Nancy Pollard: Bringing Dexterity to Robot Hands in the Real World
Professor Nancy Pollard of Carnegie Mellon University discusses practical robot dexterity at an MIT Robotics seminar, drawing on research with her team and collaborators.
Nancy Pollard, professor of robotics and computer science at Carnegie Mellon and co-founder and president of FutureHand Robotics, discusses building dexterous robot hands for real-world tasks at a CMU Robotics Institute seminar introduced by Henny Admoni.
Pollard defines dexterity as adapting movement to solve a task, even when each attempt looks different. She points to factory work such as securing and plugging in a moving connector, and field harvesting of peppers, as tasks that remain hard to automate because they require fine, often two-handed interactions.
Human hands bump into objects during everyday tasks and adjust as they go; Pollard argues that compliant systems can tolerate those contacts. She contrasts a rigid prosthetic that struggles to pick objects off a table with a soft robot hand and a two-arm pepper harvester designed to grasp fruit while a separate arm cuts it.
Pollard describes transferring human skills by considering where contact happens and what each contact does. Her group represents hand-object contacts as surface patches that can be moved or reshaped onto other objects and robot hands, and develops methods to find grasps that remain useful despite uncertainty in object size or position.
Rather than simply copy human anatomy, Pollard argues for designing hands around the tasks they need to perform. One project used foundational poses from clay sculpting to screen hand designs, while an open-source, fully soft educational hand has 12 degrees of freedom and a complete kit costs about $2,000. FutureHand Robotics aims to capture a task, build a hand, and transfer the skill in as little as five days.
Pollard summarizes the need for soft hands, contact-aware analysis, and task-specific design. In questions, she says basic contact sensing across the hand can be more useful than detailed sensing limited to fingertips, while force readings matter for tasks such as bathing. She also argues that robot hands may improve beyond human designs and warns that learned motions need to be grounded in task goals and physical principles.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Professor Nancy Pollard of Carnegie Mellon University discusses practical robot dexterity at an MIT Robotics seminar, drawing on research with her team and collaborators.