An Architecture Based on M5P Algorithm for Multiagent Systems

Alex MachadoDavid CarvalhoEsteban CluaCristiano G. DuarteMarcos V. MontanariWillian M. P. Reis

Character-based interactive storytelling, life simulation and game difficulty dynamic balancing are examples of topics that need to deal with autonomous agent evolution. Although the commercial appeal of such kind of feature, the research of new behaviors emergence in virtual societies is restricted to intelligent agents that do not learn. This work proposes a novel architecture of a multiagent system based on the M5P algorithm for generation of emergent behaviors in complex virtual worlds. Based on our obtained results, we describe the advantages of the reinforced learning based on numerical classifiers when compared with traditional interaction adaptations.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: