Extracting Semantics: LLM-Guided Automatic Population of Robot Ontology from URDF

The article presents a method for automatically generating robot ontologies from Unified Robot Description Format (URDF) models by leveraging Large Language Models (LLMs) to infer semantic relationships. The approach utilizes majority voting from multiple LLM queries and applies syntactic and schema-level validation to ensure the outputs conform to ontology constraints. This technique addresses the bottleneck of manual ontology construction, facilitating the integration of grounded knowledge necessary for effective human-robot interaction.