Re: Self-learning fuzzy controller

Pramit 'Jake' Sarma (psarma@che.iitb.ernet.in)
Mon, 10 May 1999 13:33:16 +0200 (MET DST)

The information appears to be a little sparse - so here are some initial
assumptions:

(1) One assumes by 'std.' self-organising FLC (SOFLC) you mean
Mamdani/Procyk's 1976 model [MP76], with a meta-fuzzy rulebase (FRB) for
modifying the true controller FRB.

(2) That the plant is in fact controllable.

If (1) & (2) then the scaling factors/gains G=(GE,GC,GU) need to be
fixed a priori. Random values of G will not in general produce any
proper performance. In fact, they have to be heuristically tuned - as
has been separately done in each example in [MP76]. Or, SOFLC is only
universal modulo nontrivial scaling factors.

Or, trial-and-error G tuning may be useful.

On Thu, 8 Apr 1999, Jayeeta Banerjee wrote:

> Hello everbody,
>
> Please take sometime to read this and help a hapless research student if
> you can.
>
> I am working with a standard self-learning fuzzy logic controller and
> unfortunately,I've come across some problems that I can't seem to be
> able to solve. So I was wondering if there's anyone out there, who can
> give me some suggestions.
>
> 1. The controller is trying to track a square wave. It has an empty
> rule-base initially . The output response improves within the first two
> cycles, but then it subsequently 'unlearns' and the behaviour
> deteriorates and then some more cycles down its performance stays
> constant.
> I can't seem to understand why it 'unlearns'.
>
> 2. The number of changes to the rule-base made per cycle remains
> constant and doesn't decrease. Obviously the learning isn't converging.
> What can I do?
>
> 3. If I want to regulate a load disturbance, do I have to have a
> different type of rule-base?
>
> 4. Would anyone know of any references,etc to tackle decoupling for a
> multivariable process using a fuzzy controller?
>
> Thank you for your time.
>
> Regards
>
> Jay
>
>
>
> ############################################################################
> This message was posted through the fuzzy mailing list.
> (1) To subscribe to this mailing list, send a message body of
> "SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
> (2) To unsubscribe from this mailing list, send a message body of
> "UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
> to listproc@dbai.tuwien.ac.at
> (3) To reach the human who maintains the list, send mail to
> fuzzy-owner@dbai.tuwien.ac.at
> (4) WWW access and other information on Fuzzy Sets and Logic see
> http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
> (5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html
>

============================================================================
Pramit "Jake" Sarma
[Home] [IIT-B]

10 Adyaman, S.V. Rd. Dept. of Chem. Engg.
Vile Parle (W) Indian Institute of Technology, Bombay
Mumbai 400 056 Powai, Mumbai 400 076
INDIA INDIA
Phone: (91-22)617-2062 Phone: (91-22)576-7231
Ph/Fax: (91-22)616-2320 Fax: (91-22)578-3480
e-mail: pramits@giasbmc.vsnl.net.in e-mail: psarma@che.iitb.ernet.in

Process Systems and Control
The PROCISS Group
|
{Mathematics/Physics}-<Applied NonLinear Control>-{Industrial World}
|
Process Control, Identification & Simulation Systems
============================================================================

############################################################################
This message was posted through the fuzzy mailing list.
(1) To subscribe to this mailing list, send a message body of
"SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
to listproc@dbai.tuwien.ac.at
(3) To reach the human who maintains the list, send mail to
fuzzy-owner@dbai.tuwien.ac.at
(4) WWW access and other information on Fuzzy Sets and Logic see
http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html