Abstract

Recently, autonomic computing has been proposed as a promising solution for software complexity in IT industry. As an autonomic approach, the Reactive Autonomic Systems Framework (RASF) proposes a formal modeling based on mathematical category theory, which addresses the self-* properties of reactive autonomic systems in a more abstract level.
This thesis is about the specification and implementation of the reactive autonomic systems (RAS) through multi-agent approach by laying emphasis on the fault-tolerance property of RAS. Furthermore, this thesis proposes a model-driven approach to transform the RAS model to agent templates in multi-agent model using Extensible Stylesheet Language Transformation (XSLT). The multi-agent approach in this research is implemented by Jadex, a high-level Java-based agent programming language. The intelligent agents are created in Jadex based on the Belief-Desire-Intension (BDI) agent architecture. The approach is illustrated on a case study.