Evaluation and Comparison of Agent-Oriented Methodologies: A Software Engineering Viewpoint

Numerous agent-oriented methodologies that offer a rich pool of resources to support developers of agent-based systems have been proposed. However, the use of existing methodologies in industrial settings is still limited due to the large volume of methodologies, diversity of covered scopes, ambiguity in concepts, and lack of maturity. This makes it difficult for agent technology practitioners to choose the appropriate methodology that best fits their given development context. To eliminate such agent-based development bottleneck, it is important to introduce suitable methods for evaluating, comparing, and classifying agent-oriented methodologies in order to leverage their usage among practitioners. Having systems to evaluate methodologies can effectively help developers better understand existing methodologies, realize their benefits, outline their pros and cons, and assist practitioners with selecting the best-fit methodology for a specific agent-based project. In response, this paper proposes a novel criteria-based evaluation that is influenced by software engineering practices to assess and compare agentoriented methodologies. The proposed evaluation is derived from the software engineering body of knowledge (SWEBOK) and provides a simplified method to assess the coverage degree of an agent-oriented methodology with respect to major software knowledge areas such as the requirements and testing phases. We demonstrate the applicability of the proposed evaluation by applying it to three agent-oriented methodologies (PASSI, MaSE, and Prometheus) in the software engineering requirements and testing phases.