Comparing agent architectures in social simulation: BDI agents versus finite-state machines

Abstract

Each summer in Australia, bushfires burn many hectares of forest, causing deaths, injuries, and destroying property. Agent-based simulation is a powerful tool for decision-makers to explore different strategies for managing such crisis, testing them on a simulated population; but valid results require realistic underlying models. It is therefore essential to be able to compare models using different architectures to represent the human behaviour, on objective and subjective criteria. In this paper we describe two simulations of the Australian population's behaviour in bushfires: one with a finite-state machine architecture; one with a BDI architecture. We then compare these two models with respect to a number of criteria.