Automatically Generating Virtual Humans using Evolutionary Algorithms

Abstract

Abstract Virtual Humans are used in many applications either as an embodiment of a real person (an "avatar"), or under the control of a computer program (an "agent" or "non-player character"). The automatic generation of Virtual Humans is a challenging problem if they are to look both plausible and unique within a population. We present an approach which exploits the power of Evolutionary Algorithms (EAs), and provide illustrative examples of how our methods may be realised within the context of surface-based model geometry.