GAMES THEORY INVOLVEMENT IN THE SOCIAL ROBOTS? FACIAL EXPRESSIONS

This paper presents a method based on games theory and coordination graphs with the aim of improving the social robot facial expressions to perform human-like motions. It is an alternative to the neural networks or fuzzy logic methodologies. The difficulty is raised by the increased number of servomotors which should be controlled, and the translation of the motion of more than 40 face muscles to get the proper facial expression. There are evaluated four emotions using coordination graphs. The games are described by matrices of three or five players, and the dominance and equilibrium states are evaluated using Gambit software.