hagemann@tyo.gmd.dbp.de (Thomas Hagemann) (05/14/91)
To: Distribution
From: Thomas Hagemann, GMD Bureau Tokyo (hagemann@tyo.gmd.dbp.de)
Subject: Visit to OMRONs Fuzzy Business Promotion Center, Kyoto
Date: 25 March 1991
Updated: 9 May 1991
ABSTRACT.
On 5 March 1991 I visited the Central Research and Development
Laboratory of OMRON Corporation in Kyoto to speak with members of the
fuzzy project team. My hosts have been
Kazuaki Urasaki
Tadasu Maeda
Fuzzy Technology Business Promotion Center
Omron Corporation
Shimokaiinji
Nagaokakyo-shi
Kyoto 617
Tel: +81-75-951-5111
Fax: +81-75-955-2442
It follows an overview of Omron Corp. and its fuzzy related activities,
a table of fuzzy related patent applications in Japan, and a look into
the special edition on "fuzzy" of the japanese popular science magazine
"Quark". For different aspects of fuzzy in Japan see the reports of
David K. Kahaners (24 August 1990), myself (27 December 1990) and George
J. Klir (12 April 1991).
COMPANY OVERVIEW.
Omron Corporation, founded as Tatisi Electric Manufacturing Company in
May 1933 by Kazuma Tateishi, has major business divisions for factory
automation systems, control components, automotive electronic
components, traffic control systems, electronic fund transfer systems,
health and medical equipment, and office automation systems. The number
of employees is about 14.000 including all subsidiaries, 10.000 of them
in Japan. Other figures of Omron Corp. (excluding subsidiaries), found
in the Spring/91-edition of a japanese handbook listing 2577 companies,
published by Toyokeizai-Shinposha, are
paid-in capital 38.5 BYen 200 MECU (185 Yen = 1 ECU in 9/90)
sales 350.7 BYen 1900 MECU
profit 27.1 BYen 150 MECU
investment 11.4 BYen 60 MECU
major shareholders Mitsubishi-, Sumitomo-, Taiyo-Kobe-Mitsui- Banks
employees 6957
average age 33.4 years
average salary 296.4 kYen 1600 ECU
The banks hold between 2.0 and 4.7 percent of the company's shares (5%
being the legal upper limit of share holding by financial institutions
according to paragraph 11 of the Anti-Monopoly Law), another 1.8% holds
chairman Takao Tateishi of the founders' family.
The Central R&D Laboratory with its 800 employees is one of Omron's
research facilities covering a wide range of technologies. Others are
the Communication Systems Laboratory and the Systems Laboratory (R&D in
computer systems) in Tokyo, the Life Science Laboratory in Kyoto (R&D in
medical and health-care systems), and the Tsukuba R&D Laboratory. Omron
has developed within and parallel to the SIGMA-project of MITI a UNIX-
workstation "LUNA".
FUZZY AT OMRON.
Although located in the Central R&D Lab, the Fuzzy Technology Business
Promotion Center is not part of that, but it is directly under the
management of the company president. It has one marketing and one R&D
section, gives support to fuzzy researchers in various business
divisions (there are more than 200 fuzzy researchers in Omron
altogether) by developing hardware and software tools, and it does fuzzy
R&D on its own, focussing mainly on development, since basic research of
fuzzy theory is said to be done mostly at universities. For the
researchers, who receive a three month training when entering the
company, there are regular weekly in-house seminars, sometimes with
lecturers from universities. At present, one researcher is sent to Osaka
University to attend the two year mastercourse. In general, as in all
other japanese companies, researchers often go to the factory, to the
sales department and to customers bringing back new ideas for their own
work.
Omron's major fuzzy products are used in factory and plant automation
and automotive control. The video I was shown presented, among others, a
temperature controller in a chemical plant, a vending machine selecting
cans by colour, a camera following moving objects, and a conveyor belt,
the speed of which was controlled such that boxes, coming randomly along
the belt, passed a fixed check point in constant time intervals.
The scope of fuzzy products ranges from digital fuzzy chips, a hybrid
controller (combining fuzzy and PID control), expansion boards for NEC-
PCs and IBM-PC/AT-compatibles (to be presented in April at Hannover
Fair, Germany), various controllers and control boards, and a fuzzy
human body sensor for detecting living objects in rooms, to software for
simulation and study on NEC-PCs and the LUNA-workstation, and a fuzzy
expert system for machine diagnosis. One of Omron's fuzzy devices, the
FZ-1000, is built into the prototype heavy water reactor "Fugen", where
it controls the water temperature in a tank. Under normal and stable
conditions the temperature is controlled by a PID-device, but when the
reactor starts firing or is shut down, PID-control is less suitable for
the huge amount of information to be handled during these phases,
therefore control is switched to fuzzy control under the monitoring of a
human operator. It is claimed that this still limited system is only
used in non-critical places. The Fugen reactor is operated in Fukui
prefecture, central Japan, by the Power Reactor and Nuclear Fuel
Development Corporation, a R&D-organization supervised by the Science
and Technology Agency (STA).
A digital fuzzy chip was put on the market last October for use in
automobiles and home electronics, and 100.000 of these chips are
expected to be sold in the first year. In 1990, only 0.5% of the above
listed sales figure of Omron will come from fuzzy products, but the
sales plan for 1994 is 100 BYen. This means that Omron expects a lot
from fuzzy technology in the future, and with respect to its importance
and potential, it is compared with microprocessor technology a couple of
years ago. Nevertheless it is thought that fuzzy technology cannot
replace traditional methods everywhere, although it is often easier and
cheaper to implement. For example, fuzzy control seemed to be superior
when precision of up to 1% is sufficient, but hybrid fuzzy/PID-control
is necessary for higher precision. Similarly, PID-control is better for
tasks like keeping temperature on a constant level, whereas fuzzy
control is better for reaching this level, because overshooting rarely
happens.
HISTORY OF OMRON'S FUZZY COMMITMENT.
Whenever you visit a japanese company and get an introduction to its
activities, you first will be presented the company philosophy and its
vision for the future. The past might also be mentioned, but the focus
lies on the future. Whether vision and philosophy are derived from the
company's products and business strategies, or vice versa, is not as
important as to have a philosophy at all, which is easy to understand
for the management, the employees and the customers, like "For Beautiful
Human Life" of a big cosmetics company. The aim of "Contributing to the
World Through Batteries" of the battery manufacturing subsidiary of a
huge electric company group makes sense when considering products like
walkman, mobile phones or notebook PCs, although the explanation of this
aim in a company brochure sounds somewhat far-fetched or even ridiculous
to the left brained Westener: "Adressing the changing needs of industry
and society, our focus is on the world. We are committed to opening
doors to a richer future, for the prosperity of our global society, and
the happiness of mankind through the creative use of battery power." On
the other hand, referring just to "synergy" as the motif for the
business activities, as big german companies like to do, might be
intellectually satisfying, but the normal customer will have
difficulties to find it on the shelves for sale.
Omron's philosophy fits perfectly into the company profile: "To the
machine the work of machines, to man the thrill of further creation."
Omron was always in search of technologies, which helped to free man
from burdensome work, to make more human friendly machines and let
machines act more according to human needs - not the other way round.
Fuzzy technology was early recognized as a possible candidate of such a
technology. For Omron, fuzzy is a paradigm to introduce human
subjectivity into objective science, and a method to model and use human
knowledge and senses as they are without complicating abstraction.
Back in 1966, Omron built an analog computer "DECIVAC" for decision
making based on probabilistic methods, which could handle (in modern
terminology) crisp membership functions and weighted inference rules; at
that time, Omron was not aware of the very young fuzzy theory. With some
proud I was told that this computer was even exported to West Germany.
In 1983, when the first fuzzy technology appeared in Japan (a control of
a drinking water treatment plant by Fuji Electric, and the famous subway
control by Hitachi), Takeshi Yamakawa of Kumamoto University visited
Omron+s Tokyo office in search of financial support for his fuzzy IC, a
hand made sample of which he had completed in October 1983. Omron saw
the potential of this new technology, and in October 1984, Yamakawa came
to Omron+s headoffice in Kyoto for a lecture, which was also attended by
Kazumata Tateishi, the founder of Omron. He showed a deep interest in
this new technology, and in that year, Omron started R&D on fuzzy expert
systems. In 1986, fuzzy hardware was developed under the guidance of
Yamakawa, and a medical diagnosis expert system.
1987, was the year of the beginning fuzzy boom in Japan. Omron built a
fuzzy controller and a "fuzzy computer" (a fuzzy ALU). In 1988, this was
marketed as the FZ-1000, and a special task force, the "F-Project" team
was established within Omron. A prototype fuzzy chip designed by
Yamakawa was manufactured (now the FZ-5000), and the hybrid (F+PID)
temperature controller E5AF were developed. Omron took part in
establishing the LIFE (Laboratory of International Fuzzy Engineering
Research), attended the Fuzzy Committee of the STA, and received a grant
of 600 MYen in a three year project to develop a fuzzy microprocessor
from the JRDC (Japan Research and Development Corporation, another STA-
organization).
In 1989, 10 new fuzzy products were announced by Omron, and 60 fuzzy
demonstrations appeared at the Omron Festival, an idea contest held
regularly within Omron. Prof. Zadeh became senior advisor of Omron, and
the F-Project team was transformed into the Fuzzy Technology Business
Promotion Center, while the team leader was dispatched for two years as
head of one of the three research laboratories of the LIFE in Yokohama.
In 1990, a fuzzy human body sensor, a fuzzy expert system for machine
diagnosis (together with the machinery manufacturer Komatsu), a digital
fuzzyJchip, a new tuning method for fuzzy control was established, and a
fuzzy inference unit were developed, but this listing is not complete.
In April 1991, Omron will appear at the Hannover Industrial Fair in
Germany to start its fuzzy control overseas.
FUZZY PATENTS.
Omron Corp. has filed applications of several hundred patents related to
fuzzy technology, although presently only 171 of them are made public
officially (patents are made public 18 months after application), and
none of them are registered yet (this process can take more than five
years). The following table of fuzzy related patents made public in
Japan between 1987 and 1990, detailed according to applicant and
technical content of patents, gives an impression of the current
situation:
Total Om Hi To Mi FF Ma Ni FE other
Total 319 107 35 34 30 15 14 12 10 62
Control Furnace 4 1 3
Applic. Process 4 3 1 0
Elevator 13 3 7 3 0
Aircondition 4 1 1 1 1 0
Robot 4 2 1 1
Tunnel drilling 1 1
Home Electronics 5 2 3 0
Automotive 41 9 1 2 5 11 13
Grinding, Treatment 3 2 1 0
Electric Power 3 1 1 1 0
Measurement 18 1 15 2
El. music instr. 6 6
others 21 9 1 9 1 1
Other Recognition 10 7 1 1 1
Applic. Detection, Verific. 7 4 1 2
Diagnose, Supervis. 7 2 1 1 1 2
Inference, Predict. 11 9 1 1
Retrieval 4 1 1 2
Picture Processing 2 2 0
Support, MMI 4 1 3 0
others 11 8 1 2
Method, Fuzzy Reasoning 40 11 5 4 5 2 6 7
Struct. Fuzzy Control 28 5 4 6 1 3 9
Computer Controller 12 12 0
Auto-tuning 10 8 2 0
Rules 8 7 1 0
Membership Function 28 20 1 7
Fuzzy Calculation 4 1 1 1 1
Defuzzification 4 3 1
others 2 1 1 0
Total Om Hi To Mi FF Ma Ni FE other
Om=Omron, Hi=Hitachi, To=Toshiba, Mi=Mitsubishi Electric, FF=Fuji Film,
Ma=Matsushita Electric, Ni=Nissan Auto, FE=Fuji Electric,
other=applicants with less than 10 patent applications
FUZZY EDITION OF THE SCIENCE MAGAZINE "QUARK", MARCH 1991.
On the 160 minutes shinkansen bullet train ride from Tokyo to Kyoto I
had a look at the special edition on fuzzy technology of the magazine
"Quark", which was also mentioned in David K. Kahaners short notice (15
February 1991) on the International Congress of Biomedical Fuzzy Systems
in February 1991 and in George J. Klirs report.
QUARK calls itself a "visual science magazine, revealing science for
pleasant life" and has a circulation of 120.000 copies per month. The 20
page special feature, titled "Becoming definitely strong at fuzzy",
rides heavily on the present japanese media+s fuzzy fad, epitomized in
the first of its four parts: a dialogue on "Who is fuzzy+s `mother of
invention`?" Such dialogues or round table talks are a necessary part of
any special feature on any theme in any japanese magazine, where
specialists and so-called commentators (hyoronka) talk freely about
anything considered relevant, leaving conclusions to the reader. This
dialogue was between Prof. Takeshi Yamakawa (45), the father of the
dedicated fuzzy chips, and Kazuhiko Kimura (32), a "best-selling trend
watcher" and columnist of several magazines and TV-programmes. They
talked around the reason of the high acceptance of fuzzy theory in Japan
and the post-fuzzy era. While Yamakawa tried to guarantee a certain
level of quality during the dialogue, Kimura gave a glimpse into the
shallowness of pop-science by arguing, for example, that automating
daily necessities like cooking by fuzzy rules and devices, would change
mankind considerably: "man+s secret reason for marriage has been to have
his wife do the cooking and household work, and hence fuzzy makes
marriage unnecessary (haha)".
The reason of the high acceptance of fuzzy in Japan was believed to be
its "eastern" background: oriental vagueness vs. occidental exactness.
While eastern and western people would *think* in their brains
similarly, the way of *expressing* their thoughts were different:
westerners would just express themselves straightly and rationally, but
easterners would also take into account the point of view and situation
of their listeners, thus prefering often vague expressions. The ancestor
of fuzzy theory, Lotfi Zadeh, is an Iranian (like his wife, who was born
in Japan and lived here until the age of 3) and used his eastern
experience when he tried to solve control problems that were intractable
by exact methods.
Part two of the special feature consisted of test reports of three fuzzy
consumer products: a video camera by Matsushita (NV-SI, 165000 Yen), a
washing machine by Toshiba (AW-50VF2, 95000 Yen) and a vacuum cleaner by
Hitachi (CV-AX760D, 55000 Yen); and a short description of 20 other home
electronics products.
Part three described five examples of "Automatic Control of Master
Skills": sake destillation (Gekkeikan), stock investment (Yamaichi
Securities), teeth correction (Matsumoto Dental University), subway
control (Sendai City) and prediction of "udon" noodle soup or "tofu"
bean curd sales in dependance of weather conditions (Japan Weather
Association).
Part four was a question & answer section.
Q1: How can fuzzy theory express vagueness scientifically?
Q2: What are membership functions?
Q3: How are membership functions constructed?
Q4: Where does fuzzy inferencing differ from conventional?
Q5: Why is fuzzy control suddenly used in home electronic products?
Q6: Are special integrated circuits necessary for fuzzy control?
Q7: Occur false conclusions at the end of a chain of vague inferencing?
Q8: How would a fuzzy computer look like?
Q9: What can happen, when fuzzy control malfunctions?
Q10: Where do fuzzy computers differ from neuro computers?
While most questions were dealt with in earlier reports (dkk, gjk, thg),
the answer to Q6 lists four different kinds of fuzzy IC+s presently in
use: general purpose microprocessors, which are programmed for fuzzy
inferencing (used also for fuzzy simulation, advantage of easy program
changes); analog circuits like in Yamakawa+s chip, where the values of
membership functions are represented by electric current or voltage
(advantage of high speed); digital circuits like Masaki Togai+s (AT&T)
chip, where values of membership functions are transformed into binary
digits; look-up tables, where all results of fuzzy inferences are stored
beforehand and just retrieved according to input values (advantage of
high speed and low cost, often used in home electronic products).
The last part ended with a list of ten books on fuzzy theory, authored
by well known insiders (Mukaidono, Sugeno, Terano, Yamakawa), published
by first rank companies (Iwanami Shoten, Kodansha, Nikkan Kogyo
Shinbunsha). The bestseller among them is "Recommending ambiguous
Engineering" by T. Terano of December 1981, now in its 16th printing;of
the 80000 printed copies, 75530 have been sold, for a price of 620 Yen.
The others have been published between August 1988 and November
1990,printed in copies between 2500 and 50000, with price tags between
1100 and 1450 Yen.
CONCLUSION.
Fuzzy technology is widely used in japanese products, is believed to
have an enormous future potential, is well known to the ordinary
educated Japanese, and might be an example of a technology invented and
successfully implemented in the East. (Please post comments on that into
soc.culture.japan or mail them directly to me since I only have a
restricted access to the usenet.)
--
Thomas Hagemann hagemann@tyo.gmd.dbp.de ++81-3-3586-7104
German National Research Center for Computer Science (GMD)
German Cultural Center, Akasaka 7-5-56, Minato-ku, Japan 107