Ontological Foundations for the Design of Integrated Multiagent Architectures

Hong Zhang, Dan Jong Kim, Ram Ramesh

{hz5, djk5, rramesh }@acsu.buffalo.edu

Management Science & Systems

School of Management

State University of New York at Buffalo

Abstract

Research on ontology is becoming increasingly pervasive in the computer science and information systems communities.  Founded on pragmatics, theories and even empirical determinations, an ontology basically supports not only the design process but also the implementation and deployment of information systems. While an ontology provides guidance on information capturing and representation strategies, a framework actually realizes it in a usable form. In general, a framework provides a complete set of generic applications with representation formalisms that a system designer can customize or extend to address particular needs.

We are developing an ontological framework for designing multi-agent systems. Termed RITA (Role-Infrastructure-Taskflow Agents), the framework provides a comprehensive archetype for designing multi-agent systems in general.  RITA identifies of four basic constructs: role, infrastructure, taskflow, and agent.  The role construct defines the services and responsibilities of the roles involved.  The infrastructure construct identifies the internal state and knowledge/resource base of an agent.  The taskflow construct determines the internal/external environments and interactions in a multi-agent system.  The agent construct specifies the different agent types. Our objective in this research is to develop a design tool that designers can use to “stamp out” agents and integrate their schemas to suit their application needs.  In the paper, we present the basic architecture of the proposed framework.

This work has been motivated by several reasons. Some of the most important ones are: (1) the emergence of multi-agent systems as a reality, (2) proliferation of agent architectures and even some standards which have serious compatibility and interoperability problems with other objects in a computing environment, (3) non-existence of viable frameworks for practical and efficient design of agents. Interestingly, the archetypal components of RITA essentially capture most of the pragmatics and theories underlying agent systems, and thus rendering them universal.  This work is founded on and extends the following research streams developed by us over the years: enterprise computing frameworks, object-oriented formalisms, collaboration systems and database theory.

Download paper.