Behavior Learning of Face Robot Based on the Characteristics of Human Instruction

Abstract

We have found a difficulty in the learning of a human-friendly behavior in a face robot by means of human instruction. The difficulty is mainly spotted at the learning algorithm of the robot not taking account of characteristics of human instruction. This paper suggests an effective learning algorithm constructed by the results obtained through analysing characteristics of human instruction, and evaluates its effectiveness by computer simulation and human-face robot interaction experiments.