Robust decentralized control

Dear Josh,

Let me first say (in response to your earlier email)
that I would somewhat reluctant to suggest to anyone
to enter into research on decentralized control, since quite a lot
of work has been done, and the practical significance is somewhat
limited (if you want to go to a great effort in designing a controller
then you might as well design a MIMO controller).

Then to your question about checking for RP. If you are talking about
a pure analysys of a given design then checking mu(N)<1 directly is
clearly the best.
However, if you want to design a controller that takes into account
robustness then the procedure you mention ("independent designs")
may be worthwhile (although I would probably not go to such effort;
see above).

>>From kanterjm@seas.upenn.edu Wed Feb 03 19:54:32 1999
>Delivered-To: skoge@ona.chembio.ntnu.no
>Subject: how do you get a robust design?
>
>I have a question about how design of the decentralized controller by
>"independent designs"
>based on the nominal plant, as is done in the book, and robustness
>all
>fit together. Would the design process, once pairings are known go like:
>
>step 1: design a decentralized controller that gives a good nominal
>response by following book.
>>
>step 2: see if robust performance is satisfied.
>
>Also, do you do step 2 by checking mu(N)<1 or by doing what was done with
>M.Morari. in "Robust Performance of Decentralized Control Systems By
>Independent Dersigns?"
>
>What would you do? it seems that checking step 2 with mu(N)<1 is more
>conclusive and easier to do than what was in the paper with M.Morari.
>Any help you give would be GREATLY apreciated.
>
>Josh
Received on Thu Feb 04 08:05:36 1999