Integrating Action Robot Ontology for Enhanced Human-Robot Interaction: A NAO Robot Case Study

This paper presents an approach that allows the NAO humanoid robot to respond to a question from a user and gesticulate depending on the text that it is saying. The question might also be an action command spoken by the user that the robot recognizes and executes. A Large Language Model is integrated within the approach to provide the question-answering capabilities. For the action commands, we have used an action robot ontology that we have defined in past work. We have extracted the pertinent classes and individuals and generated a three-word string for each action that is matched semantically with the user’s text. Moreover, as far as the action commands are concerned, the system can work in two modes: STATELESS and STATEFUL. When in STATEFUL mode, the robot knows its current posture and performs the command only if it is compatible with its current state.