Hierarchical emotional memory-based human robot social interaction framework

Abstract

A social robot is a robot that operates autonomously with the ability to communicate with people or other objects. Recently, The growth of the social robot market has been driven by the development of Artificial Intelligence, IoT, and Cloud technologies, and robots are getting closer to us in order to help people in the family and everyday life. As robots get closer to people, physical and mental social interactions are beginning to appear very frequently. Therefore, robots that interact with human beings need a memory structure that can efficiently learn human interaction experience. In social relationships, emotional interaction is at the core. Robots need the ability to learn emotional experiences effectively with other experiences in social interaction with humans. When the robot learns emotional experiences, it should be able to predict the emotional situation based on the experiences that were learned when similar experiences were repeated, and to provide appropriate services to the users. In addition, as the robot interacts with the user, the robot should also be able to provide the intuitive emotional expression and services to the user with the emotion of the robot itself. In this paper, I propose a hierarchical emotional episodic memory which can efficiently memorize emotional interaction experience with users by using Adaptive Resonance Theory (ART) based memory network. In addition, I propose a robot emotion generation and expression memory structure how robot's own emotions are generated and used for learning, and to show emotional expressions to users. Finally, to develop a framework for the social interaction of humans and robots, I have developed a humanoid type robotic platform with arms and head, and developed a software architecture of social interaction. In this paper, I confirmed the results of experimental analysis of the hierarchical emotional episodic memory, suggested construction of the theoretical model for emotional generation and representation memory, and demonstrated the human robot social interaction framework in which various applications work with AI technologies and robotic hardware.