An information theoretical approach to crowd simulation

Official URL: http://192.168.1.20/record=b1293778 (Table of Contents)

Crowd constitutes a critical component in many virtual environment and entertainment applications. In this thesis, we propose methods to solve two distinct problems in crowd simulation domain; automatic camera control and adaptive behavioral modeling. As the basis of our methods, we develop a framework which uses information theoretical concepts to automatically construct analytical maps of crowd's locomotion, which are called behavior maps. The developed framework contains a probabilistic model of the scene to build behavior maps. In the first part of this thesis, we propose a novel automatic camera control technique which utilizes behavior maps to find interest points which represent either characteristic behaviors of the crowd or novel events occurring in the scene. The camera is updated accordingly to display selected interest points. In the second part of this thesis, we propose a novel behavioral model which uses behavior maps to control agents' behavior adaptively with agent-crowd interaction formulations. Our model can be integrated into crowd simulators and enhance their behavioral complexity. We made comparative analyses of the presented behavior model with measured crowd data and two agent-based crowd simulators.