Poster Abstract: Contactless Friction Sensing in Robotic Systems via Fusing Sub-Terahertz Wireless Signals and Vision

Sensing the coefficient of friction (COF) is crucial for robotic and Cyber-Physical System applications, including grasping. We introduce RoboTera, a novel system for non-contact COF estimation using sub-Terahertz (sub-THz) perception in robotics. Unlike tactile sensors that require direct contact, our approach leverages sub-THz signals with sub-millimeter wavelength to capture surface roughness characteristics as an essential factor in non-contact COF inference, that conventional imaging modalities like cameras and LiDAR cannot detect. Our system enables precise COF inference by integrating sub-THz-estimated roughness with image-based material classification. Further, we exploit COF inferences to identify stable grasp configurations and improve grasping performance. Experiments show over 92% accuracy in COF estimation, with a 31.8% improvement in grasp success rates in real-world robotic tasks.