• The knowledge processing system should be efficiently built and available to use within the language culture of conventional systems.

• It should be efficient and resource saving.

• It should express, handle, and control various kinds of complicated knowledge that differ from a simple set of rules.

• It should have a simple and easy to use front end.

Lest you think this is just for computer scientists, object Fortran has arrived. Enter Fujitsu's Fortran/KR (knowledge reasoning). This allows rules, objects, event-driven, fuzzy techniques, Dempster-Shafer, and neural techniques. Fujitsu has begun shipping a development environment and objectoriented user interface. (I hope to see a demonstration of this and will report further at that time.)

K. Tanaka (Kobe University) gave an overview lecture on the developing methodology in database (DB), beginning with hierarchical (CODASYL), moving to relational (RDB), then to object-oriented (OODBMS). He felt that the use of each is bell-curve-like over time and that now we are at the center of the RDB usage, at the tailend of CODASYL use, and in the prototype stage for OODBMS usage. He went on to list the advantages he saw in using this new technology.

• Consolidation of various languages (not a special DB language)

• Data model closer to world information

• Better performance, perhaps up to two orders of magnitude

He also mentioned some of the OODBMS applications he knew about.

• Concurrent engineering and computer-aided design/ manufacturing (CAD/CAM)

• Computer-integrated manufacturing (CIM) DB

Multimedia

• Management information systems (MIS) (now called Strategic Information Systems)

• Knowledge base

Naokazu Yokoya [Electrotechnical Laboratory (ETL), E-mail: yokoya@ etl.go.jp] gave a good survey of new trends in computer vision algorithms. I trends in computer vision algorithms. I was very pleasantly surprised to learn that "constraint satisfaction and cooperative algorithms" essentially meant the use of extremely well understood methods in numerical optimization such as regularization, implemented via gradient descent, simulated annealing, or other fairly standard approaches. There was also a discussion of matrix iteration techniques necessary to solve the linearized problems that arise--again mainstream numerical computation.

A more detailed paper surveying research at Osaka University on vision research at Osaka University on vision for autonomous mobile robots was presented by Y. Shirai. Their work involves two key ideas: (1) stereo images that are brought into correspondence by iteration and (2) color image segmentation using the brightness and edge histogram, i.e. without considering the semantics of the scene.

There were a number of very interesting advanced application talks including the following.

Akiyoshi and Nishida (Mitsubishi Electric) (E-mail: {akiyoshi} {nishida} @sys.crl.melco.co.jp) are studying the issue of training of large-plant operators in how to cope with emergency situations.

Researchers from Kansai environmental and electric power institutes

have developed an expert system to assist tree planting, i.e., to deal with landscaping, soil, vegetation, tree types,

etc.

Osaka Gas scientists are developing a speech recognition system. This is becoming mature technology and there is ample work going on in the West, including Carnegie Mellon University (CMU) and other universities. Osaka's application prototypes will be ready this fall to assist gas construction workers in filling out their reports and in digital mapping. My understanding of this work is that it is using well established methods (such as close talk mike, phonetic engine, phonetic codes, two stage decision tree, finite state grammar, etc.) but that it is a complete system, neatly packaged using a DSP5601 acoustic processor in a phonetic engine, hooked by RS232 to a 68020 workstation where the linguistic decoder resides. In other words, it is ready to perform real work.

Y. Hosokawa (Sanyo) gave a detailed description of his company's CD-ROM car navigation system. There are several of these in actual use within Japan. I reported on Sumitomo's last year, which used sophisticated onboard sensors. Sanyo's also uses sensors, but in addition incorporates input from the eight GPS satellites that will be in position by 1992, allowing location precision to 30 meters. (The sensors are needed in cases when the satellites are unavailable, such as in tunnels, near high buildings, etc.) Hosokawa also gives details of the map organization on the CD-ROMs and points out that about 20% of Japan (78% of the population) has already been digitized at a scale of 1:25000, although additional building data are still needed. Development of navigation systems is being supported in Japan by the police and various ministries, including the Ministry of Post and Telephone and the Ministry of Construction. Sanyo's system also permits voice (driver) input for control.

Kenichi Sahara (Sumitomo Electric Industries, Ltd.) showed a design environment for graphical user interfaces based on the X Toolkit. Interface objects can be selected from a depository and the interface is constructed without writing explicit code. The interface objects are embedded in editing objects, which act as a glue between the user and the interface. It was interesting that the editing objects can collect information about preferences of the user and therefore accumulate knowledge about the construction of the interface.

Toshie Nakamura (Osaka University) used computers to investigate human perception of music. She found that humans can recognize a crescendo much easier than a decrescendo. Her explanation: In the evolution of humans we have learned that sounds fade out and therefore developed adoption mechanisms. These mechanisms focus on the unusual, fading a decrescendo out while amplifying a crescendo. She also found a relation between the optimum time intervals in music or speech and the breathing and the mental tempo of the listener.

Kazunori Shimamura (NTT Human Interface Laboratories) reported on their research about cooperative works supported by multimedia multipoint

telecommunication and introduced the PMTC prototype for B-ISDN that supports up to 20 users. Their survey of 708 Japanese businessmen in 35 companies shows that about 60% of the conferences requiring trips can be substituted by video conferences plus facsimile. They implemented four service concepts into their PMTC prototype. First, the system is for personal use, which means it must be based on a personal workstation. Second, there are virtual conference spaces necessary that can be occupied by a subset of all conference participants. While information from the outside flows into this space, no information flows out. Third, multimedia support (graphics, telewriting, telepointing, and also high quality full motion video and high quality voice transmission) is necessary to exchange intentional and observational information. Fourth, a sophisticated user interface in the form of a multimedia multiwindow display is a must. For the connection they use a mesh type architecture because a star type restricts the way of information exchange. Necessary network speeds are in the range of 155/620 Mbps over multiple logical channels.

Jun Murai (Keio University) gave a talk on WIDE (widely distributed

computing environment). WIDE was started in 1987 and has a budget of $600,000. It has contracts with 15 companies, but unlike other countries, it gets little assistance from the Government. It consists of one backbone in Japan that will be backed up over satellite (ISDN class D) starting from June 1992. This backbone connects directly to the Internet. It has 6 network operation centers and connects to 130 networks. There are currently 57 researchers involved in the project. Research focuses on special local characters in net technologies and multicast datagrams. ISDN is studied as a cost effective way to connect computers. Students at the Keio University campus in Fujisawa can already access their network from a public ISDN box at the campus. A project on voice interfaces might lead to system and network administration over public telephones. The person in charge can then access the computer over the phone, get a status report from the computer as voice mail, and give commands to correct a problem. An international conference called INET '92 will be held from 14-17 June 1992 in Kobe. Information is available over E-mail from inet92-info@wide.ac.jp. WIDE itself can be contacted at wide@wide.ac.jp.

and especially in the area of manufacturing. I have summarized this material below. The technical papers were the usual mix, a few are discussed below, but as the Proceedings are in English, my emphasis is to focus on those things not readily available. The visit to GINTIC was a tremendous hit by all participants and reinforced some of the points made during the earlier sessions about Singapore's commitment to this technology.

Most of the featured and tutorial sessions were given by Western scientists, including those from the United States [Gerhardt (Rensselaer), Lu (University of Illinois), Biles (University of Louisville), Mills (University of Texas-Arlington), and Tse (Stanford)] and Europe [Gillin (Swinburne Institute of Technology), Sackett (Cranfield Institute of Technology), Ranky (University of Surrey), Weston (Loughborough University of Technology), Nolan (University of Galway), Warschat (Fraunhofer Institute for Industrial Engineering), and Vamos (Hungarian Academy of Sciences)]. The majority of these highlighted talks were about how Western researchers see manufacturing in the next 5 to 10 years. The significant Singaporean exception was the excellent tutorial by

Prof. Nai Choon Ho
Director

GINTIC Institute of CIM
Nanyang Technological University
Nanyang Avenue, Singapore 2263
Tel: +65-6605551, +65-6605490
Fax: +65-7916377
E-mail: ncho@ntivax.bitnet

who has been most helpful in explaining details of Singapore's science directions to me and who also provided a great deal of statistical information for my use. Ho is also a very articulate spokesman for CIM, and people like him are a critical reason for Singapore's current achievements and probable future success.

Generally, the main lectures were by academic experts who have been thinking at a very high level and are primarily concerned with educating and planning, e.g., several papers had subtitles like "Strategic Issues for the '90s," "Is CIM the Right Choice," "Simultaneous Engineering--The Competitive Edge," and these would all make good reading. Within the contributed papers reading. Within the contributed papers there were a great many descriptions of real automated manufacturing, but most of the papers were still about plans, prototypes, simulation, or fairly small systems. However, some of the plans are for systems that will impact large manufacturing, such as one for production equipment in the Brazilian pulp and paper industry (R.M. Naveiro). There were also several talks on computer supported cooperative work (CSCW), which takes as its primary role the ways in which computer systems interact with people in a joint effort to solve problems. Good discussions of this were given by P.P. Yim (CIM Group) and T. Winograd (Stanford). People in the States and Europe (and even Japan) have been exploring this technology to support engineering teams for several years now. Xerox, Digital, and IBM, for example, have been active in this area. Fuji-Xerox has a division of almost 100 people to work on this technology. The basic idea is to extend the impact of computers on individual productivity gains to group productivity improvements.

Even among the contributed papers there were many addressed primarily to management trying to educate them to the changes needed to support CIM (see, for example, Leong's on "Building a Supporting Manufacturing Infrastructure for CIM"). (Leong lists almost 20 steps.) There were several papers about teaching CIM, the most notable about a master course at the University of Surrey (Ranky) and plans for a computer network for education on CIM in Japan (Kanda et al.). In the area of simulation, I again was treated

to a talk about numerical optimization techniques much like one I described in my report on Osaka's Computer World (see page 21). Also, there were several other papers to gladden the hearts of numerical analysts everywhere, including "Numerical Simulation as a Central Element in a CIM System" (H. Groth et al., ETH) and “ComputerAided Tolerance Chart Balancing" (B. Ann, Singapore), describing the use of linear programming for this application. There was a very good session on standardization of data exchange formats, and a particularly excellent overview paper by M. Davies (CAD-CAM Data Exchange, U.K.). In addition to the expected large number of papers from the West and from Singapore, I should also mention that Taiwan, Korea, and China were well represented. (A very interesting paper describing the Chinese view of CIM was given by J. Xingling, Shanghai, alas, with no references.) I am happy to report also that there were four good papers from scientists at the University of Technology Malaysia, including a wheel wear compensation study and a welding engineer expert system. Perhaps this very poor country will be able to bootstrap its way into modern industrial development. Surprisingly, there were no papers from Indonesia, which is the world's fifth largest as well as the fifth most populous country, has large quantities of oil, and is a neighbor of Singapore.

There was also a small vendor exhibit of about 18 companies, mostly computer-aided design/manufacturing (CAD/CAM) systems, electronic data interchange (EDI) interfaces, robotics applications, artificial intelligence (AI) packages, and related software. About half of these companies were joint ventures between Western and Singaporean companies (e.g., HewlettPackard Singapore, IBM, DEC, Sun); the remainder were Singaporean companies showing either their own or Western products. There was also one

Singaporean book publisher, World Scientific. To date most of their books have been in science, but they are moving rapidly into the engineering market.

SUMMARY

This report should be considered as a supplement to my earlier one [D.K. Kahaner, "Computing and Related Scientific Activities in Singapore," Scientific Information Bulletin 16(3), 53-63 (1991)], and so I will not repeat background material.

In addition to the technical details that were reported in depth at this conference, there was also a very clear statement about the future of manufacturing. This was made repeatedly by speakers and audience and should be thought about carefully by everyone concerned with productivity. The term "computer-integrated manufacturing" is already viewed as not correctly describing future directions. Instead, people are talking about "computer-integrated enterprises."

The main difference between CI "manufacturing" and CI "enterprise" is that researchers have now realized the only way to solve manufacturing problems is to take a "system" point of view of the whole production enterprise. Taking a system point of view, however, makes the problems much more difficult. Generally speaking, there are three levels of understanding, event, process, and methodology, before sound solutions are possible. The event level focuses on individual results (or cases) and experiences gained from these results. At the process level, attention is paid to the underlying reasons (e.g., physics, mechanisms, rationales) upon which the events were based. At the methodology level, various models and theories, empirical or analytical, are proposed to explain and guide event occurrences based on the understanding of their respective processes. It is felt that one can only arrive at a sound

system solution through a gradual evolution through these consecutive levels of understanding; any deviation within or between these levels will not lead to a useful system solution.

Cutting through the jargon, this means that the manufacturing facility of today should be thought of as one part of a larger organization that includes marketing, planning, designing, scheduling, ordering, and shipping raw materials and products, etc., and of course, the shop floor manufacturing, too. In this context, intelligent or integrated means information flow--making information widely available, using the same information consistently, and moving information rapidly from its source to where it is needed. In other words, we are moving away from the computer numerical controller (CNC) and robot manipulation world that is down on the shop floor, and beyond the purely data intensive, database, CAD/CAM world, to a knowledge intensive decision level where the tools will be much more knowledge processing related, involving AI and other high level computing models. In this world view it computing models. In this world view it is extremely clear the crucial role to be played by computing and associated information technology (IT) fields; information and its flow is seen to be at least as important as welding, grinding, etc. Any company or country that can pull these together was seen by this group as having a large leg-up on those who cannot.

Some interesting contrasting comments were made by scientists within the CIM community. For example, P. Yim asks, "Have manufacturing companies or industries, as a whole, improved through the synergy gain from the past decade's integration efforts for those who opted to invest into computerintegrated manufacturing? The answer [he feels] leans towards NO." I am sure that there would be quite a lot of disagreement with Yim's assessment. However, within the papers presented

at ICCIM'91, there was the definite sense of still trying to justify the research, that is, trying to convince others that there is something here more than just ideas. To me this feels very much like the early days of computer science, and if so it will pass in time.

There are two issues to be mentioned about this meeting. The first concerns new ideas in manufacturing research in an absolute sense and the second concerns its state in Singapore.

I believe it is fair to say that Singapore is not developing ideas that are ground breaking. The work that I saw is along very traditional lines using well established concepts and basic directions. There is a substantial amount of education related research and training, and also several interesting potential products that are in progress. Most of this work is centered around the GINTIC Institute of CIM at Nanyang Technological University. Asmaller amount is at the National Computer Board's Information Technology Institute. GINTIC and the program to train Singaporeans are very strongly modeled after Western models, such as the programs at Rensselaer and at Fraunhofer. In fact,

Dr. Lester A. Gerhardt
Associate Dean for Research
School of Engineering

Prof. Electrical, Computer & Systems
Engineering Dept
Rensselaer Polytechnic Institute
(RPI)

Troy, NY 12180-3590
Tel: (518) 276-6203/6400
Fax: (518) 276-8788

E-mail: gerhardt@rpitsmts.bitnet

was one of the featured speakers at ICCIM'91, and he has been to Singapore many times. I asked him specifically about research here and he pointed out that while the country has made huge progress over the past 5 years, it is still mostly instituting programs and projects from the West and trying to adapt