Yitian Shao

Wide-Bandwidth Tactile Sensing for Machines

Humans naturally exploit rich tactile information from wide-bandwidth and distributed dynamic tactile signals (DTS) during touch. In comparison, robotic tactile sensing still falls short of human sensing capabilities. This talk will present research on how structural design and deep learning can enhance robotic tactile perception through wide-bandwidth dynamic tactile sensing, including: 1) efficient large-area tactile sensing enabled by customized robot finger structure utilizing sensors mounted on its base to estimate contact location and force across the entire finger surface; 2) delicate peg-in-hole maneuvers enabled by robot inferring contact conditions based on captured transient DTS produced from contacts between the grasped object and the insertion slot; 3) enhanced naturalness of haptically-rendered surface textures by synthesizing DTS captured by biomimetic finger using data collected by rigid robot end-effectors.