The Dragon Dialogue Navigation Port project adapted a dialogue-driven navigation system for the Stretch robot using ROS2 and Python. The system integrated perception, planning, and dialogue modules to enable natural language commands for navigation and fetching tasks in real-world environments.

Key optimizations included asynchronous execution and embedding caching within the NLU, CLIP, and VQA pipelines, significantly reducing command latency and improving responsiveness. Real-world experiments demonstrated reliable voice-to-motion execution, obstacle-free navigation, and safe task performance, underscoring the potential of dialogue-based AI for intuitive human–robot interaction.

Technologies Used

• ROS/ROS2
• Docker
• Python
• Natural Language Understanding
• Real-Time Systems Optimization