Increasing accessibility in agent-oriented methodologies

Even though there is evidence of the suitability of the multi-agent approach to cope with the complexity of current systems, its use is not widespread in other areas of computing science, nor in industrial and commercial environments. This can be explained, particularly for agent-oriented methodologies, by the absence of key software engineering best practices. In particular, we have identified three groups of drawbacks that limit the use of agent-oriented methodologies: incomplete coverage of the development cycle, a lack of tools for supporting the development process, and a high degree of dependence on specific toolkits, methods or platforms. Although these issues negatively affect the applicability of the multi-agent approach in general, it is arguably for open systems that their effect is particularly noticeable. In this thesis, therefore, we aim to address the issues involved in taking existing agent-oriented methodologies to a point where they can be effectively applied to the development of open systems. In order to do so, we consider the combination of organisational design and agent design, as well as the methodological process itself. Specifically, we address organisational design by constructing a software engineering technique (software patterns) for the representation and incorporation of standard organisations into the organisational design of a multi-agent system. The agent design aspect is addressed by constructing an agent design phase which uses standard agent architectures through a pattern catalogue. Based on this, we develop a methodological process that combines the organisational and agent designs, and that also considers the use of iterations for making the development of a system more agile. This methodological process is exemplified and assessed by means of a case study. Finally, we address the problem of monitoring the correct behaviour of agents in an open system, by constructing a model for the specification of open multi-agent systems.