Keep it Simple: Understanding Natural Language Commands for General-Purpose Service Robots

Abstract

Service robots are designed to perform useful tasks for humans, which involve managing and combining a variety of skills in the form of global and local plans to solve any given task. In this work, we propose a framework to process natural language commands for general-purpose service robots where the robot should perform an arbitrary spoken command requested by a non-expert operator. Our system uses a Natural Language Processing (NLP) parser and a Conceptual Dependency (CD) builder to create CD structures that an expert system can employ to generate a global plan that the robot will execute. An extra simplification step using Large Language Models (LLM) was tested and evaluated in order to improve the accuracy of the system. Finally, our system has been tested in challenging environments in robot competitions and we have achieved promising results.