Discovery and Imitation of Other Individual by Robot Based on Projection of a Self-model

Abstract

This paper proposes a novel model which enables a humanoid robot infant to discover and imitate other individual(e.g. human parent)focusing on the develop-mental process of human infants. In this work, the authors define "other individual" as an actor which can be predicted by a self-model. For modeling the developmental process of imitation ability, the following three approaches are employed. (i)Projection of a self-model for predicting other individual's actions.(ii)Mediation by a physical object between self and other individual.(iii)Introduction of infant imitation by parent. For creating the self-model of a robot, we apply Recurrent Neural Network with Parametric Bias (RNNPB) model which can learn the robot's body dynamics. For the other-model of a human, conventional hierarchical neural networks are attached to the RNNPB model as "conversion modules". Our target task is an "emulation" of moving an object. For evaluation of our model, human imitation experiments by the robot projecting its self-model were conducted. The results proved the adequacy of our model to enable the robot to recognize a human as an imitable being, and to imitate his object manipulations from his perspective.