UML for Behavior-Oriented Multi-agent Simulations

Developing multi-agent simulations seems to be rather straight forward, as active entities in the original correspond to active agents in the model. Thus plausible behaviors can be produced rather easily. However, for real world applications they must satisfy some requirements concerning verification, validation and reproducibility. Using a standard framework for designing a multi-agent model one can gain further advantages like fast learnability, wide understandability and possible transfer. In this paper we show how UML can be used to specify behavior-oriented multi-agent models. Therefore we focus on activity graphs and the representation of different forms of interactions in these graphs. 1 Motivation A multi-agent model captures the behavior of a system on the level of the active entities and maps them onto simulated active entities. Thus, on a rather coarse grain level of analysis, it is far more intuitive to describe the model in terms s and roles, using goals or activities, t...

• Type: Conference
• Year: 2001
• Where: CEEMAS
• Authors: Christoph Oechslein, Franziska Klügl, Rainer Herrler, Frank Puppe
• Keywords: Active Entities, CEEMAS 2001, Intelligent Agents, Multi-agent Model, Multi-agent Simulation