Developing multiagent systems: The Gaia methodology

Reviewer: Rafael Corchuelo

In this paper, the authors present a new version of the Gaia methodology, which focuses on multiagent systems (MAS), and was introduced originally by Wooldridge et al. [1]. This new version is based on the so-called organizational metaphor, which provides several modeling tools to deal with open MASs. This metaphor leads to a general characterization, in which different groups of agents form communities that interact and communicate with each other and their environment. The authors argue that the metaphor is general enough to be used in typical MASs, since it helps reduce the conceptual distance between the software system and the real-world system it needs to support. However, this metaphor can only be applied to systems that are highly predictable and reliable; otherwise, the well-known ants, societal, or physical metaphors might be more suitable. The authors make crystal-clear what the advantages of Gaia are with respect to four methodologies, but they draw a veil over many other well-known proposals. I recommend that they pay a visit to http://www.science.unitn.it/~recla/aose/ to examine an up-to-date collection of about 30 such methodologies. Summing up, I think that Gaia provides a good set of modeling artifacts, but that it is not complete. No guidelines for requirements elicitation or implementation are provided, no supporting tools are available, and it lacks a process. Online Computing Reviews Service

Reviewer: George K. Adam

This paper's focus is on key issues related to the identification of appropriate abstractions for agent-based software engineering, and on the definition of a suitable methodology for the analysis and design of complex applications in terms of multiagent systems (MASs). The work reported on in the paper is appropriate for the current state of this field; the paper describes the organizational abstractions that are necessary for agent-oriented software engineering. This differs from previous work in that it focuses on the particular organizational abstractions that are appropriate for the software engineering process, rather than on the specific technologies to implement them. The authors argue that agent-oriented computing is an appropriate software engineering paradigm for the analysis, design, and development of many contemporary software systems. To demonstrate this, they provide a lengthy and detailed description of their key concepts for agent-based computing and multiagent systems development methodology. An extended literature review of related work is provided; the authors present many of the problems encountered in developing multiagent systems, and clearly state their extensions. Their concepts for the analysis and design of a multiagent system are documented in a methodology called the Gaia methodology. The authors illustrate their arguments by using, as examples, two real-world applications: a manufacturing system, and a conference management system. Although a considerable part of this work addresses the key concepts of agent-based computing, as they pertain to agent-oriented software engineering, the paper itself is rather important, due to its development and provision of a methodology, centered around organizational abstractions, for the analysis and design of open multiagent systems. Potential applications of this could be the development of software applications for the management of open multiagent systems, such as electronic transactions via the Web. It is not easy, however, to identify or predict the potential results, which may be positive. Finally, the authors claim that much work remains to be done on advancing the widespread acceptance of agent-based computing as a software engineering paradigm, and to this end, they make suggestions for future work. This is a research study intended for researchers interested in developing multiagent systems. Although, in general, this is a well-written paper, I would expect, in such a lengthy work, to see some fragments of software implementations (code examples of applications), if only in an appendix. This would have helped readers to better understand the authors' views on agent-oriented software engineering. The level of novelty of this work, in relation to the extensive comparative literature review included, appears to be satisfactory. The credibility of this work-although it seems to be validated initially by the investigation of two real-world applications, and later on with a well-documented description of the Gaia methodology-could be extended further with more practical example cases. Online Computing Reviews Service