|Ph.D Student||Sturm Arnon|
|Subject||Developing and Evaluating an Object-Process|
Methodology-Based Multi-Agent Systems Framework
|Department||Department of Industrial Engineering and Management||Supervisors||Professor Dov Dori|
|Dr. Onn Shehory|
In recent years, researchers and practitioners have both recognized the advantages of applying the agent-based paradigm for systems development. Yet, the number of deployed commercial agent-based applications is small. One of the reasons for this is the lack of industrial methodologies for agent-based application development. Analyzing the problems within the domain of agent-oriented methodologies, we developed and applied a framework for comparing and evaluating agent-based methodologies. The evaluation results showed that agent-oriented methodologies are deficient in addressing software engineering criteria. Based on this finding, we devised a new agent-oriented methodology based on the Object-Process Methodology. The new methodology extends OPM with the capabilities to support domain analysis. We adopted OPM to gain method flexibility, because there is no agreement on a standard for agent-oriented methodology. Utilizing the domain extension of OPM, we devise a multi-agent system metamodel, which includes the definitions and functionality of the building blocks required to model the MAS domain. We exemplified the use of the method in expressing system structure and behavior, communication aspects, and mobility. Evaluating our approach, we compared OPM/MAS with known agent-oriented methodologies: GAIA, MaSE, and Tropos. We compared them using the framework we had developed by a students’ survey and by an external framework. All three evaluation techniques showed similar results, and indicated that OPM/MAS is preferable over other methodologies, mainly with respect to software engineering criteria.