Human-Centric Autonomous Systems With LLMs for User Command Reasoning