A multiagent based fitness function for evolutionary architecture

Previous multiagent system (MAS) methods for modelling discrete pedestrian dynamics have been shown to be brittle in the face of increasing density. We present an alternative method, which incorporates the novel idea of a local-to-global linkage for path discovery and add a heuristic to control critical information. This method has been successful in reproducing high-level crowd patterns and we present a preliminary approach to the evolutionary design of building architecture using a crowd model based on this method. We present a genotype /spl rarr/ fitness mapping and results of evolutionary runs in an exit sign layout scenario. We discuss important issues and how these may affect future research directions.