petek@ida.liu.se (Peter Eklund) (06/06/91)
************DEADLINE NOW JULY 1**********
ANNOUNCEMENT - CALL FOR PARTICIPATION
FUZZY CONTROL WORKSHOP
IJCAI-91, SYDNEY, AUSTRALIA
AUGUST 24, 1991
WORKSHOP COMMITTEE
Prof. Dr H J Zimmermann, RWTH Aachen, Aachen
Dr M Reinfrank, Siemens AG, Munich
Prof. L A Zadeh, University of California at Berkeley
Prof. M Sugeno, Tokyo Institute of Technology
Dr D Driankov (Chairman), University of Linkoping
During the past several years fuzzy control has emerged as one of
the most active and fruitful research areas in the application of
fuzzy set theory, especially in the realm of industrial processes
which do not lend themselves to control by conventional methods.
Fuzzy control has appeared as a qualitative extension of classical
control theory and is very similar to AI knowledge representations
in that both model the ``common sense'' knowledge of an experienced
human operator. In essence the theory of fuzzy control provides for
an algorithm which can convert the control knowledge of an operator
into an automatic control strategy. In particular fuzzy control
theory appears very useful when;
* linearity and time-invariance can not be assumed; the responses
to change in manipulated variables are non-linear and highly
sensitive in certain regions. There are significant transport
lags in the process and the process itself is subject to random
disturbances.
* it is difficult to derive differential/difference equations
representing the process i.e. there is a lack of a well-posed
mathematical model. At the same time the ability of the exper-
ienced operator to cope with such a process is recognized and
these operators can describe their knowledge of control actions
linguistically as a set of rules.
* the human understanding of the process and its conventional
mathematical description are alien and this results in a lack of
an effective man-machine interface.
However, despite of the indisputable success of the theory of
fuzzy control, there remain a number of issues which are consider-
ed to be its weak points requiring further investigation and more
solid treatments. It is these issues which will be the focus of
the workshop;
* efficient systematic methods for knowledge acquisition. So far the
process of transferring the operator's knowledge into a usable
knowledge base have been time consuming and non-trivial.
* conception and design of fuzzy control systems that have the
capacity to learn from experience, that is a combination of
techniques from both fuzzy logic and neural networks can improve
the learnability and adaptability of a fuzzy controller in a
changing environment.
* well-founded formal procedures for fuzzy controller design based
on fuzzy models of the process. The need for the development of
fuzzy dynamic systems theory is urgent with its emphasis on the
modeling of the linguistic structure of the process which extends
in a qualitative way the fundamental notions of state, controll-
ability and stability.
It is at this last juncture that the theory of fuzzy control and
recent developments in qualitative reasoning in AI meet each other
and can be cross-fertilized. However, these two approaches have
developed independ- ently from one another and there has been almost
no exchange of ideas between the two scientific communities. In
control theory the terms fuzzy rule-based formalism can be likened
to a qualitative input/output model whereas the AI approach is akin
to a qualitative state-space description and performs the function
of an internal representation of the process. Thus, the fuzzy control
representation describes what an operator does rather than why he
does it. The knowledge about the later can only come from the internal
representation of the process i.e. its model. In this context the
workshop will provide a framework within which the similarities and
differences between the two approaches can be highlighted and
discussed in depth.
Speakers will be by invitation. For participants, a short abstract
of the author's experience in fuzzy control or qualitative reasoning
should be delivered to the workshop secretary at the address below
by 1 July, 1991.
Peter Eklund (Secretary/Organizer)
Department of Computer and Information Science
University of Linkoping
S-581 83 Linkoping, Sweden
tel. (+46) 13 281950
fax. (+46) 13 142231
pwe@ida.liu.se (internet)
pwe@seliuida (bitnet)