Professor

Mechanical Engineering, Northwestern Univ.

Dexterous manipulation remains one of the frontiers of robotics. No robot hand ever built closely matches the engineering specs of the human hand (strength, speed, sensorization, durability, etc), not to mention the versatility and performance of human manipulation. Rapid advances in AI give this century-old challenge new impetus, yet many difficult problems – and the need for fundamental advances – remain. In this talk, I’ll introduce a new U.S. National Science Foundation Engineering Research Center for Human AugmenatioN via Dexterity (HAND), and I’ll explain how the Center plans to advance dexterity. I’ll go on to discuss some of the work in our own group, focusing on the challenge of human-in-the-loop training of a dexterous robot; i.e., training via telemanipulation. Even the best available dexterous telemanipulators yield performance far below direct human manipulation. Although the reasons for this are not fully understood, the lack of high-quality haptic feedback is certainly one factor. I’ll describe several projects that aim to boost telemanipulation performance via tactile and kinesthetic feedback, and a separate project that aims to deliver high-bandwidth, high-resolution tactile feedback.