Lapas attēli
PDF
ePub

• The knowledge processing system

• Concurrent engineering and have developed an expert system to should be efficiently built and computer-aided design/ assist tree planting, i.e., to deal with available to use within the language manufacturing (CAD/CAM) landscaping, soil, vegetation, tree types, culture of conventional systems.

etc. • Computer-integrated manufactur- Osaka Gas scientists are develop• It should be efficient and resource ing (CIM) DB

ing a speech recognition system. This is saving.

becoming mature technology and there • Multimedia

is ample work going on in the West, • It should express, handle, and con

including Carnegie Mellon University trol various kinds of complicated • Management information systems (CMU) and other universities. Osaka's knowledge that differ from a simple (MIS) (now called Strategic Infor- application prototypes will be ready set of rules.

mation Systems)

this fall to assist gas construction workers

in filling out their reports and in digital • It should have a simple and easy to Knowledge base

mapping. My understanding of this work use front end.

is that it is using well established methods

Naokazu Yokoya (Electrotechnical (such as close talk mike, phonetic engine, Lest you think this is just for computer Laboratory (ETL), E-mail: yokoya@ phonetic codes, two stage decision tree, scientists, object Fortran has arrived. etl.go.jp) gave a good survey of new finite state grammar, etc.) but that it is Enter Fujitsu's Fortran/KR (knowledge trends in computer vision algorithms. I a complete system, neatly packaged using reasoning). This allows rules, objects, was very pleasantly surprised to learn a DSP5601 acoustic processor in a event-driven, fuzzy techniques, that “constraint satisfaction and coop- phonetic engine, hooked by RS232 to a Dempster-Shafer, and neural tech- erative algorithms” essentially meant 68020 workstation where the linguistic niques. Fujitsu has begun shipping a the use of extremely well understood decoder resides. In other words, it is development environment and object- methods in numerical optimization such ready to perform real work. oriented user interface. (I hope to see a as regularization, implemented via Y. Hosokawa (Sanyo) gave a detailed demonstration of this and will report gradient descent, simulated annealing, description of his company's CD-ROM further at that time.)

or other fairly standard approaches. car navigation system. There are sevK. Tanaka (Kobe University) gave There was also a discussion of matrix eral of these in actual use within Japan. an overview lecture on the developing iteration techniques necessary to solve I reported on Sumitomo's last year, methodology in database (DB), begin- the linearized problems that arise--again which used sophisticated onboard senning with hierarchical (CODASYL), mainstream numerical computation. sors. Sanyo's also uses sensors, but in moving to relational (RDB), then to A more detailed paper surveying addition incorporates input from the object-oriented (OODBMS). He felt research at Osaka University on vision eight GPS satellites that will be in that the use of each is bell-curve-like for autonomous mobile robots was position by 1992, allowing location over time and that now we are at the presented by Y. Shirai. Their work precision to 30 meters. (The sensors center of the RDB usage, at the tail- involves two key ideas: (1) stereo images are needed in cases when the satellites end of CODASYL use, and in the that are brought into correspondence are unavailable, such as in tunnels, prototype stage for OODBMS usage. by iteration and (2) color image seg- near high buildings, etc.) Hosokawa He went on to list the advantages he mentation using the brightness and edge also gives details of the map organization saw in using this new technology. histogram, i.e. without considering the on the CD-ROMs and points out that semantics of the scene.

about 20% of Japan (78% of the • Consolidation of various languages There were a number of very inter- population) has already been digitized (not a special DB language) esting advanced application talks includ- at a scale of 1:25000, although additional ing the following

building data are still needed. • Data model closer to world infor- Akiyoshi and Nishida (Mitsubishi Development of navigation systems is mation

Electric) (E-mail: {akiyoshi} {nishida} being supported in Japan by the police

@sys.crl.melco.co.jp) are studying the and various ministries, including the • Better performance, perhaps up to issue of training of large-plant oper- Ministry of Post and Telephone and two orders of magnitude

ators in how to cope with emergency the Ministry of Construction. Sanyo's situations.

system also permits voice (driver) input He also mentioned some of the Researchers from Kansai environ- for control. OODBMS applications he knew about. mental and electric power institutes

Kenichi Sahara (Sumitomo Elec- telecommunication and introduced the computing environment). WIDE was tric Industries, Ltd.) showed a design PMTC prototype for B-ISDN that started in 1987 and has a budget of environment for graphical user inter- supports up to 20 users. Their survey of $600,000. It has contracts with 15 faces based on the X Toolkit. Interface 708 Japanese businessmen in 35 com- companies, but unlike other countries, objects can be selected from a deposi- panies shows that about 60% of the it gets little assistance from the Govtory and the interface is constructed conferences requiring trips can be ernment. It consists of one backbone in without writing explicit code. The substituted by video conferences plus Japan that will be backed up over satinterface objects are embedded in editing facsimile. They implemented four ser- ellite (ISDN class D) starting from June objects, which act as a glue between the vice concepts into their PMTC proto- 1992. This backbone connects directly user and the interface. It was interest- type. First, the system is for personal to the Internet. It has 6 network opering that the editing objects can collect use, which means it must be based on a ation centers and connects to 130 netinformation about preferences of the personal workstation. Second, there are works. There are currently 57 researchers user and therefore accumulate knowl- virtual conference spaces necessary that involved in the project. Research focuses edge about the construction of the can be occupied by a subset of all con- on special local characters in net techinterface.

ference participants. While informa- nologies and multicast datagrams. ISDN Toshie Nakamura (Osaka Univer- tion from the outside flows into this is studied as a cost effective way to sity) used computers to investigate space, no information flows out. Third, connect computers. Students at the Keio human perception of music. She found multimedia support (graphics, telewrit- University campus in Fujisawa can that humans can recognize a crescendo ing, telepointing, and also high quality already access their network from a much easier than a decrescendo. Her full motion video and high quality voice public ISDN box at the campus. A project explanation: In the evolution of humans transmission) is necessary to exchange on voice interfaces might lead to syswe have learned that sounds fade out intentional and observational informa- tem and network administration over and therefore developed adoption tion. Fourth, a sophisticated user inter- public telephones. The person in charge mechanisms. These mechanisms focus face in the form of a multimedia multi- can then access the computer over the on the unusual, fading a decrescendo window display is a must. For the con- phone, get a status report from the out while amplifying a crescendo. She nection they use a mesh type architec- computer as voice mail, and give comalso found a relation between the opti- ture because a star type restricts the mands to correct a problem. An intermum time intervals in music or speech way of information exchange. Neces- national conference called INET '92 and the breathing and the mental tempo sary network speeds are in the range of will be held from 14-17 June 1992 in of the listener.

155/620 Mbps over multiple logical Kobe. Information is available over Kazunori Shimamura (NTT Human channels.

E-mail from inet92-info@wide.ac.jp. Interface Laboratories) reported on Jun Murai (Keio University) gave a WIDE itself can be contacted at their research about cooperative works talk on WIDE (widely distributed wide@wide.ac.jp. supported by multimedia multipoint

INTERNATIONAL CONFERENCE

ON COMPUTER-INTEGRATED
MANUFACTURING 1991 (ICCIM'91)

The International Conference on Computer-Integrated Manufacturing 1991

(ICCIM'91), held from 30 September to 4 October 1991 in Singapore, is
summarized. Singapore manufacturing science activities are focusing on

building value-added solutions using modern but well known ideas and
equipment and have a very aggressive education/training plan to increase the
number of specialists in advanced manufacturing, exactly what is needed for a
small country with few natural resources. Also, the Singapore/Japan AI Center
and the National University of Singapore are discussed. The AI Center provides

significant advantages to both countries.

by David K. Kahaner

[blocks in formation]

Country

No.

Also, while companies in Japan are

more advanced and active in "engiYugoslavia

2 neering” research than universities, this Poland

2 is changing very rapidly; their universiDenmark

2 ties are picking it up at a great speed. Finland

2 The recent Japan Intelligent ManufacNew Zealand

1 turing System (IMS) program puts very Romania

1 strong emphasis on university-based South Korea

1 research, and joint university/industry Israel

1 projects are a very powerful way to France

1

speed up this change. I am told that if Turkey

1 one attends international conferences

in Europe, for example, there are some Surprisingly, there were only a handful very good engineering papers from of Japanese attendees and almost no Japanese universities presented. Japanese speakers. Either advertising Three days of research papers (almost was weak in Japan or the Japanese 150 papers in five parallel sessions) were unable to attend for other rea- were preceded by two days of tutorials sons. It may be that some Japanese still in three parallel sessions. (Tutorial see an international conference held in speakers from the West were invited Singapore as not truly of international and expenses paid to encourage their standard or flavor. Perhaps they thought attendance.) On the last Saturday there they could get more exposure or learn was a visit to the GINTIC Institute of more new things by attending interna- CIM on the campus of Nanyang Techtional conferences in the West. This nological University. Aspects of the would apply equally well to Japanese tutorials that were most useful were participation in conferences in other the discussions by Singaporeans of their developing countries.

country's science and technology plans and especially in the area of manufac- Generally, the main lectures were to a talk about numerical optimization turing. I have summarized this mate- by academic experts who have been

by academic experts who have been techniques much like one I described rial below. The technical papers were thinking at a very high level and are in my report on Osaka's Computer the usual mix, a few are discussed below, primarily concerned with educating and World (see page 21). Also, there were but as the Proceedings are in English, planning, e.g., several papers had sub- several other papers to gladden the my emphasis is to focus on those things titles like “Strategic Issues for the '90s,” hearts of numerical analysts everywhere, not readily available. The visit to "Is CIM the Right Choice," "Simulta- including "Numerical Simulation as a GINTIC was a tremendous hit by all neous Engineering--The Competitive Central Element in a CIM System" participants and reinforced some of Edge,” and these would all make good (H. Groth et al., ETH) and “Computerthe points made during the earlier ses- reading. Within the contributed papers Aided Tolerance Chart Balancing” sions about Singapore's commitment there were a great many descriptions

there were a great many descriptions (B. Ann, Singapore), describing the use to this technology.

of real automated manufacturing, but of linear programming for this applicaMost of the featured and tutorial most of the papers were still about tion. There was a very good session on sessions were given by Western scien- plans, prototypes, simulation, or fairly standardization of data exchange fortists, including those from the United small systems. However, some of the mats, and a particularly excellent overStates (Gerhardt (Rensselaer), Lu plans are for systems that will impact view paper by M. Davies (CAD-CAM (University of Illinois), Biles (Univer- large manufacturing, such as one for Data Exchange, U.K.). In addition to sity of Louisville), Mills (University of production equipment in the Brazilian the expected large number of papers Texas-Arlington), and Tse (Stanford)] pulp and paper industry (R.M. Naveiro). from the West and from Singapore, I and Europe (Gillin (Swinburne Insti- There were also several talks on com- should also mention that Taiwan, Korea, tute of Technology), Sackett (Cranfield puter supported cooperative work and China were well represented. (A Institute of Technology), Ranky (CSCW), which takes as its primary very interesting paper describing the (University of Surrey), Weston role the ways in which computer sys- Chinese view of CIM was given by (Loughborough University of Technol- tems interact with people in a joint J. Xingling, Shanghai, alas, with no ogy), Nolan (University of Galway), effort to solve problems. Good discus- references.) I am happy to report also Warschat (Fraunhofer Institute for sions of this were given by P.P. Yim that there were four good papers from Industrial Engineering), and Vamos (CIM Group) and T. Winograd scientists at the University of Technol(Hungarian Academy of Sciences)]. The (Stanford). People in the States and ogy Malaysia, including a wheel wear majority of these highlighted talks were Europe (and even Japan) have been compensation study and a welding about how Western researchers see exploring this technology to support engineer expert system. Perhaps this manufacturing in the next 5 to 10 years. engineering teams for several years now. very poor country will be able to bootThe significant Singaporean exception Xerox, Digital, and IBM, for example, strap its way into modern industrial was the excellent tutorial by

have been active in this area. Fuji-Xerox development. Surprisingly, there were

has a division of almost 100 people to no papers from Indonesia, which is the Prof. Nai Choon Ho

workon this technology. The basicidea world's fifth largest as well as the fifth Director

is to extend the impact of computers on most populous country, has large quanGINTIC Institute of CIM individual productivity gains to group

tities of oil, and is a neighbor of Nanyang Technological University productivity improvements.

Singapore. Nanyang Avenue, Singapore 2263 Even among the contributed papers There was also a small vendor exhibit Tel: +65-6605551, +65-6605490 there were many addressed primarily of about 18 companies, mostly Fax: +65-7916377

to management trying to educate them computer-aided design/manufacturing E-mail: ncho@ntivax.bitnet to the changes needed to support CIM (CAD/CAM) systems, electronic data

(see, for example, Leong's on “Build- interchange (EDI) interfaces, robotics who has been most helpful in explain- ing a Supporting Manufacturing Infra- applications, artificial intelligence (AI) ing details of Singapore's science direc- structure for CIM”). (Leong lists almost packages, and related software. About tions to me and who also provided a 20 steps.) There were several papers half of these companies were joint great deal of statistical information for about teaching CIM, the most notable ventures between Western and my use. Ho is also a very articulate about a master course at the University Singaporean companies (e.g., Hewlettspokesman for CIM, and people like of Surrey (Ranky) and plans for a Packard Singapore, IBM, DEC, Sun); him are a critical reason for Singapore's computer network for education on the remainder were Singaporean comcurrent achievements and probable CIM in Japan (Kanda et al.). In the panies showing either their own or future success.

area of simulation, I again was treated Western products. There was also one Singaporean book publisher, World system solution through a gradual at ICCIM'91, there was the definite Scientific. To date most of their books evolution through these consecutive sense of still trying to justify the research, have been in science, but they are moving levels of understanding; any deviation that is, trying to convince others that rapidly into the engineering market. within or between these levels will not there is something here more than just

lead to a useful system solution. ideas. To me this feels very much like SUMMARY

Cutting through the jargon, this the early days of computer science, and

means that the manufacturing facility if so it will pass in time. This report should be considered as of today should be thought of as one There are two issues to be mena supplement to my earlier one part of a larger organization that includes tioned about this meeting. The first [D.K. Kahaner, “Computing and marketing, planning, designing, sched- concerns new ideas in manufacturing Related Scientific Activities in uling, ordering, and shipping raw mate- research in an absolute sense and the Singapore,” Scientific Information rials and products, etc., and of course, second concerns its state in Singapore. Bulletin 16(3), 53-63 (1991)], and so I the shop floor manufacturing, too. In I believe it is fair to say that Singapore will not repeat background material. this context, intelligent or integrated is not developing ideas that are ground

In addition to the technical details means information flow--making infor- breaking. The work that I saw is along that were reported in depth at this mation widely available, using the same very traditional lines using well estabconference, there was also a very clear information consistently, and moving lished concepts and basic directions. statement about the future of manu- information rapidly from its source to There is a substantial amount ofeducafacturing. This was made repeatedly where it is needed. In other words, we tion related research and training, and by speakers and audience and should are moving away from the computer also several interesting potential prodbe thought about carefully by everyone numerical controller (CNC) and robot ucts that are in progress. Most of this concerned with productivity. The term manipulation world that is down on work is centered around the GINTIC "computer-integrated manufacturing" the shop floor, and beyond the purely Institute of CIM at Nanyang Technois already viewed as not correctly describ- data intensive, database, CAD/CAM logical University. Asmaller amount is ing future directions. Instead, people world, to a knowledge intensive deci- at the National Computer Board's Inforare talking about "computer-integrated sion level where the tools will be much mation Technology Institute. GINTIC enterprises."

more knowledge processing related, and the program to train Singaporeans The main difference between CI involving AI and other high level are very strongly modeled after Western "manufacturing” and CI “enterprise” computing models. In this world viewit models, such as the programs at is that researchers have now realized is extremely clear the crucial role to be Rensselaer and at Fraunhofer. In fact, the only way to solve manufacturing played by computing and associated problems is to take a “system” point of information technology (IT) fields; Dr. Lester A. Gerhardt view of the whole production enter- information and its flow is seen to be at Associate Dean for Research prise. Taking a system point of view, least as important as welding, grinding, School of Engineering however, makes the problems much etc. Any company or country that can Prof. Electrical, Computer & Systems more difficult. Generally speaking, there pull these together was seen by this Engineering Dept are three levels of understanding, event, group as having a large leg-up on those Rensselaer Polytechnic Institute process, and methodology, before sound who cannot.

(RPI) solutions are possible. The event level Some interesting contrasting com- Troy, NY 12180-3590 focuses on individual results (or cases) ments were made by scientists within Tel: (518) 276-6203/6400 and experiences gained from these the CIM community. For example, Fax: (518) 276-8788 results. At the process level, attention P. Yim asks, “Have manufacturing com- E-mail: gerhardt@rpitsmts.bitnet is paid to the underlying reasons (e.g., panies or industries, as a whole, improved physics, mechanisms, rationales) upon through the synergy gain from the past was one

was one of the featured speakers at which the events were based. At the decade's integration efforts for those ICCIM'91, and he has been to Singapore methodology level, various models and who opted to invest into computer- many times. I asked him specifically theories, empirical or analytical, are integrated manufacturing? The answer about research here and he pointed out proposed to explain and guide event she feels) leans towards NO.”I am sure

she feels) leans towards NO."I am sure that while the country has made huge occurrences based on the understand- that there would be quite a lot of dis- progress over the past 5 years, it is still ing of their respective processes. It is agreement with Yim's assessment. mostly instituting programs and projfelt that one can only arrive at a sound However, within the papers presented ects from the West and trying to adapt

« iepriekšējāTurpināt »