In an enormous space at an undisclosed U.S. location, a four-wheeled autonomous vehicle the size of a microwave oven rolls up to a robotic arm and stops suddenly. The arm swivels, inserts a probe into the vehicle’s plastic bin, and, with a whooshing sound, retrieves a box of Hamburger Helper.
Amazon and other retailers still use humans to pick up items, because replicating the human hand, which can just as easily grab a lightbulb as it can a pack of AA batteries, is one of the toughest problems in robotics. The key technical breakthrough was suction cups, says Matt Mason, Berkshire Grey’s chief scientist and a Carnegie Mellon University computer science professor who specializes in grip.
Mason and Berkshire Grey’s 100 employees, spread across a Lexington, Mass., headquarters and a Pittsburgh research and development lab, designed a system that uses cameras connected to artificial intelligence software to scan and identify each item, verifying that it belongs in the shipping queue and determining where to grab it.