METAOSE: Meta-analysis for Agent Oriented Software Engineering

In view of the existing methodologies for agent-oriented software engineering (AOSE) the development of multi-agent systems (MAS) is still a difficult challenge. The learning curve for mastering MAS model properties and problem's domain characterization is steep. The software engineers hesitate to use MAS since choosing a MAS-based methodology leads to fix the type of models that will be involved without inter-methodologies reusability. We think these are some of the reasons restricting the dissemination of multi-agent methods. This paper presents a self-organized MAS-based intelligent process to assist the engineer. This process comprises three stages: problem features and domain characterization, MAS components matching and meta-analysis. It aims to reduce the difficulties of starting an MAS-based solution to disseminate the use of MAS. The process is presented as a guiding tool for the engineers, especially for those less experienced in MAS, creating a path in the preliminary stage of the MAS conception. This approach is situated just before choosing a MAS framework or methodology to deploy a solution. We show how it works and we present a study case to compare our preliminary results.