18 - Dynamic Structuring of Intelligent Computer Control Systems

Abstract

Fundamentally, automation is concerned with the manipulation of physical properties of real-world objects, which if left alone do not achieve desirable behavior. The challenge facing automation research is not only to provide methods for achieving the desired behavior, but also to evolve a theory that defines and establishes the types of behavior that can be automated. Distributed intelligent control systems capture aspects of societal behavior and utilize these to solve complex or large-scale control problems. Such systems can be implemented using ideas from the fields of distributed computer control systems and multiagent systems. The properties of the knowledge representations and manipulations that can be incorporated into multiagent systems are analyzed in this chapter. In addition to traditional knowledge representation techniques, multiagent systems introduce a knowledge representation that can capture behavior exhibited by communities. Coordination between autonomous entities is essential in such a community environment and different approaches to this task are discussed in this chapter. To assist in understanding the knowledge processing that takes place in a multiagent system, a distinction between a priori and operational knowledge is introduced. This chapter also shows that this important distinction facilitates the rational development of multiagent systems that can dynamically structure themselves to address the control problem at hand. Three examples of experimental multiagent systems are presented in the chapter to illustrate the concepts and principles underlying this kind of controller.