Robert J. Wood, a robotics researcher at Harvard University, discusses designing soft manipulators for delicate and dexterous tasks, including deep-sea sampling, in a GRASP Lab seminar.
Wood contrasts traditional hands, built for dexterity, with soft robotics, which can handle delicate objects. He proposes using a robot’s compliance to reduce the burden of modeling contacts, object shapes, and positioning errors. His target measure of dexterity is in-hand manipulation, where an object moves without repositioning the manipulator.
Wood describes developing pneumatic soft fingers with David Gruber to collect fragile marine organisms intact. A finger with coupled bending and twisting wraps around a target, simplifying control in an unstable, remotely operated vehicle. The team used the devices on coral, sea cucumbers, and a crab, and later extended the approach to a soft arm controlled with a haptic glove.
Wood presents F.O.A.M. actuators, which use vacuum and a folded internal structure to produce programmable contractions. A 2-kilogram arm lifted a payload of about 20 kilograms, a payload-to-arm-weight ratio of roughly 10. To recover dexterity, the group added parallel and serial finger motions, using simulation to compare designs and compliance to grasp thin materials despite placement errors.
Wood treats the palm as an active part of the hand: it can support objects against gravity, change its shape, and switch friction to guide motion. These features enable translation, rotation, and reorientation within a grasp. The team also found that finger arrangements suit different object shapes, and that adjustable friction on the digits adds another way to control objects.
Wood describes SoMo, an open-source soft-robot simulator built around PyBullet, and a lab-scale injection-molding process that uses 3D-printed molds. He also highlights fully soft robots, small climbing robots for engine inspection, habitat maintenance, and whale-data collection. In Q&A, he notes that prosthetics and surgery could benefit from soft manipulation, while mobile devices still face power and actuation challenges.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Robert J. Wood, a robotics researcher at Harvard, describes work on the mechanical side of artificial intelligence at a CMU Robotics Institute seminar.
