Vers une architecture pour la simulation microscopique de foule

We present in this study an architecture for microscopic crowd simulation, i.e. which simulates each entity by an autonomous agent. We treat in this article all the necessary steps for such a process. First, we propose a representation of the environment which is computationally efficient, automatically calculated, and informed with pre computed data which can be used by the entities at low cost. Then, we propose a description engine for the interactive objects of the simulation, which enhances the behavioural potential of the simulated entities. For these interactions, we emphasise the fact that the interactive objects are directly integrated inside the path planning process of the humanoid. This path planning process is based on an individual knowledge base about the environment, takes into account several decision criteria at the same time, and manages different completion conditions which may be as well geographical as conceptual. Finally, we show how these models are integrated in our architecture to obtain a crowd simulation platform.