Robots have come a long way, but their sense of touch has lagged behind — cue the Internet’s endless clips of machines smashing eggs and breaking bottles. To improve their dexterity, a team of Columbia engineers has created a robot learning system with pressure sensors that “feel” objects with unprecedented precision and adjust the grip in real time. In tests, the robot deftly grasped fragile items like eggs and grapes and performed complex hand maneuvers, such as reorienting a hex key and serving food with a spatula. “Our tactile sensors bring robots’ sensing capabilities a step closer to human levels,” says Yunzhu Li, an assistant professor of computer science and the project’s lead investigator.
