Social Spatial Behavior for 3D Virtual Characters

This thesis presents a social navigation solution for virtual game characters, capable of generating sensible human-like spatial behavior in social scenarios. In a social setting with several groups of virtual characters, our model generates group-joining, group-leaving and group-revisiting behaviors for an individual character. We consider interest as the main motivation behind character’s interactions with the groups. Thus, our social navigation model not only navigates the character toward interesting groups, but also continuously evaluates interestingness of groups and utilizes it to build group-leaving and group-revisiting mechanisms. In an engineering approach, we use the psychological knowledge on social spatial behavior to produce an internal representation of interest; then combine it with existing social navigation models to build our solution. We describe the two-stage implementation of our model, consisting of planning and realization of social spatial behavior. Finally we present simulation results of four testcase scenarios as proofs of concept for our model.