A Normative Model for Behavioral Differentiation

In multi-agent systems simulations, reproducing realistic behaviors is a crucial issue. Their variety and consistency are important factors, usually not specifically considered. In this paper, we propose a behavioral differentiation model designed (1) to generate various and consistent behaviors, and (2) to control the determinism of this generation. Based on a normative system and a nondeterministic generation engine, it allows users adapting it easily to their various needs. Finally, we show its application to the traffic simulation software developed and used at Renault, SCANeR(c) II.