Conference heron::euro_swas_ai

350.0. "fuzzy logic trip report" by ULYSSE::ROACH (TANSTAAFL !) Wed Jul 17 1991 11:22

 

                  I N T E R O F F I C E   M E M O R A N D U M

                                        Date:     04-Jul-1991 11:53am CET
                                        From:     ORCIUCH
                                                  ORCIUCH@YIPPEE@MRGATE@HUGHI
                                        Dept:      
                                        Tel No:    

TO:  PAT ROACH@VBO


Subject: fuzzy logic trip report

From:	AIAG::PUNDIT "ASDS Research, dtn 291-8100  03-Jul-1991 1227"  3-JUL-1991 18:30:46.93
To:	csse::herbener
CC:	@asdsmgrs,@aitclist,news
Subj:	I: Trip Report - Fuzzy Logic Conf

On June 27-28, 1991 I attended the Industrial Conference on Fuzzy Systems
in Austin sponsored by MCC* and the U of Texas. It was attended by over 300
people representing Govt., industry, and universities. (IBM was conspicuous
by its absence., Digital had 10, Motorola 48, TI 12, Kodak 8, ...)

(*MCC, Microelectronics and Computer Corp, is a consortium of some 22 US
companies in which DEC is a member. MCC has a very modest program in Fuzzy
Systems.)

The purpose of the conference was to explore why Japan has large number of
applications of Fuzzy Logic whereas US has few, and the future direction.
Here are some highlights:

1. Spectacular successes -- Japanese consumer electroics claimed the most
spectacular successes. A washing machine senses weight, soil, and texture of
the clothes, determines the right amount of detergent, water, soaking, washing,
spinning, ... with impressive results and user simplicity of operating an
on/off switch. A wall-mounted air-conditioner has infra-red sensor to monitor
human traffic, flow of hot or cold air by a door or window opening, ... and
adjusts the vents, the fan speed, and the compression. Similar ideas in
automatic cameras. These applications used a chip incorporating the fuzzy
logic. Patents filed in Japan now exceed 2,000.

The enabling piece (beyond human ingenuity) is the application of fuzzy logic
(which deals with approximate reasoning) and sometimes a "defuzzification"
scheme which translates the fuzzy rules into (step-wise) continuous functions.

Other interesting developments include: balancing the pendulum; decomposition
of the 2-hinged pendulum balancing problem; smoothed imaging which compared
very well in quality with the (computationally expensive) Kalman filtering
solution; ...

Japan certainly leads in applications. And, MITI is pouring money into it.
Japan's consortia of univ and industry under MITI funding have very ambitious
goals from fundamental research to industrial applications to future tools.

Why Japan? There is no satisfactory answer. However, possible reasons cited
ranged from conducive Japanese culture (the ease of acceptance),
discouraging western culture (of dealing in black and white), Japanese 
tenacity in overcoming brute force by ingenuity, Zadeh's feeling that
US throws too much mathematics at a problem, ... However, there is significant
progress in Europe as well.

2. Future - Where do we go from here?

Prof. Bart Kosco at USC has just published a book with a foreword by the
founding father Prof Lofti Zadeh in which Zadeh considers this book to be
the first complete textbook on the subject he initiated some 30 years ago.
Kosco gave an excellent 2 hr tutorial. And, Zadeh was the post-dinner
keynote speaker. (Earlier, I had lunch with Prof Zadeh whom I have known
since a long long time.) Here are some salient points:

a. We will see a greater sophistication in the products. For example, Minolta
has just launched an SLR camera 7xxi last month which now appeals to
professional photographers as well as amateurs.

b. We will see the use of neural nets with fuzzy logic. The initial
control value could be based on the fuzzy logic but refined over time with
neural net.

c. There could be a marriage of fuzzy logic with expert system. 
That is, the "rules" could be enriched in both cause and effect parts
using fuzzy logic (with defuzzification), and values derived from neural nets. 

d. Some ambitious efforts are: in NASA (20 km tether maneuvering in space);
control of 4-rotor helicopters, and  an integrated graphics workstation
environment combining expert system and fuzzy logic - in Japan's Laboratory
for International Fuzzy Engineering Research (LIFE).

3. What can DEC do ?

As DEC moves closer to commodity market, we can not afford to ignore the
developments in fuzzy logic. We can start by learning, participating with 
consortia (both US and Japan), launching  head on by extending our AI  
capability with suitable applications, and untertaking tool development.

Many details have been purposely left out. However, I can be reached for
additional information you may need. The purpose of this report is to whet
your appetite first.

Regards,
Neil

    
    Look at 	EUROPE::EERP_ABSTRACTS not 49. There is good expertise in
    Europe on this subject.
    
    Brian