Automated Generation of Legible Nonverbal Robot Expressions

This thesis develops automated methods for generating clear and adaptable nonverbal communication for robots collaborating with humans. It progressively integrates reinforcement learning (RL), large language models (LLMs), and optimization techniques to produce expressive robot motions and sounds, reducing reliance on time-consuming manual design. Together, these approaches form a set of frameworks that enable robots to convey their internal status more effectively to humans in collaborative human-robot interaction (HRI) settings. The research advances scalable, generalizable methods for nonverbal robot communication across diverse tasks and expressive modalities.