Modeling adaptive perception system of virtual agent with emotion based on Q-learning

Human make behavior by perception, modeling believable perception for a virtual agent is a meaningful topic in computer games. A model of adaptive perception of a virtual agent with emotion is proposed based on Q-learning method; the formulas of emotion value are presented. A demo system is realized on PC; an agent with the model can dynamically control the range of the perception according to virtual environment, and express believable emotions.