Currently, most commercial robots can only “pick and place” objects. Thus, most commercially available robots are limited to “pick and place” tasks, which is far from the capabilities of a human hand. However, this will change, thanks to MIT students in the Computer Science and Artificial Intelligence Lab, who discovered how household robots could employ lateral thinking for greater dexterity. In one paper, Jennifer Barry, PhD student, describes an algorithm that allows the robot to push an object to the edge, so that the object can be grasped. In another paper, Annie Holladay, MIT senior, describes how a two-armed robots can steady an object in place.
Although both Barry and Holladay’s algorithms allowed parameters of the objects to be plugged in, the ultimate goal of this research was independent learning. The robot itself was to infer how to manipulate new types of objects. Although robot movements are still far off from human dexterity, this research is a step towards solving the more general motion-planning problem.
Jasmine Yan, AME
Image credit: news.mit.edu