A Flexible Multi-Expert Meta-Architecture for Virtual Agents

Etienne de Sevin (1) , Quentin Reynaud (1, 2) , Vincent Corruble (1)

1 LIP6 - Laboratoire d'Informatique de Paris 6
2 Thales Services

One of the main motivations in behavioral simulation of virtual agents is the design of agent architectures which are sufficiently flexible, modular, consistent and generic to manage at the same time reactive and cognitive behaviors. We present in this article FlexMex, a flexible multi-expert meta-architecture for virtual agents. The main challenge lies in the structuration and organization of different modules addressing each a specific type of intelligence, each producing its own desires, goals, plans or motivations for behaviors. Our meta-architecture is independent of the module contents. The propagation of the behaviors has to be in a flexible and manageable manner to the decision process. We instantiate it in the context of autonomous pedestrians inhabiting a virtual city according to the real-time, scalability and complexity constraints.

Domaines: Intelligence artificielle [cs.AI]

Etienne de Sevin, Quentin Reynaud, Vincent Corruble. A Flexible Multi-Expert Meta-Architecture for Virtual Agents. Workshop Affects, Compagnons Artificiels et Interaction, Nov 2012, United States. ⟨hal-00876089⟩