Analyzing a Multi-Agent-System Decision Architecture Aiming to Model the Behavior of Virtual Humans

In this paper a novel approach on how to model the decision-making of virtual humans in the domain of multi-agent-systems is analyzed. The developed decision architecture embeds strategic behavior for intelligent humanoid agents (Hoogendoorn & Bovy 2004). The aim is to solve the action selection problem in the domain of naturalistic decision-making related to pedestrian movements. Embedding individual, motivational, reactive and proactive dynamic decision-making in virtual humans is the key to fulfill essential characteristics to model high level cognitive behavior of synthetic humanoid agents (de Sevin & Thalmann 2005). The decision architecture is based on the activation of motivations for the virtual humans and uses Agent-Based Modelling (ABM) and the methodology of System Dynamics (SD). A set of motivations is defined for each virtual human and activation processes are dependent on agent-intrinsic needs modeled as motivational reservoirs and agent-extrinsic multipliers aggregating information coming from the environment. The decision architecture fully distinguishes action selection from navigation and locomotion and fulfills therefore the three-layer architecture originally coming from the area of robotic research (Blumberg & Galyean 1995). The analysis of the developed decision architecture is important for calibration and validation purposes of multi-agent microworlds that enable to conduct risk-free experiments in which realistic virtual human beings – for instance virtual event visitors, virtual airport and train station passengers or virtual parade attendees – need to be embedded. The sophistication of the developed commonly called action selection mechanism (Tyrrell 1993) is done for a case study of a music festival.