2008 | OriginalPaper | Buchkapitel
Training Time and Team Composition Robustness in Evolved Multi-agent Systems
verfasst von : Russell Thomason, Robert B. Heckendorn, Terence Soule
Erschienen in: Genetic Programming
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Evolutionary algorithms are effective at creating cooperative, multi-agent systems. However, current Island and Team algorithms show subtle but significant weaknesses when it comes to balancing member performance with member cooperation, leading to sub-optimal overall team performance. In this paper we apply a new class of cooperative multi-agent evolutionary algorithms called Orthogonal Evolution of Teams (OET) which produce higher levels of cooperation and specialization than current team algorithms. We also show that sophisticated behavior evolves much sooner using OET algorithms, even when training resources are significantly reduced. Finally, we show that when teams must be reformed, due to agent break down for example, those teams composed of individuals sampled from OET teams perform much better than teams composed of individuals sampled from teams created by other methods.