use in 'control
> problems' and those for use in 'DSP-sister' uses. &#4294967295;There may be certain
> applications that lean one way or another (i.e. for adaptive control
> your Kalman will almost always be time-varying, while if you're doing an
> observer it'll be whatever the plant is, if you're demodulating AM then
> it'll be time varying again, etc.)
>
> But I'm ready to have my horizons broadened.
>
> --
> Tim Wescott
> Control system and signal processing consultingwww.wescottdesign.com

AM wrote:
(top posting fixed)
> On May 20, 4:14 pm, Tim Wescott <t...@seemywebsite.now> wrote:
>> AM wrote:
>>> -Hello all,
>>> Does anyone have some good references on H-inf filter and Kalman
>>> filters for a controls person? I am pretty familiar with how these
>>> work (and the associated algorithms) in a controls context and I am
>>> trying to broaden my horizons with their DSP-sister uses. Thanks for
>>> your time.
>> I don't see much difference between Kalman filters for use in 'control
>> problems' and those for use in 'DSP-sister' uses.  There may be certain
>> applications that lean one way or another (i.e. for adaptive control
>> your Kalman will almost always be time-varying, while if you're doing an
>> observer it'll be whatever the plant is, if you're demodulating AM then
>> it'll be time varying again, etc.)
>>
>> But I'm ready to have my horizons broadened.

> @hardy and @rune, thanks for the books, i will take a look at them.
> 
> @Tim
> 
> I am probably splitting hairs with the DSP v. controls distinction. My
> experience with these methods is entirely from a controls context. So
> I figured that learning about them from a different viewpoint would
> ensure that i really understand them (as opposed to being able to get
> them to work for a controller only).
> 
> -Andrew

I only have one book -- "Optimal State Estimation" by Simon.  That's if 
you don't count "Estimation and Detection Theory" by Van Trees, which 
has you derive the Kalman filter in a few homework problems (!), and a 
couple of adaptive control books.  I don't know how it stacks up against 
others for completeness, but it's been sufficient for me (and I've been 
_using_ it to generate revenue for over a year).  He has a very down to 
earth, practical style, without forgetting that Kalman filtering is a 
mathematical process and the math has to be paid attention to.

-- 
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com

On 20 May, 14:12, AM <quarkthed...@gmail.com> wrote:
> -Hello all,
>
> Does anyone have some good references on H-inf filter and Kalman
> filters for a controls person? I am pretty familiar with how these
> work (and the associated algorithms) in a controls context and I am
> trying to broaden my horizons with their DSP-sister uses. Thanks for
> your time.
>
> AM

It's not that far between what the controls guys do and what DSP
people do, as Tim says.

This is a paper I did many moons ago that uses an extended KF approach
to do demodulation of polynomial phase signals:

http://espace.library.uq.edu.au/view/UQ:11308

And it's not really H-infinity filtering, but this is some stuff I did
even more moons ago that takes some H-infinity ideas and turns them
into an FIR filter design/approximation approach:

http://espace.library.uq.edu.au/view/UQ:10631
http://espace.library.uq.edu.au/view/UQ:11246

Ciao,

Peter K.