Emotion Generation Model with Growth Functions for Robots

Abstract

In this paper, we propose an emotion model with growth functions for robots. Many emotion models for robots have been developed using Neural Networks (NN), which focus on the functions of emotion recognition, control, and expression. One problem that affects these emotion models for robots is the development of a “simplified” emotion generation algorithm. Users readily lose interest in “simple” systems. Most models have attempted to generate complex emotional expressions, whereas no previous studies have considered the “growth of a robot.” Therefore, we propose a growth model for emotions based on changes in the network structure of a self-organizing map. We also applied a multilayer perceptron NN to generate more sophisticated expressions of emotion using growth functions. This model generated a similar behavior to the concept of affective change described in genetic psychology. Our results showed that this emotion model was more suitable for producing a robot with growth functions based on a psychological model.