Automatic generation of moving crowd using chaos model

Abstract

Concerns the automatic generation by computer of moving crowd graphics. Motion of animal flocks has been studied, but seldom human crowds. We are concerned with modeling and generation of a crowd in a 3D virtual city to increase reality. Virtual cities often have few inhabitants, and those are controlled by the users or only repeat behavior based on strict rules. For a more natural atmosphere, many data and rules are needed. A general view of a crowd reveals significant characteristics. Though the behavior of each figure is independent, the crowd, as a huge mass of it, has a pattern as a whole. Chaos is very convenient to represent such features. It is controlled by a few parameters. For instance, since chaos is influenced quite sensitively by initial value, it can produce various nonmutual values that decide the next actions of the figures. Moreover, these values form themselves into an attractor. As the attractor is transformed by changing a parameter, we can easily control the shape of the crowd. Although random numbers are often applied to such purposes, it is difficult to create and control such patterns. Utilizing these characteristics of chaos, we constructed the models of the crowd in the street and the folk dance. The prototype system was developed for evaluation and it was confirmed that users felt easy to generate and control the crowd.