T6 - HALF DAY

==============================================================

RATIONAL ACTION IN AUTONOMOUS AGENTS

Mike Wooldridge, Simon Parsons


=============================================================
ABSTRACT OF THE TUTORIAL

Following an introduction to the basic idea of autonomous agents, we
survey the main theoretical approaches to rational decision making, in
which rational decision-making is viewed as utility
maximisation. After discussing the limitations of such approaches, we
introduce the key techniques proposed in the literature for overcoming
these problems. We focus in particular on the belief-desire-intention
(BDI) model of agency, which we illustrate by means of examples from
mobile robotics.
After attending this tutorial, an attendee will:
* understand the concept of an autonomous agent;
* understand basic economic and decision-theoretic approaches to
  rational action in autonomous agents;
* understand the limitations of naive utility maximising approaches to
  rational action in resource-bounded agents;
* understand the main alternatives to utility maximisation approaches,
  including bounded-optimality, decision-theoretic planning, and
  partially observable Markov decision processes;
* understand the belief-desire-intention (BDI) model of agency, as a
  computational approach to rational decision making.

===================================================================
INTENDED AUDIENCE: WHY THE TUTORIAL IS OF INTEREST?


Rational action is at the heart of the design of autonomous
agents. When we construct agents, we want to be able to program them to
do our bidding in some environment. They have to perceive their
environment and to decide what to do, weighing the costs of action,
the possible outcomes of actions, and their overall goals. In other
words they have to act rationally. 


Despite the centrality of rational action, there is a lack of material
which introduces the central ideas in a easy to understand
format. Many of the ideas are buried in the technical literature. This
tutorial seeks to rectify the situation, giving a broad overview of
some of the key topics, while providing enough detail that delegates
come away with sufficient understanding to apply the techniques they
need.



==================================================================
BACKGROUND KNOWLEDGE REQUIRED


The tutorial is intended for any delegates who want to gain an
appreciation of the basic mechanisms for building agents capable of
rational action. 


The tutorial is suitable for any delegate with a basic grounding in AI,
(including such basic issues as search and knowledge representation)
but no detailed knowledge of agents, decision theory, rational action,
bounded rationality, or the BDI model.


===================================================================
DETAILED OUTLINE


MISSING!!!!



=================================================================
BIOGRAPHY OF PRESENTERS

Michael Wooldridge
       Department of Computer Science
       University of Liverpool
       Chadwick Building
       Liverpool L69 7ZF, UK
      +44 151 794 7790 (Mike Wooldridge)
      +44 151 794 3715    
m.j.wooldridge@csc.liv.ac.uk


Michael Wooldridge is Professor of Computer Science in the Department
of Computer Science at the University of Liverpool. He has been active
in the research and development of multi-agent systems for ten years,
gaining his PhD for work in the theoretical foundations of multi-agent
systems from the University of Manchester, UK in 1992.  Prof
Wooldridge has published many articles in the theory and practice of
agent-based systems, and has edited eight books in the area. He has
served on many program committees for conferences and workshops in the
area, and serves as an associate editor of the International Journal
of Autonomous Agents and Multi-Agent Systems (Kluwer), and an
editorial board member for the Journal of Applied AI (Taylor &
Francis). He was also the founding coordinator of AgentLink, the
ESPRIT-funded European Network of Excellence in the area of
agent-based computing (see http://www.AgentLink.org/).




Simon Parsons
       Department of Computer and Information Science
       Brooklyn College, City University of New York
       2900 Bedford Avenue,
       Brooklyn 11210, NY 11210, USA
      +1 718 951 5538
      + 1 718 951 4842
parsons@sci.brooklyn.cuny.edu

Simon Parsons is an Associate Professor in the Department of Computer and
Information Science at Brooklyn College, City University of New
York. Dr Parsons received his PhD from the University of London in
1993 and subsequently worked at the Imperial Cancer Research Fund, the
University of London, the University of Liverpool, and MIT. Dr Parsons
has written over one hundred papers on techniques supporting decision
making in intelligent systems and co-edited three collections and
written a monograph on the same subject. Dr Parsons is the editor of
the Knowledge Engineering Review, published by Cambridge University
Press, and was the recipient of a 1998 Younger Engineers Achievement
Medal, awarded by the IEE, for his work in informatics.

======================================================================