ISAC Humanoid: An Architecture for Learning and Emotion

A report on the development of a software architecture to facilitate interactions between a person and a humanoid robot is given. The architecture employs independent, asynchronous software agents that run in parallel on a set of PCs in a local area network. It comprises four major components: a Self Agent which controls the robot’s general interactions with the world and monitors its own status, a human agent which models the person and controls the robot’s interaction with her or him, a spatiotemporal short-term memory called the Sensory EgoSphere, and a long-term memory called the DataBase Associative Memory. The structures of these components and a description of their interactions is given. Current research efforts in learning and emotion are described. Outstanding problems are listed. Keywords– Robotics, Humanoids, Software agents, Affective Computing, Kansei, Sensory Ego-Sphere.