Evaluation of modeling techniques for agent-based systems

Abstract

To develop agent-based systems, one needs a methodology that supports the development process as common in other disciplines. In recent years, several such methodologies and modeling techniques have been suggested. An important question is, to what extent do the existing methodologies address the developers' needs. In this paper we attempt to answer this question. In particular, we discuss suitability of agent modeling techniques to agent-based systems development. In evaluating existing modeling techniques, we address criteria from software engineering as well as characteristics of agent-based systems. Our evaluation shows that some aspects of modeling techniques for agent-based systems may benefit from further enhancements. As we show, these aspects include distribution, concurrency, testing and communication richness. We also find space for (relatively small) improvements in aspects such as the refining of the models throughout the development process and the coverage and consistency checking of the suggested models.

FAML: A Generic Metamodel for MAS Development

“…Tran and Low [48] developed an evaluation framework using various feature analysis frameworks applied to evaluating AOSE methodologies [55], [56], [57], [58], identifying and integrating four categories of evaluation criteria: Process-Related, Technique-Related, Supportive Features-Related, and Model-Related. Only concepts represented by ModelRelated Criteria refer to the workproducts of a methodology.…”

A Preliminary Comparative Feature Analysis of Multi-agent Systems Development Methodologies

“…The framework proposed in [1] was developed from the synthesis of various existing feature analysis frameworks, including those for evaluating conventional system development methodologies -namely [8], [9], [10] and [11], and those for evaluating MAS methodologies -namely [12], [13], [14] and [15]. The framework therefore improves on the existing work by extensively assessing both agent-specific (or MAS specific) and generic system engineering dimensions.…”

Suitability assessment framework of agent-based software architectures

“…[24,59,62,64,65,68]. For instance, [24] identifies methodologies that may address problem domain-specific criteria for autonomy, concurrency and distributedness.…”

“…Other examples include [65] identifying methodologies that address domain-specific criteria for communication and [64] identifying a methodology's applicability to multiple domains in terms of "expressiveness" which generalizes domainspecific criteria. However, it remains that [24,59,62,64,65,68] have not provided criteria that assesses the suitability of a methodology (or MAS) for a given problem domain. These works provide particular views to assess methodologies e.g.…”