Evaluation of modeling techniques for agent-based systems

TLDR

Developing agent‑based systems requires a methodology similar to other disciplines, and several such methodologies and modeling techniques have been proposed recently. The paper seeks to determine how well existing methodologies meet developers’ needs for agent‑based systems. We discuss the suitability of agent modeling techniques for agent‑based systems development and evaluate them using software engineering criteria and agent‑based system characteristics. The evaluation indicates that while some aspects of modeling techniques benefit from further enhancements—particularly distribution, concurrency, testing, and communication richness—there is also room for modest improvements in model refinement, coverage, and consistency checking.

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.