Lapas attēli
PDF
ePub

CONCLUDING REMARKS With respect to their now taking a leading focus their R&D, simply by creating

role in fuzzy systems research, it has funding opportunities in desired areas From a very early age, Japanese been proposed that this is because Japan while withdrawing funding in others. school children are frequently reminded is one of the Eastern cultures, which They are also not afraid to invest occaof their country's vulnerable economic are reputed to be more receptive of sionally in high-risk ventures. condition. They live in an island nation, vagueness and imprecision. My per- Just as in the United States, howwith a population equal to a third of sonal opinion is that this is only a small, ever, the granting agencies do make the United States crowded onto a land and perhaps negligible, part of the overall their decisions on the basis of advice mass slightly larger than California, cause. It seems rather that Japan's invest- received from their leading scientists. and of which almost three-fourths is ment in fuzzy systems is only one of And this points to another possible mountainous and essentially uninhab- many instances of the same overall reason why fuzzy systems have flouritable. They have no natural resources strategy. They constantly scour the ished in Japan. Japanese university and not enough space to grow their horizon, watching for new scientific professors are all paid on a 12-month own food. They have been grappling ideas, which in effect are regarded only basis, so that they have no need for with the related problems for several as other raw materials, items waiting to summer salaries and are typically progenerations.

be transformed into marketable com- vided at least the minimal funding Their involvement in the Second modities. This approach to technolog- necessary to do their research. Hence, World War was largely an act of des- ical progress is evident even in their there is not as much incentive as in the peration, to increase their space by policy toward funding research at large United States for faculty to compete acquiring part of China. When that corporations and in universities. The for research funds, which means there failed they determined that their only prevailing emphasis has been on appli- is less cause to discredit ideas other option was to succeed in the interna- cations rather than theory. A nice illus- than one's own. Discussions regarding tional market place, earning enough tration of this fact is the following the pros and cons of a new idea can money to buy what they need to sur- breakdown of presentations at the take place in a much more friendly and vive. Their strategy since that time has congress of the International Fuzzy collegial manner, simply because one's been to import raw materials and turn Systems Association (IFSA-91) held in livelihood and/or academic survival is them into products that can be sold Brussels, Belgium, last July. The papers not so much at stake. When a few abroad.

were subdivided into four areas: Artifi- Japanese professors took an early interIn order to accomplish this it has cial Intelligence (AI); Engineering est in fuzzy systems, they were at least been necessary for them to excel in (Engin); Computer, Management, and treated with respect, if not actively business and technology. They are a System Sciences (CM&SS); and Math- encouraged. people who have no choice but to work ematics (Math). Note not only the dis- Fuzzy systems research has been hard and live by their wits. In order to tribution but the proportion of total suppressed in the United States largely accomplish their goals they are accord- participation by Japanese.

because the leading Al figures, and ingly inspired to acquiesce in a rather

principal advisors to the granting agentightly controlled and conservative

Total Ву

% cies, come from the traditional backsociety, with a high value placed on Topic Papers Japanese Japanese ground of symbolic approaches to AI. social harmony. The Japanese are well

Accordingly, the models of semantic aware that they cannot permit exces

59
23

39 inference proposed by Zadeh seemed sive squabbling among themselves, for Engin 65

37

57 not only strange but also threatening. they otherwise cannot live together in CM&SS 74

23
31

Out of this came an apparent (although such a small space and probably could Math

60
3

5 unstated) policy to provide only mininot collectively survive. This feeling

mal funding in this area. Also, because inspires in them a desire to care for, All

U.S. academics do not value social and cooperate with, one another in

harmony and mutual respect as much ways mostly unthinkable in the West. It The strong emphasis on applica- as do the Japanese, there was great also allows them to respond rapidly tions has been orchestrated largely by sport made in some circles of ridiculing and in unison to meet new challenges. MITI and STA through their funding and alienating anyone who was identi

Against this backdrop we have seen policies. In contrast with the United fied with the fuzzy systems movement. this nation rise rapidly to the forefront States, MITI can to a certain extent This only exacerbated the problem of in the realm of information technology. dictate where large corporations will gaining support.

AI

258

86

33.3

[ocr errors]

As to what might have been done, exercise their own vision as to what Whether projects such as LIFE are or what could be done in the future, to research should be supported. Within truly worthwhile has yet to be estabprevent similar phenomena from occur- this context, there should also be spe- lished, but more programs that proring, I can only offer a few tentative cial funding set aside for high-risk mote collaboration between universisuggestions. There is, of course, little ventures, with the understanding that ties, government, and industry would that can be done about the lack of the director is not to be penalized if a certainly be in order. American ingeimportance placed on values like social particular project fails to bear fruit. nuity should be able to provide the harmony and mutual respect, but then The performance of the director would needed new bridge between theory and this may not be altogether necessary. be judged not only by his superiors but applications. Out of rugged individualism and com- also by feedback actively solicited from petitive self-promotion comes also a the sector of academia in concern. Also Daniel G. Schwartz is an associate wealth of both scientific, philosophic, as in ONR, NSF directorships should professor in the Department of and artistic creativity. Indeed it is still have indefinite duration, thereby allow- Computer Science at Florida State the case that the deep theoretical break- ing directors to lay plans and carry them University, Tallahassee. He holds B.A throughs are being accomplished in out, and it should be possible for and M.S. degrees in mathematics, the West. This is because we continue researchers to establish a relationship and he completed his Ph.D. in systo support the intellectual infrastruc- with the agency, based on past perfor- tems science at Portland State Uniture necessary for this to occur. What is mance, rather than having to start anew versity in 1981. His dissertation and needed, therefore, is really only a better every couple of years. It would also

subsequent research concerned forsystem for (1) protecting and encourag- help to increase the overall budgets of mal axiomatizations of fuzzy logic ing fledgling ideas and (2) transform- these agencies; there are too many sci- and associated reasoning systems, ing theory into applications.

entists whose talents are being wasted. whereas his more recent work has For the former I would propose a

The latter aim constitutes a much involved the development of a new partial reorganization of the NSF, more difficult issue, as the United States symbolic approach to approximate creating a branch organized somewhat has strict rules regarding government reasoning. Dr. Schwartz has over 30 along the lines of the Office of Naval involvement in private enterprise. It published articles on these and related Research (ONR). Specifically this would would seem, though, that some agency topics. His current interests include amount to setting up directorships in such as MITI could be established to the integration of fuzzy rule-based major funding areas wherein the direc- aid corporations in identifying poten- reasoning with other reasoning parators themselves have the authority to tially fruitful lines of R&D and to grant digms and their applications both in be proactive, i.e., to formulate and incentives for working in those areas. expert decision support systems and

[graphic]

MATHEMATICAL THEORY OF NETWORKS

AND SYSTEMS '91 (MTNS '91)

The Mathematical Theory of Networks and Systems '91 conference and site
visits in Japan, Hong Kong, Macao, Singapore, and India are summarized.

by Biswa N. Datta

GENERAL OVERVIEW OF relationship existed in the United States in the areas of interest to the conTHE CONFERENCE as well.

ference. All these lectures had some

The conference was truly interna- tutorial flavor. The international symposium on the tional in nature. There were 558 In his opening address, Prof. Mathematical Theory of Networks and attendees and they came from all over Katsuhisa Furuta, a well-known conSystems (MTNS'91) was held in Kobe, the world: Australia, Belgium, Brazil, trol theorist from the Tokyo Institute Japan, from 17 through 21 June 1991. Canada, France, Hong Kong, India, of Technology, discussed how intelliThis was the ninth MTNS meeting and Israel, Italy, Japan, Korea, Malaysia, gent control can be effectively used in the first to be held in the Far East. The Mexico, The Netherlands, Poland, robotics. Intelligent control is usually previous meetings were held in the China, Taiwan, Spain, Sweden, defined as the "activities in the interUnited States, Canada, The Netherlands, Switzerland, Turkey, the United section of automatic control and artifiSweden, and Israel. MTNS meetings Kingdom, the United States, Vietnam, cial intelligence"; however, Furuta have been organized biannually since and Yugoslavia.

coined a new definition of intelligent 1973. MTNS '91 was held at the Inter- The

purpose

of the conference was control. He defined intelligence as "the national Conference Center on the to bring together research engineers, ability to adapt to the environment” manmade port island in Kobe City. mathematicians, and computer scien- and intelligent control as “the control Kobe is the largest port city in Japan tists to discuss mathematical problems to provide the intelligence to the sysand one of the most popular tourist of systems theoretical in nature arising tem.” Based on this definition, he prospots, surrounded by centuries-old in applications of current interest. It is posed a mechanism that appears to be historical cities like Kyoto, Osaka, Nara, a fact of life that engineers and mathe- effective for "sensor based robot conand Nagoya.

matical scientists hardly communicate trol such as compliance control (and) The local organizing committee of with each other. On the other hand, vision based servo and coordination the meeting, co-chaired by Profs. many research projects in applied sci- control of multiple arms”. H. Kimura and S. Kodama, both of ence and engineering are heavily depen- In another plenary talk on a similar Osaka University and internationally dent upon successful interdisciplinary topic titled “Systems Theory and Intelrenowned control theorists, did a super collaborations of mathematicians, ligence,” Dr. M. Vidyasagar, the direcjob. One of the remarkable skills of the engineers, and computer scientists. tor of the Center for Artificial Intelliorganizing committee was to attract The scientific program of the con- gence and Robotics in Bangalore, India, sponsorship support from most of the ference consisted of plenary lectures, showed that there are several areas of leading corporations of Japan, about special topic lectures, mini course, intelligence which give rise to interest50 of them. I have attended several of invited and contributed sessions, and a ing and challenging problems in systhe past MTNS conferences, but this poster session.

tems theory. By “intelligence" Dr. was the only conference with so many

Vidyasagar referred to those aspects corporate sponsorships. The sponsor- PLENARY LECTURES

of (rational) human behavior that comships by such a large number of corpo

puters find difficult to replicate, and by rations clearly indicate the strong The goal of the plenary lectures was “systems theory” he referred to a wide industry-university relationship that to provide the participants with the variety of topics such as complexity exists in Japan. I wish that a similar major research developments in topics theory, stochastic algorithms, simulated

annealing, neural networks, etc., besides • F.M. Callier, Univ. Notre-Dame de linear control systems in state-space the traditional areas of control theory. la Paix, Belgium, “Recent Advances modeling can be studied eloquently using

The other plenary lectures were given on Spectral Factorization and LQ- the techniques of combinatorics. These by Profs. B.D.O. Anderson of the Optimal Regulation for Multivari- discussions were strongly inspired by Australian National University; Israel able Systems"

the earlier work of C.T. Lin on strucGohberg of Tel Aviv University, Israel;

tural controllability. The far-reaching and A. Isidori of the University of Roma, • Bruce Francis, Univ. of Toronto, goal is to develop a combinatorial Italy.

Canada, “Sampled-Data Control analogue of dynamical systems theory Anderson's plenary talk was on Theory"

in a matroid theoretic framework. optimizing the discretization of con

Prof. M. Silva's talk was on how tinuous time controllers. A standard • M. Halsler, Ecole Polytechnic, tools from linear algebra could be used problem in digital control is replacing Switzerland, “Topological Methods in the analysis and synthesis of Petrinets. a continuous time controller with a for the Qualitative Analysis of discrete time controller so that, as far Nonlinear Circuits”

MINI COURSES as possible, closed-loop properties are preserved. The standard approaches to • Pramod Khargonekar, Univ. of There were two mini courses: one solve the problem do not make any use Michigan, “State-Space H. and on recent advances in neural network at all of the plant, whereas the closed- Robust Control Theory"

theory given by S. Amari of the Univerloop properties clearly depend on the

sity of Tokyo and M.I. Jordan of the plant as well as the controller. Anderson • M. Ikeda, Kobe Univ., Japan, Massachusetts Institute of Technology proposed some novel approaches for “Decentralized Control of Large- and the other on HQ/H_lu synthesis solving the problem that make use of Scale Systems”

given by John Doyle of Caltech, the knowledge of the plant. Two aspects

K. Glover of Cambridge University, of the closed-loop properties, stability • Pradip Pandey, UC Berkeley, “Solv- U.K., and Andy Packard of UC Berkeley. and the closed-loop transfer matrix, ing the Algebraic Riccati Equations The goal of these courses was to eduare emphasized.

on a Supercomputer”

cate the participants in the selected Gohberg's talk was on the interplay

research areas. The presentations were between interpolation problems and • A. Morse, Yale Univ., “A Concep- tutorial in nature, with particular attensystems theory. He described some tual Framework for Parameter tion given to motivation, definitions, recent developments in interpolation Adaptive Control”

and examples. theory and showed how these developments have greatly influenced modern • K Murota, Tokyo Univ., "Combina- SCIENTIFIC COMPUTING theoretical systems theory research. torial Systems Theory”

AND CONTROL THEORY Isidori's talk was on robust regulation of nonlinear systems. Research in • A. Ohsumi, Kyoto Institute of Activities in MTNS '91 nonlinear systems is still in its infancy. Technology, “Derivation of the Isidori showed how certain problems Schrodinger Equations from Sto- The design and analysis of linear in nonlinear systems can be solved using chastic Control"

control systems give rise to a variety of linearization techniques while for others

interesting computational linear algethese techniques are not useful. • M. Silva, Spain, “Linear Algebraic bra problems. Some of the well-known

Techniques for the Analysis and ones are: controllability and observaSPECIAL TOPICS LECTURES Synthesis of Petrinets"

bility problems, stability and inertia

problems, feedback stabilization and The goal of the special topics lec- Two of the lectures that I found extremely eigenvalue assignment problems (the tures was to present research advances interesting are described below. so-called pole assignment problems), in certain selected areas of interest to The talk by Prof. K. Murota of Tokyo frequency response problems, and matrix the conference; the lectures were really University was on combinatorial sys- equation problems (such as Lyapanov, designed for specialized subgroups of tems theory. Murota showed how cer- Sylvester, Riccati). Because of the the participants.

tain basic concepts such as control- importance of these problems, they have The special topics lectures were as lability, observability, stabilizability, etc. been very well studied both in mathefollows:

that arise in the design and analysis of matics and control literature. There

« iepriekšējāTurpināt »