Structured learning of a partner robot based on perceiving-acting cycle

This paper proposes a learning method for constructing the interrelation between a partner robot and human from the psychological point of view. First, the concept of perceiving-acting cycle is applied for learning the relationship between perception and action of a partner robot. Furthermore, modular neural networks are used for specifying and learning action segmentations through the interaction with a human. In this paper, we focus on the visual system for perception of the partner robot. A perpetual module is prepared for extracting specific perceptual information according to the current state. Some experimental results show the interactive learning can be done in the communication process between a partner robot and a human.