Intelligent Hands
The past few decades have been filled with extraordinary feats in robotics -- from image segmentation and pose estimation to SLAM and legged locomotion. Over this time, we have continued to see manipulation as a particularly difficult task to tackle, as it still remains a largely unsolved problem today. Taking a step back, the promise of robot manipulation is that we, as roboticists, are able to build a complete, complex system capable of both delicately and forcefully interacting with the world. Service robots of the future will be tasked with an arrange of tasks – from handling light, delicate glassware to transporting a jug of laundry detergent. This dichotomy has been particularly challenging to address as we battle the physical limits of scaling factors (motors, sensors, bearings, etc.). Hands are the epitome of this venture. How can we even begin to mimic the control, the sensing, and ultimately, the dexterity of the human hand?
In this talk, I want to discuss the history of hand design in a bit more detail. I will present approaches of the past, what worked and what didn't, and approaches of the present. As we analyze this further, we will start to notice a cycle of recurring themes, and it is healthy for us in the robotics community to think about these in more detail. To end, I will share with you some thoughts on what it may take to build more "capable" manipulators and end effectors.
Bio:
Andrew (Andy) Morgan is a Research Scientist at The Robotics & AI Institute in Cambridge, MA. He obtained his PhD from Yale University's GRAB Lab under the supervision of Prof. Aaron Dollar in March of 2023. In this adventure, he was the recipient of the NSF GRFP and the RSS: Pioneers awards. His research mainly focuses on planning, control, and design for robot hands, typically for the use-case of robot in-hand manipulation. During his PhD, he also spent time at Amazon Robotics AI and TU Darmstadt working with Prof. Jan Peters. At RAI, he focuses on the design and control of low-impedance, high-power manipulators to hopefully someday help bridge the gap in robot capabilities. For more information, visit his website: https://asmorgan24.github.io