AGENT ROLE LOCKING (ARL): THEORY FOR MULTI AGENT SYSTEM WITH E-LEARNING CASE STUDY Salaheddin J. Juneidi, George A. Vouros Department of Information and Communication Systems Engineering School of Sciences University of the Aegean ABSTRACT Advances in methods and techniques for software engineering are crucial for industrial and commercial applications, as these systems are required to operate in increasingly complex, distributed, open, dynamic, unpredictable, and inherently highly interactive environments. This article presents Agent Role Locking (ARL) theory supported by a case study as an example of engineering complex systems with autonomous entities, and managing their inherent complexity during analysis, design and implementation. Agent Role Locking (ARL) theory provides a new conceptualization of the relation between agents and roles in MAS. ARL calls on modification of UML interaction diagrams by introducing AIP diagram to preserves the distinguishing characteristics of agent software entities. KEYWORDS Agent Oriented Software Engineering (AOSE), Multi Agent System (MAS),AUML, super role, atomic role, Agent Interaction Protocol (AIP). 1. INTRODUCTION Agent oriented software engineering (AOSE) has been introduced as a new paradigm for engineering complex software systems. Indeed, advances in software engineering are crucial for industrial and commercial applications, as software systems are required to operate in increasingly complex, distributed, open, dynamic, unpredictable, and inherently highly interactive application environments. Agent based computing appears to be very promising for building such systems, integrating entities with agency characteristics into software applications [Bradshaw 1997, Wooldridge 1995]. Although many tools and applications are considered as agent software integrated applications [Odell 2002], but they mostly, do not follow any formal software engineering process. Recent surveys and evaluation of the most prominent approaches to AOSE [Teveit 2001,Juneidi & Vouros 2004] in terms of software engineering criteria as well as agent based computing criteria [Juneidi & Vouros 2004,Ardis 1995,Shehory 2000], reveal that non of the given methodologies are fully satisfactory to AOSE . Most AOSE approaches agree on the importance of roles during analysis [Jennings 2000,Bradshaw 1997,Odell 2003,Odell 2004]. However, roles are not reflected in the final system design and implementation, which creates an engineering gap. This gap affects the development of complex systems with autonomous entities, restricts the level of complexity that can be handled during analysis, design and implementation of systems, resulting into systems with rigid structure, and with no adaptation abilities. As stated in [Juneidi 2004] a new engineering paradigm and a new way of thinking must emerge to adopt agents smoothly into the software development process. This paper introduces Agent Role locking (ARL) theory which is an approach to AOSE that integrates role-oriented and goal-oriented system analysis, emphasizing on agents, roles and on their interplay, clearly distinguishing between agent and role entities. ARL has its own view of Multi-Agent System (MAS), gives agents entities the freedom to perform any “appropriate” role. ARL has its own methodology for analyzing MAS by defining environment; organizations belong to the environment, super roles and atomic roles of these organizations. It defines a new dynamic structure of agents’ interactions, Agent Interaction Protocol ISBN: 972-99353-6-X © 2005 IADIS 442