Site: OBAYASHI CORPORATION
Mitsuwa Ogawa-machi Bldg. 3F
3-7 Kanda Ogawa-machi
Tokyo 101, Japan
Date Visited: March 25, 1992
The Obayashi Corporation is one of the world's major general construction contractors, among the top five in Japan. As of December 1990, there were approximately 12,000 employees, including 161 research scientists and technicians and 140 systems engineers. Revenues for 1990 were close to $10 billion. The corporation has over a dozen overseas subsidiaries, including three in the United States (Obayashi America, James E. Roberts - Obayashi, and Obayashi Hawaii).
After opening remarks by Mr. Matsuoka, Mr. Nakao discussed two expert systems at Obayashi. The first is an automatic drilling control system for shield tunneling (a method of construction first used in England in the 19th century. This system controls the direction of drilling, which must be accurate to within five centimeters per one kilometer of tunnel length. The non-linearity of the problem precluded the use of a mathematical model. The system employs fuzzy control to solve the problem. The system is now in use in some (not all) of Obayashi's drilling operations.
The second system is a swing control system for a cable crane, which is used to transport casting concrete for dam construction. Controlling the swing of the basket as it is accelerated, moved and decelerated to its final position is a non-linear control problem that was solved by using rules obtained from skilled crane operators. The system is being implemented directly in C, and will be integrated with a programmable controller using fuzzy control techniques. There are approximately 2,000 rules now, with more expected after the testing phase. Expected payback is a fourfold reduction in personnel.
We were also shown an interesting video on a "Super Construction Factory" planned for the 21st century, in which high-rise buildings are built almost entirely by specialized robots. The realization of this vision would require extensive research in advanced robotics and robot planning.
Obayashi has built 25 expert systems for internal use, and one for external use. Of these, six are in routine operation and nine more are at the field testing stage. Most of the systems (14) were characterized as advisory systems. The automatic direction control system for shield tunneling was built on top of two shells, called AI-DNA and AI-RNA, sold by AdIn, a Japanese AI company. The system is stand-alone, uses the C language, contains about 10,000 lines of code, and runs on a personal computer. The system was designed and built by two civil engineers, two mechanical engineers, and one systems engineer (plus programmers), and is maintained by the domain expert and two technicians. Development time was one year, including the testing period. The primary payback has been in improvement in the quality of direction control, and in the reduction of personnel required for this task, from three to one.
Obayashi reports that 70 percent of ES projects that are started get as far as a prototype, and 30 percent actually get to an operational system. Using existing tools they can envision building systems up to the size of a few thousand rules. Future systems planned by the corporation include other automatic control systems and intelligent CAD. The primary perceived problems with present technology are knowledge acquisition, constructing design systems, the need to incorporate model-based and case-based reasoning, and the need for machine learning.
The automated building construction concept (CIC: Computer Integrated Construction) and the highly automated tunneling systems were very impressive in concept. To the extent Japan succeeds in fulfilling this vision of CIC, another U.S. industry may be at risk.
Obayashi started with AI ten years ago. Our hosts stated that they want to use CADAM and AUTO CAD, but these products need to be made Japanese language capable. Hence, since they have no access to these products, they feel they are three to four years behind U.S. construction companies. This is based on written reports that Mr. Matsuoka has read of a scheduling ES used by Bechtel International, a pipe laying application by Stone and Webster, and structural design and visualization systems that Skidmore, Owens uses. Moreover, he believes that Europe is also ahead of Japan because ESPRIT has started on standardization of models (STEP) and in Japan they are just starting to discuss that issue. Mr. Matsuoka also feels that Japan may be three to four years behind in user interface technology.
AI, the dream, still exists in Japan. However, interest in AI and ES is decreasing in Japan, though many down-to-earth applications are occurring. Many companies use ES in special areas only and that activity is increasing. Initially, there were a high number of prototypes, but the research budget for AI prototypes has been reduced. Production systems are increasing; people now know what works and what doesn't.
Of Obayashi's own applications, many prototypes in the planning stage don't make it to use. Why? It is hard to make an indispensable system. Obayashi's people find that they can use AI in automatic control applications. They are using neural network learning to set up the parameters of a fuzzy control system.
Why does Obayashi use ESs?
Number (1) is the main driver.
Obayashi often undertakes projects to learn the technology.