A Methodology for Designing Agent-Based Models: Agent from « UP » for Complex Systems

摘要

The designing and implementation of an agent-based model for a system called complex are often guided by clear objectives and often meet a need. Such kinds of models are often used to help for understanding real systems mechanisms and to be at last a decision support. The evolution of a theory that is supported by validation stages through a simulation needs to be guided by a methodology. This one will make possible to organize the work and communicate with the end users in order to achieve the objectives with a reasonable time. We propose here a methodology that extends the Unified process. It provides a set of tools to carry out the modeling work successfully. This methodology is iterative and incremental and is guided by a collaborative approach based on communicating with the end users in one side and specialist in another side. The communication is also supported by tools and a common ontology. The present method formalism is based on Agent UML combined with GAIA. A practical case is also presented with an agent based model on the Rift valley fever.