Most of the companies the JTEC team visited are large enough to have their own research laboratories. Like their counterparts in the U.S., industrial laboratories of large companies in Japan engage in basic research and advanced development in a variety of areas. One example is Toshiba Electric, where the research and development infrastructure is similar to other computer companies active in selling ES tools and developing ES applications.
Toshiba has five research laboratories that conduct long-range (five to ten years) research in areas ranging from computer hardware, to manufacturing, to software. In addition, Toshiba has eight development laboratories that look ahead three to five years. In 1991, with annual net sales of $22.9 billion, the company spent 8.2 percent of that on research, close to double the ratio in 1981, when it spent 4.8 percent of sales on research (see Figure 3.7).
Toshiba's Systems and Software Engineering Laboratory (SSEL) was set up in 1987 with a charter to conduct research and basic development in systems and software areas and to develop corporate strategies in these areas. SSEL was also chartered to support operational divisions. SSEL has four divisions: (1) AI and human interface; (2) systems, covering distributed, neural network, and fuzzy systems; (3) software engineering; and (4) the systems and software development department, the mission of which is technology transfer. Although there are some good basic research activities -- for example, research on the use of multiple cases in case-based reasoning (Tanaka, Hattori et al. 1992) -- SSEL's primary strength lies in applied research.
Figure 3.7. Toshiba Corporation R&D Expenditures of Non-Consolidated Base (Source: Toshiba)
By applied research we mean the building of novel or complex application systems that use and extend the latest techniques from diverse areas. It is at once a technology transfer task as well as technology integration, refinement, and improvement task. Examples of advanced systems that are being built by SSEL include: model-based plant control systems, described in Chapter 2, which use advanced techniques, such as model-based reasoning, qualitative simulation, and knowledge compilation, in addition to the rule-based technique (Suzuki, Sueda et al. 1990) (research funded by ICOT); and model-based automatic programming systems (Nakayama, Mizutani et al. 1990). In addition to building advanced application systems, SSEL has developed prototype ES shells such as DiKAST (a diagnostic ES shell) and PROKAST (a scheduling ES shell). Both are knowledge acquisition, as well as problem-solving, tools integrating the results of SSEL's research in knowledge acquisition (Araki and Kojima 1991).
As mentioned earlier, the best-selling Japanese tools were first developed by, and continue to be developed by, Japanese computer manufacturers. In the U.S., in contrast IBM's TIRS has done poorly in the market. DEC uses ES technology extensively inside the company, and makes money in training and consulting with other companies on the uses of the technology, but it does not have its own tool. U.S. tools are developed and marketed by a dozen or so small companies rather than major hardware companies. The Japanese can continue to invest in the research and development of new tools (which they are doing) and are in a better position to survive lean times. In contrast, American vendors must work with short-term objectives and lean cash reserves.
On the research front, the U.S. fares somewhat better. As is the case with Toshiba in particular, industrial research in Japan is generally a mix of some basic research and advanced development, as in the United States. However, basic research activities of Japanese companies are on a relatively small scale compared to some U.S. companies such as IBM, Bell Labs, and Xerox PARC.
On the other hand, Japanese applied research is very extensive and closely integrated with the development and sales arms of the companies. In all the computer companies many, if not all, of the tool prototypes were first developed and tested in research laboratories. The first explorations of ES technology occurred in the research laboratories of Fujitsu, Hitachi, and NEC. Similar exploration also occurred in other large industrial laboratories such as those of Canon, Nippon Steel, and NTT. The first applications were joint efforts of researchers, system developers, and customers. Thus, the researchers serve as technology explorers and technology transfer agents, both in the computer companies and in the user companies.