Nicholas V. Findler
Arizona State University
Department of Computer Science and Engineering, and Artificial Intelligence Lab
Box 875406
Tempe, Arizona 85287-5406
Phone: (602) 965-5934
Fax: (602) 965-2751
E-mail: nicholas.findler@asu.edu
URL: http://www.eas.asu.edu:80/~csedept/people/faculty/findler.html
A STATE SPACE FRAMEWORK FOR COORDINATION
Principal Investigator: Nicholas V. Findler
Co-Principal Investigator: Raphael M. Malyankar
Publication submitted on grant work:
Theoretical Tools for Intelligent Agent Societies; another is being prepared.
Summary
This project investigates mathematical model-based methods as a
framework for coordination in intelligent agent societies.
The first step is identifying the relevant aspects of intelligent
agent behavior. These aspects are mapped to state variables that
describe the behavior of a society of agents. Dependencies between
activities then manifest themselves as relationships between these
aspects or state variables. Coordination, the process of managing
dependencies between activities, is viewed as the exertion of control
over these relationships. Such control can best be understood when the
state variables' dynamics can be represented by constructs based on
logic or ordering (partial or total), or both. The question of imposing
logical structure and partial or total orderings on the identified
variables (and hence, on the corresponding aspects of intelligent social
behavior) is therefore being examined.
Once the relationships governing a society are known, coordination
problems are reduced to optimization, constraint satisfaction, and
constraint propagation problems. Existing techniques and research in
these fields then become available for formal approaches to coordination.
The project is aimed at demonstrating the viability of the
approach. This objective is being achieved by constructing mathematical
models of typical intelligent agent societies using the state variable
framework described above. Simulations of typical artificial intelligent
agent societies are built ab initio (separately from the mathematical
models). These simulations are used to test and validate the
mathematical model(s). Given a preliminary validation of the framework
based on simulation models, subsequent research may examine its validity
and applicability to various actual artificial intelligent agent societies.
The framework described will facilitate theoretical and simulation-based
approaches to coordination mechanisms, and produce valuable insights
into the fundamental problems of coordination. This should lead to (i) a
taxonomy of intelligent agent societies; (ii) a methodology for modeling
intelligent agent societies using such taxonomy; and (iii) foundations for
modeling and analyzing coordination using such methodology.
Return to ITO Workshop Abstracts
Return to ITO Workshop Home Page