Study On Behavior Modeling Of Virtual Characters In Intelligent Virtual Environments

Intelligent virtual environments has currently been a new research direction of virtual reality technology, as the interdisciplinary subject of virtual reality, artificial intelligence and artificial life, and it focus on simulating the virtual characters with life characteristic in intelligent virtual environments. This paper was partially supported by Scientific Research Fund of Heilongjiang Provincial Education Department(No:11521006), which studies primary technologies of behavior model architecture and behavior model, and establishes cognitive model to approximate by artificial neural network for virtual characters.This paper introduces primary contents of intelligent virtual environments and behavior model, analyzes the hierarchy model of virtual characters, behavior model of the brain architecture and behavior model based on Agent and so on, then designs behavior model which combined their specialty and excellence to program cognitive layer and behavior layer of virtual characters respectively, so that virtual characters could adapt the change of virtual scene quickly and neatly.Behavior model is tested by adopting the experiment of the herding and the driving which are implemented by using finite state machine and artificial neural network respectively. The realization of cognitive model which is approximated by artificial neural network includes acquiring training samples, pre-processing samples and post-processing samples, neural network training and application. Training samples which are dealt with conflicting samples and normalized are acquired by program controlling and manual controlling, so that the proceeding of training could reflect the mapping relation of inputs and outputs.The experiment shows virtual characters whose cognitive models are approximated by artificial neural network are more intelligent and independent, compared to the traditional characters.