Universität Bielefeld - Sonderforschungsbereich 360

The Dynamics of Intelligence:
Constraint-Based Architecture for Situated Agents

Alan K. Mackworth
Laboratory for Computational Intelligence
Department of Computer Science
University of British Columbia, Canada

Abstract

A robot is, typically, a hybrid intelligent system, consisting of a controller coupled to its plant. The Constraint Net (CN) model is a formal and practical model for building hybrid intelligent systems as Situated Agents. In CN, we model a robotic system as a symmetrical coupling of a robot with its environment. Even though a robotic system is, generally, a hybrid dynamical system, its CN model is unitary. Most other robot design methodologies use hybrid models of hybrid systems, awkwardly combining offline computational models of high-level perception, reasoning and planning with online models of low-level sensing and control.

Many robots can be designed as online constraint-satisfying devices. A robot in this restricted scheme can be verified more easily. Moreover, given a constraint-based specification and a model of the plant and the environment, automatic synthesis of a correct perceiver-controller becomes feasible. In our lab we have developed a testbed for multiple, visually-controlled, cheap robot vehicles performing a variety of tasks, including playing soccer. The ideas developed in this talk are illustrated by application to the challenge of building robot soccer players. This work is joint with Ying Zhang and several others.


Anke Weinberger, 1997-11-24