rick@cs.arizona.edu (Rick Schlichting) (04/12/91)
[Dr. David Kahaner is a numerical analyst visiting Japan for two-years
under the auspices of the Office of Naval Research-Asia (ONR/Asia).
The following is the professional opinion of David Kahaner and in no
way has the blessing of the US Government or any agency of it. All
information is dated and of limited life time. This disclaimer should
be noted on ANY attribution.]
[Copies of previous reports written by Kahaner can be obtained from
host cs.arizona.edu using anonymous FTP.]
To: Distribution
From: D. Kahaner ONR Asia [kahaner@xroads.cc.u-tokyo.ac.jp]
Re: Fuzzy Logic (George Klir report)
12 April 1991
ABSTRACT. A report by George Klir (SUNY Binghamton) on Fuzzy Logic
research in Japan, based on a visit during Feb 1991.
INTRODUCTION.
Professor George Klir
Dept of Systems Science
Thomas J. Watson School
SUNY at Binghamton
Binghamton, NY 13901
Tel: (607) 777-6509, Fax: (607) 777-4000
Email: CJOSLYN@BINGVAXU.CC.BINGHAMTON.EDU
spent two weeks in Japan assessing Fuzzy Logic research and attending
the Biomedical Fuzzy System Congress in Tokyo. His report, which
includes some additional tutorial material, will appear in a forthcoming
Scientific Information Bulletin, published in my office. The less
mathematical portions are reproduced below along with some additional
comments that I have added. Readers should also note earlier reports
[fuzzy, 24 August 1990], and [fuzzy.gmd]. These provide references to
several of the same projects, people and products. In many cases
complete mailing addresses are given for the researchers.
"JAPANESE ADVANCES IN FUZZY THEORY AND APPLICATIONS"
George J. Klir
AN HISTORICAL OVERVIEW
Fuzzy theory emerged in the second half of this century by
challenging basic assumptions of three classical theories: the
assumption of sharp boundaries in classical set theory; the assumption of
classical (Aristotelian) logic that each proposition must either be true
or false; and the assumption of additivity in classical measure theory
and, in particular, probability theory. The first challenge came from
fuzzy set theory founded by ZADEH in 1965 (Ref. 24) even though some key
ideas of the theory were envisioned by BLACK in 1937 (Ref. 2). The
second challenge came from fuzzy logic, which emerged as an outgrowth of
fuzzy set theory (Ref. 6) as well as a generalization of the LUKASIEWICZ
infinite-valued logic defined on the unit interval (Ref. 11). The third
challenge came from fuzzy measure theory founded by SUGENO in 1974 (Ref.
14), even though the basic ideas of fuzzy measures, monotonicity and
continuity, were already present in CHOQUET capacities introduced in 1953
(Ref. 3).
In its initial stage, fuzzy theory encountered a lot of skepticism
and, on some occasions, open hostility (Ref. 27). In spite of the
opposition, the development of fuzzy theory became quite strong in the
1970s. New important concepts were introduced such as fuzzy numbers,
fuzzy topology, and various kinds of fuzzy relations. An extension
principle was introduced in 1975 (Ref. 26) by which other concepts and
theories of classical mathematics can readily be fuzzified. Operators
for aggregating fuzzy sets were investigated in a comprehensive way,
fuzzy sets of more general types were introduced, a theory of dynamic
fuzzy systems was developed, and various categories of fuzzy sets and
relations were formulated in category-theoretic terms. All these
advances influenced the development of fuzzy logic. For example, fuzzy
arithmetics is crucial for dealing with fuzzy quantifiers, fuzzy
operations of complementation, union, and intersection can readily be
mapped into the corresponding logic operations of negation, disjunction,
and conjunction, and fuzzy relation equations play an important role in
implementing fuzzy rules of inference.
Fuzzy measure theory was also significantly advanced in the 1970s.
In particular, several theories that generalize or complement probability
theory were introduced during this decade. They include probability
theory of fuzzy events (Ref. 25), random set theory (Ref. 8), theory of
Sugeno ~-measures (Ref. 14), Dempster-Shafer theory of evidence (Ref.
12), and possibility theory (Ref. 4).
Some ideas of prospective applications of fuzzy theory also emerged
in the 1970s. For example, fuzzy control (Ref. 15), fuzzy decision
making (Ref. 28), and fuzzy pattern recognition (Ref. 1). These ideas
were still "halve-baked" and of interest almost exclusively to the
academic community only. Industries, business, and government showed
little interest in this new area. In spite of this general lack of
interest, fuzzy theory continued to advance rapidly, as documented by the
need for a specialized journal, Fuzzy Sets and Systems, which was
established in 1978. Applications of the theory, however, were
hopelessly behind the theory itself.
The situation gradually changed in the 1980s, primarily to four
factors: (i) the theory became sufficiently mature; (ii) several
organizations promoting the theory and applications emerged (e.g., North
American Fuzzy Information Processing Society (NAFIPS) in 1981,
International Fuzzy Systems Association (IFSA) in 1984, Japan Society for
Fuzzy Theory and Systems (SOFT) in 1989); (iii) the theory became
recognized as a respectable academic subject by a growing number of
academic programs, where courses covering various aspects of the theory
and applications were initiated, which, in turn, contributed to the
publication of first textbooks in this area; and, above all, (iv) some
applications of fuzzy theory because sufficiently appealing to attract
the attention of industries and other non-academic constituencies.
The increasing interest in fuzzy theory and applications during the
1980s was most pronounced in Japan, a country that is now the
undisputable leader in this area. The United States, unfortunately, is
not only far behind Japan (especially in applications), but also behind
China, Soviet Union, and a host of other countries.
Research in fuzzy theory in Japan started early. Although it was
initially pursued only by a small group of researchers, these were
renowned scholars who managed to obtain a stronger support for the
research than it was possible in other countries. It is important to
notice that these scholars paid considerably more attention to
applications of fuzzy theory than their colleagues in other countries.
A genealogical tree of Japanese researchers and a list of early
papers by Japanese (1968-1971) in the area of fuzzy theory was prepared
by ZADEH (Ref. 27). The following is a chronological list of more recent
major events regarding the development of fuzzy theory and applications
in Japan:
1985, Japanese Chapter of IFSA was founded; it organized the first of
annual meetings called Japan Fuzzy Symposia.
1987, Hitachi successfully implemented fuzzy control of the subway
system in Semdai City in Northern Japan.
1987, Second IFSA Congress was held in Tokyo.
1988, Japan Agency of Science and Technology began to study "Expected
Fields for Fuzzy Engineering".
1988, IFSA Workshop on Applications of Fuzzy Systems was held in
Iizuka.
1989, SOFT was established and started to publish its own quarterly
journal; in February 1991, the Society had over 1,700 members,
60% of whom were engineers from more than 100 companies.
1989, Biomedical Fuzzy System Association was founded at Kawasaki
Medical University in Kurashiki.
1989, Laboratory for International Fuzzy Engineering Research (LIFE)
was initiated with a budget of approximately 40 million dollars
for 6 years; the support comes from the Ministry of
International Trade and Industry (MITI) and a consortium of 48
companies; the Laboratory has a staff of 25 researchers.
1989, Japan Agency of Science and Technology initiated a National
Project on Fuzzy Systems and Their Applications to Human and
Natural Problems with a budget of approximately 4 million
dollars for 5 years.
1990, Fuzzy Logic systems Institute was established as a private
research foundation supported initially by 14 companies.
1990, International Conference on Fuzzy Logic and Neural Networks was
held in Iizuka resulting in a publication of 1,236 pages of
Proceedings and Tutorials.
1991, Center for Promotion of Fuzzy Engineering was established at
the Tokyo Institute of Technology.
1991, First International Congress of Biomedical Fuzzy Systems was
held in Tokyo.
1991, International Fuzzy Engineering Symposium is scheduled in
Yokohama.
CURRENT RESEARCH ON FUZZY THEORY AND APPLICATIONS IN JAPAN
It is well known that research on fuzzy theory and applications has
been given high priority in Japan for the last few years by the
government as well as most industries. The motivation to support
research in this area quite generously is likely an outcome of some
highly successful industrial applications of fuzzy control in Japan in
the late 1980s. The first significant application of fuzzy control was
the automatic drive fuzzy control system for subway trains in Sendai
City. Although it took seven years to complete this project, the final
product was extremely successful. It is generally praised as superior to
other comparable systems based on classical control. The fuzzy
controller achieves not only a higher precision in stopping at any
designated point (to within 7 cm), but makes each stopping more
comfortable by lowering the frequency and speed adjustments. In
addition, it saves about 10% of energy.
Many other industrial projects that employ fuzzy control have been
completed in Japan since the opening of the subway system in Sendai City
in 1987. A complete list would be too long. The following are just a
few representative examples to illustrate the great utility of fuzzy
control:
-- chlorine controller for water purification plants
-- elevator control systems
-- traffic control systems
-- control of bulldozers
-- air conditioning systems
-- control systems for cement kilns
-- control of working machines, vacuum cleaners, video cameras,
refrigerators, etc.
[INSERT FROM D. KAHANER. Mr. Hideyuki Takagi, Central Research
Laboratories, Matsushita Electric Industrial Co., Ltd., provided me with
details of two new fuzzy products.
(1) TV meeting systems
Mitsubishi Electric Cooperation announced that their new TV meeting
systems are on the market as of April 1, 1991. They give very little
explanation of how fuzzy logic is applied, but apparently this decides
the degree of image movement and this is used in the decision about codec
(code-decode) variables. An ordinary TV system uses thirteen images per
second; this system uses ten images per second. Ten images/sec is
normally enough for static or slow movement of images, but when there is
a great deal of movement the resulting image is unnatural. Fuzzy logic
detects the degree of movement and changes the codec variables, so the
image quality of ten images/sec is better than that of thirteen. In
fact, a newspaper reported that fuzzy logic in this system chooses the
best values of ten parameters in deciding image quality according to the
image movement, and that the image quality is about equal to a
conventional system at twenty frames/sec. Though the concrete method of
fuzzy logic implementation may be different, the basic philosophy seems
to be quite similar to the movement detection system in a Panasonic
camcoder.
Specs: MELFACE810: 56kbps - 384 kbps, MELFACE850: 64kbps - 1536 kbps
Price: 7.3 million Yen - 19.5 million Yen
One site can have a meeting with up to 8 sites bidirectionally, and up to
64 sites can get the meeting image at the same time unidirectionally.
This system has 67% body size and 70% weight of ordinary systems because
of new seven codec ICs.
Takagi remarks that generally there appear to be two current features in
the Japanese consumer field. One is to adopt not only fuzzy logic but
also neural networks, for example as seen in products from Matsushita,
Hitachi, and Sanyo. The other is that application into AV (audio &
visual) equipment is increasing, for example quality control of TV image,
auto-focus and auto-iris, code compression, auto-stabilizer, auto-volume
control of car audio, color correction of printer, etc.
(2) Fuzzy & neuro electric fan
Sanyo Electric (overseas brand name is Fisher) put a new electric fan
on the market in March 1991. Its most interesting function is that it can
find the direction of its remote infrared master. Fuzzy reasoning infers
the infrared strength of the remote controller, and neural networks
decide angle using three sensors and estimated strength of ultrared. Thus
the electric fan can find out the position of user and send cool wind in
that direction.
END OF INSERT.]
Most of these products are well publicized. Extensive lists of products
that employ fuzzy control can be found, for example, in special issues of
the Japanese magazines Trigger (July 1989) and Quark (March 1991). Less
information is available about current research in the area of fuzzy
theory, which is perhaps less conclusive at this time, but involves
greater long term implications. The following is an overview of current
activities at several key organizations involved in research on fuzzy
theory and applications. The information is based on personal visits of
these institutions in February 1991.
Laboratory for International Fuzzy Engineering Research (LIFE)
LIFE seems to be currently one of two principal centers devoted
fully to research on fuzzy theory and applications. The other center is
the Fuzzy Logic Institute in Iisuka, Fukuoka.
LIFE was founded on March 28, 1989, with the purpose "to vitalize
fuzzy theory basic study, research on it efficient utilization by
strengthening ties between industrial and academic circles, and to
promote international technological exchange." It is located in Yokohama
and headed by DR. TOSHIRO TERANO, Professor at Hosei University, who is
one of the earliest and most important contributors to fuzzy theory in
Japan.
Since its very beginning, the principal aim of LIFE has been to
study comprehensively the many issues associated with human friendliness
of machines. It was early realized that there are two principal
requirements every user-friendly machine must satisfy: it must be
sufficiently intelligent and its communication with user must be smooth.
It is believed that fuzzy theory is capable of contributing significantly
to achieving these requirements.
After an initial stage, during which the focus was on formulating
specific research projects compatible with the overall aim, LIFE settled
on 9 projects, which are now organized under 3 research groups. The
three groups are: Decision Support Group, Intelligent Robot Group, and
Fuzzy Computing Group.
Projects under Decision Support Group are oriented to the
investigation of intelligent support systems for dealing with various
problems involving large-scale systems models. In the first project, the
aim is to investigate fuzzy expert systems that can deal not only with
all types of numerical or statistical data, but also with various types
of news data expressed in natural language. As a concrete theme, it was
chosen to develop a prototype of foreign exchange support system by which
macroscopic predictions of exchange rates can be made on the basis of
market participants and economic situation. The second project is aimed
at the development of intelligent support systems for human plant
operators. A power plant was chosen as a specific testing ground for the
project. The third project is oriented to the study of anticipating
control systems based upon fuzzy dynamic models.
The Intelligent Robot Group consists of three projects whose aim is
to develop a robot that combines sophisticated visual perception
capability with the capability of understanding natural language. The
ultimate goal is to design a home robot. Such a robot must be much more
adaptable to changes in environment than common industrial robots and
must also be able to understand intentions of its owner. The first
project is concerned with the issues involved in natural language
understanding, such as the ambiguity and vagueness inherent in natural
language, the dependence of meaning on context, the intention of the
language producer, and the ability to apply the experience to present
situations. This is clearly an extremely challenging project, in which
fuzzy logic plays an essential role. The aim of the second project is to
develop high level visual perception capabilities. The research is
centered on the knowledge-based model description of objects and
appropriate reasoning methods to deal with the model and actual image
data. The third project is concerned with the overall capability of
robots to make intelligent decisions and the required control skills to
resemble the behavior of human beings. The primary foci are on sensor
fusion, global path planning, and decision making for autonomous motion.
The orientation of projects under the Fuzzy Computing Group is the
investigation of various aspects of computer systems from the standpoint
of fuzzy theory. The first project focuses on fuzzy neural networks.
The aim is to combine advantages of fuzzy logic and neural network
technology. This idea, which was originated by BART KOSKO in the U.S.
(Ref. 7), is now far more rapidly being developed in Japan (not only at
LIFE) than in the U.S. The goal of the second project is to develop a
fuzzy expert shell called "LIFE FESHELL" for building fuzzy expert
systems associated with the other projects, such as the foreign exchange
support system or the image understanding system. The third project
focuses on the development of a computer specifically designed for fuzzy
information processing. The project will result in specifications for
the entire architecture of the computer and the necessary hardware and
software technology to implement the architecture. Research on software
involves the design of a language in which fuzzy sets and operators on
fuzzy sets can be easily implemented, aiming at a software development
support system capable of flexible fuzzy information processing.
Research on hardware involves the development of a high-speed fuzzy set
operation chip as well as the study of how parallel processing can be
best utilized in the fuzzy computer.
Fuzzy Logic Systems Institute (FLSI)
FLSI, located in Iizuka (Southwest of Japan), was established one
year after LIFE, on March 15, 1990, with the following aim: "To offset
the disadvantages of existing deterministic methods of information
processing by conducting experimental research into fuzzy information
processing and neuroscience, and to promote the wider use of fuzzy
information processing and neuroscience." The Institute is headed by Dr.
TAKESHI YAMAKAWA, Professor of Kyushu Institute of Technology, who is
well known for his pioneering work on hardware technology implementing
fuzzy inference rules for fuzzy controllers.
Since FLSI was less than one year old when I visited it, its program
was not as well defined and final as the program of LIFE. Nevertheless,
the following areas seem to form the primary orientation of the
Institute:
1. Research and development of high-speed hardware to support fuzzy
logic. This is a natural outgrowth of previous work of Professor
YAMAKAWA. In his publications (e.g., Refs. 18-22) an electronic circuit
is described where speed is 10 Mega FIPS (Fuzzy Inferences Per Second).
The ultimate goal is to use the hardware developed at FLSI for designing
and building a fuzzy computer (in cooperation with LIFE), which will be
capable of using rules of fuzzy logic at very high speed. This part of
computer is usually referred to in Japan as the sixth generation
computer.
2. Research on fuzzy neural networks, which involves relevant
theory, hardware development, and various applications. Currently, the
main focus seems to be on hardware development of an artificial fuzzy
neuron, a neuron in which the weights of synaptic junctions are
represented by fuzzy numbers rather than ordinary numbers. Although
research on fuzzy neural networks is currently very active in Japan and
not restricted to FLSI, the latter plays undoubtedly a leadership role in
this area. This was exemplified by its principal sponsorship of the
first International Conference on Fuzzy Logic and Neural Networks (July
22-24, 1990), which happened to be a very successful event (Ref. 10).
Among the applications of fuzzy neural networks that are currently
researched at FLSI is pattern recognition of handwritten characters. A
combination of designability of neural networks with their usual learning
capabilities is explored. At this time, a hardware system for
handwritten character recognition was implemented by one layer of fuzzy
neurons, with each neuron designed for recognizing one particular
character, whose speed is 10 microseconds per character recognition.
3. Research on fuzzy control systems. Emphasis is on systems in
which the needed inference rules are implemented in hardware, which
itself is also studied and developed at the Institute. A good example of
the effectiveness of fuzzy control, associated with FLSI, is a fuzzy
controller designed to stabilize an inverted pendulum. The performance
is excellent even under sever noise such as the movement of mice placed
in a container atop the pendulum or pouring water into the container. It
is amazing that only 7 fuzzy inference rules are needed to achieve such a
high performance.
4. FLSI is also increasingly getting involved in research of
various aspects of biomedical fuzzy systems, as documented by its major
role in organizing, jointly with the Biomedical Fuzzy Systems
Association, the First International Congress of Biomedical Fuzzy Systems
in Tokyo (February 13-15, 1991). A list of papers and their authors is
given at the end of this report.
The congress, which was preceded by two-day tutorials (organized
also by FLSI) on fuzzy theory for medical doctors, showed that fuzzy
theory can offer a great deal to the medical profession and the medical
professionals are becoming to recognize this potential. Topics that were
particularly well covered at the congress include fuzzy expert systems in
the various area of medicine, the role of fuzzy neural networks in the
organization of medical knowledge, and the role of fuzzy control in
medicine. My expectation is that the importance of biomedical fuzzy
systems will grow rapidly in Japan within the next few years and FLSI
will undoubtedly play a major role in this development.
Tokyo Institute of Technology
The center of research on fuzzy theory and applications at the Tokyo
Institute of Technology (TIT) is referred to as Sugeno Laboratory.
Associated with the Dept. of Systems Science in Yokohama, the Laboratory
is headed by Professor MICHIO SUGENO, well known for his pioneering work
on fuzzy measures, fuzzy integral, and fuzzy control. The current
research at the Laboratory involves the following areas:
1. Fuzzy measure theory and fuzzy integral. Professor SUGENO
himself originated the concepts of a fuzzy measure and fuzzy integral in
his dissertation at TIT in 1974. He and his group have been active in
this area over the years. Current work focuses on investigating the
concept of fuzzy t-integral, which subsumes three types of fuzzy
integrals as special cases, Choquet integral, Sugeno integral, and Weber
integral.
2. Sugeno laboratory has been a leader in the area of fuzzy control
(Ref. 15). It currently focuses on two projects in this area: the
design and construction of a microprocessor-based fuzzy controller for
general purposes, and the design of a fuzzy controller for unmanned
helicopter. The latter project, which is supported by the Ministry of
Transportation, is particularly challenging. A helicopter is a highly
unstable object, especially under rapidly varying wind conditions. To
model its behavior adequately requires to consider 15 input variables and
4 output variables, which are strongly interrelated. Attempts to design
a classical model-based controller for this purpose have not been
successful thus far. Fuzzy control, on the other hand, seems to work
quite well, at least on the basis of simulation experiments. I saw
results on video of some simulation experiments under various wind
conditions and for various remote control oral instructions (fly
straight, turn left, hover, loud, etc.). The performance was very
impressive. Experiments with a real helicopter are scheduled for March
1992. Compared in these experiments will be fuzzy control and
conventional control, the latter being developed independently at another
laboratory. This will be an important test which, I suspect, will
demonstrate the superiority of fuzzy control in this and similar
applications, which are characterized by high instabilities,
nonlinearities, and time-varying conditions. The helicopter flight
control project is a two-year project, which was initiated in April 1990.
It consists, in fact, of these subprojects: (1) to develop a remote
control of a helicopter by oral instructions; (ii) to develop a control
for automatic autorotation entry and landing in case of engine failure;
and (iii) to develop a full control of a helicopter for sea rescue
operations based on control instructions from a mothership and
information from a satellite about the position of the helicopter.
3. Sugeno laboratory is also involved in researching some problems
associated with the development of a fuzzy computer. This research,
which is performed in cooperation with LIFE, consists of two projects:
(a) Linguistic modelling of images using fuzzy case-based reasoning. The
aim of this project is to develop fuzzy logic technology for high-level
image understanding similar to that of humans.
(b) Analysis of natural language in the context of the prospective fuzzy
computer. The principal aim is to develop methods for linguistic
modelling and simulation based upon both numerical and linguistic data.
Hosei University
One of the most active academic groups in the area of fuzzy theory
and applications in Japan is housed at the Department of Measurement and
Control of the Hosei University in Tokyo. This is a result of long-term
leadership of Professor TOSHIRO TERANO (currently the Director of LIFE)
and more recent leadership of Dr. KAORO HIROTA. The group is primarily
involved in three areas:
1. Great variety of applications of fuzzy control have been
investigated, for example, the tracing of a randomly moving object,
control of a yo-yo, semi-automatic control of a bulldozer, control of a
helicopter, control of a double inverted pendulum and, more recently,
even a triple inverted pendulum. Based on this extensive experience with
variety of applications of fuzzy control, the group came to the
conclusion that fuzzy control is very cost effective, robust, and easily
implementable even in dealing with processes that involve nonlinearities,
instabilities, and varying conditions. In addition to the work on
applications of fuzzy control, the group is also involved in theoretical
research regarding the problem of how to analyze stability of fuzzy
control systems (for example, how to identify fuzzy control rules that
effect stability). Although fuzzy control has been very successful in
practice, its theory is still in its infancy. It is generally recognized
that appropriate stability analysis for fuzzy control systems is
currently the most needed component of the emerging theory.
2. Fuzzy pattern recognition and image processing. One problem in
this category, which was investigated in the late 1980s, was the problem
of recognizing crops by fuzzy logic. Results obtained by working on this
particular problem are now being utilized in other fields. In the area
of image processing, focus is on linguistic description of scenes. Such
a description can capture not only characteristics of objects on the
scene and their relationships, but also the season (winter, summer,
etc.), time (early morning, noon, etc.), and other characteristics like
these. Since 1989, the group has also investigated the combination of
neural networks and fuzzy techniques for dealing with the problem of
image reconstruction. Research on 2-dimensional image reconstruction
based on this approach was apparently completed at the time of my visit
(February 1991); currently, they work on 3-dimensional image
reconstruction.
3. The group is particularly strong in applications of fuzzy logic
to robotics. In fact, most of the work on fuzzy control, pattern
recognition, and image processing within the group has been motivated by
its utility in robotics. The goal is not to develop a universal, human
friendly robot ( in the sense of the robotics project at LIFE), but
rather to implement special, highly complicated (and, in some instances,
unusual) skills of humans by robots. The following are some of these
implementations (all based upon fuzzy logic), which has recently been
completed: a robot playing two-dimensional ping-pong; a robot capable of
throwing arrows to a target in the same way as humans do, Japanese flower
arrangement by a robot equipped with knowledge of a human expert (encoded
in eight fuzzy inference rules); inspection and evaluation of carnation
seedlings in a plant factory with the aim of deciding for each seedling
whether it is sufficiently healthy for planting or should be discarded
(the results agree very well with judgements made by skilled inspectors
from whom the knowledge was elicited and encoded in appropriate fuzzy
inference rules involving shapes, colors and other characteristics of the
inspected seedlings); and Chinese calligraphy by a robot (painting
Chinese characters by a brush, which in Japan and China is considered as
an art). The problem of learning by robots based on fuzzy models has
also been investigated. In analogy to Bayes' theorem of conditioning in
probabilistic models, fuzzy integral and conditional fuzzy measures play
similar role in the context of fuzzy models.
In addition to the described research, the group also developed a
computer-aided instruction system to teach engineers in various
industries fundamentals of fuzzy theory and its existing and potential
applications. In one year, almost a thousand copies of the system were
purchased by various companies. This indicates, again, how strong the
current interest in fuzzy theory is in Japan.
Research in Osaka
Until 1971, Osaka had been the only place in Japan (according to
known publications) where research on fuzzy set theory took place. Both
of the major universities in Osaka were involved at this early stage, the
Osaka University and the University of Osaka Prefecture. Since Osaka is
an industrial city with a long tradition in strong cooperation between
universities and industries, it is not surprising that some industries in
Osaka have been influenced for more than two decades by research on fuzzy
theory at the local universities.
The strong cooperation between universities and industries was
clearly exhibited during my visit to Osaka. Prior to my visit, my only
acquaintances in Osaka were at the universities. Instead of arranging a
meeting at one of the university, they proposed, quite thoughtfully, to
have a joint meeting at Central Research Laboratories of the Matsushita
Electric Industrial Company (MEIC).
MEIC is a comprehensive electronics manufacturer. The primary task
of its research laboratories, particularly its Intelligence Sciences
Laboratory, is to develop products that are human friendly. MEIC is one
of the founding members of LIFE.
The Intelligence Sciences Laboratory of MEIC, senior
researchers of which I met during my visit, is currently involved
in research pertaining to the following areas:
-- speech synthesis and recognition
-- text, graphic, and image recognition
-- 3-dimensional image processing
-- fuzzy data processing
-- multi-stage reasoning and judgement
-- fuzzy data and neural networks
-- hypermedia
-- information structuring and classification
-- multi-media conversion and integration
Fuzzy theory is involved in virtually all of these areas.
MEIC has already developed and currently manufactures a number of
consumer products that use fuzzy control. They include washing machines,
refrigerators, air conditioning systems, vacuum cleaners, kerosine
heaters, microwave ovens, hot and cold water mixing units, and video
cameras. The fuzzy control automatic washing machine, for example,
senses the quantity of work load, the fabric type, the intensity and type
of dirt, and both the room and water temperature. Based on these sensory
data, it adjusts the wash, rinse, and spin cycle times automatically on
the basis of six fuzzy inference rules that adequately capture knowledge
of an experienced operator.
A fascinating new product of MEIC is a video camera that not only
adjust zoom and flash automatically, but has also an image stabilizer
that significantly reduces the movements of the image caused usually by
shaking hands of the user. The image stabilizer compares pictures taken
at two sufficiently close time instants and, using appropriate fuzzy
inference rules, makes a judgement of whether any recognized change in
the pictures is due to a moving object in the scene or due to a movement
of the camera itself. This judgement is then employed for choosing an
appropriate corrective action. The performance of the fuzzy stabilizer
is outstanding. [This camera has impressed all those who have seen it.
Also see the remarks by Takagi earlier. DKK]
As any other institution I visited in Japan, MEIC is also heavily
involved in research combining fuzzy theory and neural networks. Their
principal interest in fuzzy neural networks is to employ them for
determining appropriate membership grade functions (by learning from
input-output data) for new products.
CONCLUDING REMARKS
There is no doubt that fuzzy theory is currently well respected and
supported in Japan. In the U.S., on the contrary, it is still regarded
largely with suspicion or even hostility. As a result, Japan is far
ahead not only in the theory itself, but above all in its practical
utilization.
Fuzzy control is currently the most successful application area of
fuzzy theory in Japan, but successes in other areas, such as image
processing, pattern recognition and robotics, are by no means negligible.
Fuzzy theory seems intimately connected with the notion of user
friendliness of machines and with the development of the sixth generation
of computers.
Why fuzzy theory and its applications are so successful in Japan? A
combination of three factors may give a reasonable, even though somewhat
speculative answer. The first factor seems to be the Japanese culture,
which is known to be much more receptive to vagueness and other types of
uncertainty than the various Western cultures. The second factor is
likely the early and rather strong support given to fuzzy theory by some
influential and highly respected Japanese scholars. In the U.S., on the
contrary, some highly influential scholars were quite hostile toward
fuzzy theory during the early stages of its development. The third
factor seems to be the Japanese talent for applications. Once the
pragmatic value of fuzzy theory was established by its successful
applications, it was easier to obtain support and that, in turn, helped
to further advance the theory itself and, at the same time, explore new
applications.
REFERENCES
1. Bezdek, J. C., Pattern Recognition with Fuzzy Objective Function
Algorithms, Plenum Press, New York, 1981.
2. Black, M., Vagueness: "An exercise in logical analysis." Philosophy
of Science, 4, 1937, pp. 427-455.
3. Choquet, G., "Theory of capacities." Annales de L'Institut Fourier,
5, 1953-54, pp. 131-295.
4. Dubois, D. and H. Prade, Possibility Theory. Plenum Press, New York,
1988.
5. Kaufmann, A. and M. M. Gupta, Introduction to Fuzzy Arithmetic. Van
Nostrand Reinhold, New York, 1985.
6. Klir, G. J. and T. A. Folger, Fuzzy Sets, Uncertainty, and
Information. Prentice Hall, Englewood Cliffs, N.J., 1988.
7. Kosko, B., "Fuzzy associative memories." In: Fuzzy Expert Systems,
ed. by A. Kandel, Addison-Wesley, Reading, Mass., 1987.
8. Matheron, G., Random Sets and Integral Geometry. John Wiley, New
York, 1975.
9. Moore, R. E., Methods and Applications of Interval Analysis. SIAM,
Philadelphia, 1979.
10. Proceedings of International Conference on Fuzzy Logic of Neural
Networks, Kyusu Inst. of Technology, Iizuka, Fukuoka, Japan, July
22-24, 1990 (2 volumes and tutorials).
11. Rescher, N., Many-valued Logic. McGraw-Hill, New York, 1969.
12. Shafer, G., A Mathematical Theory of Evidence. Princeton University
Press, Princeton, N.J., 1976.
13. Sims, J. R. and Z. Wang, "Fuzzy measures and fuzzy integrals: an
overview." Intern. J. of General Systems, 17, Nos. 2-3, 1990, pp.
157-189.
14. Sugeno, M., Theory of Fuzzy Integrals and Its Applications. Ph.D.
Thesis, Tokyo Institute of Technology, 1974.
15. Sugeno, M. (ed.), Industrial Applications of Fuzzy Control, North-
Holland, New York, 1985.
16. Wang, Z. and G. J. Klir, Fuzzy Measure Theory. Plenum Press, New
York, 1991.
17. Yager, R. R., S. Ovchinnikov, R. M. Tong, and H. T. Nguyen (eds.),
Fuzzy Sets and Applications: Selected Papers by L. A. Zadeh.
Wiley-Interscience, New York, 1987.
18. Yamakawa, T., "High-speed fuzzy controller hardware system."
Information Sciences, 45, No. 2, 1988, pp. 113-128.
19. Yamakawa, T., "Intrinsic fuzzy electronic circuits for sixth
generation computer." In: Fuzzy Computing, ed. by M. M. Gupta and
T. Yamakawa, Elsevier, 1988, pp. 157-171.
20. Yamakawa, T., "Stabilization of an inverted pendulum by a high-speed
logic controller hardware system." Fuzzy Sets and Systems, 32, No.
2, 1989, pp. 161-180.
21. Yamakawa, T., "High-speed fuzzy controller hardware system: the
Mega-FIPS machine." Information Sciences, 45, No. 2, 1988, pp. 113-
128.
22. Yamakawa, T. and K. Hirota (eds.), Special Issue of Fuzzy Sets and
Systems on "Applications of Fuzzy Logic Control to Industry." 32,
No. 2, 1989, pp. 137-244.
23. Yen, J., "Generalizing the Dempster-Shafer theory to fuzzy sets."
IEEE Trans. on Systems, Man, and Cybernetics, 20, No. 3, 1990, pp.
559-570.
24. Zadeh, L. A., "Fuzzy sets." Information and Control, 8, No. 3,
1965, pp. 338-353.
25. Zadeh, L. A., "Probability measures of fuzzy events." J. Math.
Analysis and Applications, 23, 1968, pp. 421-427.
26. Zadeh, L. A., "The concept of a linguistic variable and its
application to approximate reasoning." Information Sciences, 8,
1975, pp. 199-249, 301-357; 9, pp. 43-80.
27. Zadeh, L. A., "Birth and evaluation of fuzzy logic--expectation of
Japan's role." J. of Japan Soc. for Fuzzy Theory and Systems, 2, No.
2, 1989, pp. 182-200.
28. Zimmermann, H. J., Fuzzy Sets, Decision Making and Expert Systems,
Kluwer, Boston, 1987.
BMFS-TOKYO '91
Proceedings of the International Congress of Biomedical Fuzzy Systems &
the Third Annual Meeting of Biomedical Fuzzy Systems Association
The ABC Kaikan Hall - Tokyo, Japan
February 13-15, 1991
Special Lectures
Biomedical Application of Fuzzy Neural Networks from Medical
Diagnosis to Pattern Recognition
E. Sanchez (Faculty of Medicine, University of Marscille, and
Neurinfo Research Department, IMT, Marscille, France)
Measures and Principles of Uncertainty
G.J. Klir (Department of Systems Science, Thomas J. Watson
School, State University of New York, Binghamton, New York
13901, U.S.A.)
Knowledge Representation
Knowledge Representation in Neural Network and Fuzzy Logic
K. Yoshida, (Dept. of Preventive Medicine and Public Health,
School of Medicine, Keio University)
Fuzziness in the Cortical Column Model
F. Blanc, E. Sanchez (Neurinfo Research Department, IMT,
Marseille, France)
A Clinical Report and Retrieval System Based on Fuzzy Set Theory
M. Jamzad, A. Uchiyama (School of Sci. & Eng. Dept. of Elec. &
Comm. Waseda Univ. Tokyo)
H. Toyama (Tokyo Metropolitan Institute of Gerontology)
Expression of a Multiple Fuzzy Set Defined on Dependent Coordinates
System and Reducing Transformation to COSMOS-DIAGRAM
T. Yanaru, T. Hirota (Department of Artificial Intelligence,
Kyushu Institute of Technology, Iizuka, Fukuoka, Japan)
Handling Fuzzy Items in Diagnostic Knowledge Bases
H. Bandemer (Freiberg Mining Academy, Germany)
Decision Making
Uncertainties in Estimating Dose-Effect Relationships Under
Emergency Conditions
Y. Nishiwaki (University of Vienna. Inst. for Medical Physics)
C. Preyssl (European Space Agency, Noordwijk, Netherlands)
Optimization of Radiation Protection Based on Fuzzy Risk Analysis
C. Preyssl (European Space Agency, Product Assurance and Safety
Department, 2200AG Noordwijk, The Netherlands)
Y. Nishiwaki (University of Vienna, Faculty of Medical Physics,
1010 Vienna, Austria)
Medical Application of Fuzzy Decision Making
K. Yoshida (School of Medicine, Keio University)
E. Tazaki (Dept. of Systems and Control, Tohin Yokohama
University)
The Basic Consideration About the Relationship Between Doctors'
Diagnostic Rules and Membership Functions
T. Takahashi, T. Imamura, S. Kaihara (Hospital Computer Center,
University of Tokyo Hospital)
Diagnosis (I)
Differential Diagnosis of Gastric Cancer and Gastric Ulcer Using a Fuzzy
Inference
K. Saito, A. Uchiyama (School of Sci. & Eng. Dept. of Elec. & Com.
Waseda Univ., Tokyo, Japan)
H. Hashimoto (Tokyo Women's Medical College, Tokyo, Japan)
An Automatic Skin Allergy Test Using Fuzzy Image Processing
M. Park, B.S. Jeong & M.G. Lee (Department of Electronics
Engineering, Yonsei University, Seoul, Korea)
Application on Fuzzy Expert System for Orthodontic Diagnosis
H. Kawagoe, T. Hirose, T. Itoh, M. Matsumoto (Department of
Orthodontics, Fukuoka Dental College)
An Application of Fuzzy System to Orthodontic Diagnosis in Orthosurgical
Case
K. Tsuji (Laboratory of Medical Information, The Jikei
University School of Medicine)
Y. Yoshikawa (Department of Orthodontics, Matsumoto Dental
College)
Demonstration
Fuzzy Measuring from Blurred Micro Pictures (Short Message)
H. Bandemer, A. Kraut (Freiberg Mining Academy, Germany)
Diagnostic System for Diabetes Mellitus based on the Response of Glucose
Tolerance Test Using a Fuzzy Inference
S. Arita (Department of Mathematics, Kawasaki Medical School)
M. Yoneda (Department of Endocrinology, Kawasaki Medical School)
Y. Hori (Computer Center, Kawasaki Medical School)
Diagnosis (II)
Description and Validation of PNEUMON-LA: A Medical Expert System for
Pneumonia Diagnosis
R. Lopez de Mantaras, C. Sierra (C.S.I.C. Centre d'Estudis Avancats
de Blanes, Spain)
A. Verdaguer, A. Patak, F. Sanz (Dept. of Biomedical Computer
Sciences, I.M.I.M., Barcelona, Spain)
A. Verdaguer (Department of Internal Medicine, L'Alianca Mataronina,
Mataro, Spain)
Application of Fuzzy Measure in Classification of Childhood Bronchial
Asthma
H. Inada, T.J. Kim, S. Kusuda, T. Sugahara (Osaka Children's
Hospital for Allergic Disease)
K. Nishimaki, S. Niihira, S. Nakajima, G. Isshiki (Osaka City
University School of Medicine, Pediatric Department)
Diagnostic System for Diabetes Mellitus based on the Response of Glucose
Tolerance Test using a Fuzzy Inference
S. Arita (Department of Mathematics, Kawasaki Medical School)
M. Yoneda (Department of Endocrinology, Kawasaki Medical School)
Y. Hori (Computer Center, Kawasaki Medical School)
ROC Analysis Using Fuzzy Inference in Radiographic Screen-Film System
S. Ueda, Y. Murano (Department of Central Radiology, Kinki
University Hospital, Osaka)
T. Hamada (Department of Radiology, Kinki University School of
Medicine, Osaka)
S. Arita (Department of Mathematics, Kawasaki Medical School,
Kurashiki)
An Application of Fuzzy Inference to the Ultrasonographic Diagnosis of
Liver Cirrhosis
K. Hamabata, T. Shibue (2nd Dept. of Internal Medicine, Faculty
of Medicine, Kagoshima University)
S. Arita (Dept. of Mathematics, Kawasaki Medical School)
Application of Fuzzy Inference to the Diagnosis of Prostatic Cancer by
Transrectal Ultrasonography
T. Fujioka, H. Koike, T. Kubo, T. Ohhori (Department of
Urology, Iwate Medical University School of Medicine)
S. Arita (Department of Mathematics, Kawasaki Medical School)
T. Saito (Department of Pharmacology, Kawasaki Medical School)
Y. Hori (Computer Center, Kawasaki Medical School)
Application of Fuzzy Logic to Diagnoses of Ovarian Tumors with MRI
R. Takeda (Department of Obstetrics and Gynecology, Faculty of
Medicine, Kinki University)
K. Makino (Hospital Pharmacy, Faculty of Medicine, Kyushu
University)
H. Kawagoe (Department of Orthodontics, Fukuoka Dental College)
Control (I)
Fuzzy Control of Superconducting Actuators for Potential Biomedical
Applications
M. Komori, T. Kitamura (College of Computer Science & System
Engineering, Kyushu Institute of Technology, 680-4, Kawazu,
Iizuka City, Fukuoka, 820 Japan)
An Attempt to Regulate High Frequency Ventilation by Fuzzy Logic
M. Noshiro, K. Takakuda (Tokyo Medical and Dental University,
2-3-10 Kanda-Surugadai, Chiyoda-ku, Tokyo)
T. Fujino, J. Ikebe (Tokyo Engineering University, 1404-1
Katakura-machi, Hachioji, Tokyo)
The Fluid Replacement Program with the Spread Sheet Applying the Fuzzy
Theory for the Resuscitation of Severe Burn Patient
S. Wakamatsu, K. Imamura, T. Hirayama (Department of Plastic
and Reconstructive Surgency, Tokyo Women's Medical College)
M. Ishijima (Institute of Medical Engineering, Tokyo Women's
Medical College)
Control (II)
Alternative Mechanisms for Fuzzy Logic Control
R.R. Yager (Machine Intelligence Institute, Iona College, New
Rochelle, NY 10801, USA)
The Intelligent Controller for Physiological System - Application to
Blood Pressure Regulation
T. Masuzawa (Department of Artificial Organ, Research
Institute, National Cardiovascular Center, Japan
Y. Fukui (Department of Applied Electronic Engineering, Tokyo
Denki University, Japan)
M. Suzuki (Research Institute for Tuberculosis and Cancer,
Tohoku University, Japan)
Clinical Application of Fuzzy-logic Control of Blood Pressure through
Anesthesia
T. Tsutsui, S. Hamada (Department of Anesthesiology,
Shimonoseki Kousei Hospital)
S. Arita, Y. Hori (Department of Mathematics and Computer
Science, Kawasaki Medical School)
Categorization of Anesthesiologists Thinking Represented by Artificial
Intelligence and Comparison of Each Category (A Computer Control System
of Applying Anaesthesia Using Fuzzy Logic for Medical Operation)
T. Imamura, T. Takahashi, S. Kaihara (Hospital Computer Center,
University of Tokyo Hospital, 7-3-1 Hongo, Bunkyo, Tokyo 113
Japan)
T. Horiuchi, H. Okuda, M. Uchida (Department of Anaesthesia,
Kansai Medical University, 1 Funizonocho, Moriguchi, Osaka,
570 Japan)
Demonstration
Clinical Application of Fuzzy-logic Control of Blood Pressure through
Anesthesia
T. Tsutsui, S. Hamada (Department of Anesthesiology,
Shimonoseki Kousei Hospital)
S. Arita, Y. Hori (Department of Mathematics and Computer
Science, Kawasaki Medical School)
--------------------END OF REPORT------------------------------------------