Modeling software agents identification and design

Agent-based computing has materialized as a powerful technology for developing complex software systems [1]. Having emerged, like so many other disciplines, from artificial intelligence, it is now a melting pot of many different research areas (artificial intelligence, software engineering, robotics, and distributed computing). Agent-Oriented Software Engineering (AOSE) stems from a line of research including the autonomous software agent as an autonomous element, with reactive and proactive social ability, trying to accomplish its own task [2], the multiagent system (MAS) as a set of autonomous agents that interact with each other, each representing an independent focus of system control [3], and agent societies, where the social role of the agents and social laws delimit agent operation [4]. AOSE is well suited for tackling the complexity of today's software systems [5] This line of research resulted in a new paradigm: the agent paradigm. Matured in research laboratories by developing resources and applications, with the appearance of architectures, models and methodologies, this paradigm is now acquiring the status of an engineering discipline. This effort to convert agent-oriented system development into a genuine engineering discipline capable of large-scale MAS development has led to a variety of methodological proposals [6–8]. Although they all have played an important role in establishing this field, they have omissions that are an obstacle to the formulation of a natural process for developing an MAS system or an agent society from system requirements. A good methodology should not force a given architecture (object-oriented, agent-oriented, etc) upon developers from the beginning. It is the system specifications analysis that should point developers towards the best suited architecture for solving the problem. In this paper, we describe a methodological approach for naturally producing an MAS or agent society from the system requirements. In section 2, we explain what problems MAS developers face. Section 3 details the omissions of current MAS development methodologies. Section 4 describes the structure of the proposed SONIA methodology. Section 5 outlines a case study: the ALBOR project. Finally, section 6 states the conclusions on the natural development of MAS.