This research bridges the gap between natural human speech and robot motion control, enabling robots to respond consistently to varied verbal instructions.

• Developed a Speech-to-Trajectory (S2T) framework that transforms diverse verbal commands into consistent robot movements
• Implemented a robust training method combining human demonstrations with data augmentation
• Achieved 84.3% success rate in real-world testing with untrained users
• Created a novel dataset of 600+ verbal instructions paired with robot trajectories

This breakthrough enables robots to work alongside humans in factories, healthcare, and homes without requiring users to learn specific command phrasing, making human-robot collaboration more intuitive and accessible.

Speech-to-Trajectory: Learning Human-Like Verbal Guidance for Robot Motion