Improving robots’ grasp requires a new way to measure it in humans

May 30, 2019

HUMAN BEINGS can pick up and manipulate objects and tools with hardly a thought. This seemingly simple task, however, requires the precise, co-ordinated movement of individual fingers and thumbs, each applying the correct amount of pressure, at exactly the right places, to the object handled.

That people can do this successfully is thanks to special nerve endings, called mechanoreceptors, found in their skin. These provide instant tactile feedback to the brain of the shape, feel and weight of whatever is being grasped. With time and experience, people learn to vary their grip instinctively when lifting a golf ball, for example, as opposed to an egg.

Replicating that dexterity in robots is hard. A machine usually picks things up by, first, identifying the object via a camera and appropriate software, and then using a preprogrammed grasping strategy appropriate to what it thinks it has seen. This approach has improved greatly in recent years, thanks to advances in machine learning and vision. Further improvement will, however, be best served by a more precise understanding of the mechanics of how people themselves manipulate objects. A new “smart” glove, from computer scientists at the Massachusetts Institute of Technology, promises to do just that.

Writing in this week’s Nature, Subramanian...


Other news

Cookies help us deliver our services. By using our services, you agree to our use of cookies.