Lessons learned from software engineering multi-agent systems

The popularity of agent-based systems has increased rapidly in recent years. A number of software engineering frameworks/methodologies have been proposed to support multi-agent systems construction. How well are software engineering methodologies in support of multi-agent systems development? What are the strengths and weaknesses of such attempts? In this paper, we report our experience from the application of a software engineering framework for multi-agent systems. The framework is rooted in the belief desire intention (BDI) formalism and an extension of the Unified Modeling Language (UML). Our focus is to present lessons learned by applying this framework to an agent-based intelligent elevator system. We also identify some research issues.