Options
2008
Conference Paper
Titel
Learning agent-behavior for agent-based simulation using genetic algorithms
Abstract
Agent-based simulation is a convenient method to analyse different strategies for specific scenarios, such as evacuating a crisis region or driving a convoi through unsecure areas. The quality of the simulation results depends heavily on the accuracy and diversity of the modelled agent-behavior. This paper presents a method to learn such agent-behavior automatically in order to acquire near-optimal strategies.