Active Clarification in Human-Robot Interaction via Object State Perception and Reasoning

Abstract

Embodied agents operate in human-centered environments such as homes and restaurants, where Human–Robot Interaction (HRI) is often affected by uncertainty, vagueness, and ambiguity. These challenges become more pronounced when interactions involve fine-grained object states (e.g., clean vs. dirty) rather than object categories alone. This thesis investigates how embodied agents can acquire and process object-state knowledge to enable robust interaction.