Interactive Dialogue for Behavior Teaching to Robots based on Primitive Behaviors with Fuzzy Voice Commands

This paper proposes a method to teach a robot interactively based on behavior primitives. Fuzzy voice commands are used to activate the behaviors in a fuzzy coach player system scenario. An intention control module is implemented in the interaction manager to handle the interactivity and a hierarchical behavior memory is introduced to store the learned behaviors. The proposed system is used to learn complex behaviors based on simple posture movements of a manipulator. The fuzzy voice joint commands are used to trigger the primitive behaviors which represent the single joint movements. The proposed method is demonstrated with a PA-10 robot manipulator.