The application of intelligent agent technology to simulation

We have been successfully applying intelligent agent technology to several real world simulation problems such as semiconductor manufacturing processes, train operations, and electric motor assembly lines. Agent technology is a methodology to realize an autonomous decentralized system with cooperative interactions among agents that model each element of the system. An intelligent agent has problem solving and learning skills, as well as the knowledge for that purpose. An agent is characterized as having autonomy, sociality, reactivity, and proactiveness. In the development of a software system using agent technology, each element of the system is described in an independent and modular program code. This method makes addition, change, and deletion of an element much easier than the case of conventional programming. In almost any real world system, each element works mostly in an independent and parallel manner yet with interactions with each other, and matches well with the concept of an agent. We have chosen a platform called the PIM, parallel inference machine, that describes and executes multiple agents in the independent and parallel manner. It makes the simulation application software development for a real world system much more straightforward than conventional computing platforms. This paper describes the agent technology, its application to real world simulation and the platform "PIM". It describes in some detail actual simulator application examples: semiconductor process system management, train operation management, and electric motor assembly line change. (C) 2003 Elsevier Science Ltd. All rights reserved.