Hybrid Problems in Smart Matter Control

Hybrid Systems and AI: Modeling Analysis and Control of Discrete Plus Continuous Systems

Volume

Papers from the 1999 AAAI Spring Symposium

Contents

The remarkable increase in computer capabilities per unit price has led to an explosion of computer applications in processing information. Similarly, the significant increase in sensor and actuator capabilities per unit price now under way combined with the aforementioned computer advances will enable a rapid increase in the number of control systems, i.e., systems that can sense and manipulate their environment. Many of the machines of the industrial age can be rearchitected using a multitude of sensors, actuators, and control systems if the requisite component prices are sufficiently low. In particular, the number of controllers can be sufficiently large that the statistical properties of the ensemble dominate over specific details of individual elements. Such systems have become known as smart matter [1,2]. Unlike traditional matter, the components are capable of complex continuous and discrete actions. Such changes in capability will require control algorithms capable of operating a multitude of interconnected discrete and continuous sensors, actuators, and control systems in a robust and adaptable manner. In this paper, some of the challenges associated with creating such hybrid control systems for large numbers of components will be discussed along with some of our initial work in this area.