New MIT algorithm helps robots rapidly pick up and sort items

Though we're not yet at the point where robots can take fully take over warehouse jobs, technology is rapidly developing toward that end goal. One big issue surrounding robots in these jobs is their ability to sort items. Unlike boxes, which are neatly stacked and easy to grasp, small unpackaged items introduce randomness into the mix, requiring a robot to do a bunch of thinking about how it will grasp the items. This takes a fair bit of time...or at least it used to.

Researchers with MIT have detailed the development of a new algorithm that speeds up a robot's ability to grasp and adjust its grip on an object, something that would typically require several minutes of calculations. The new algorithm enables robots to figure out these maneuvers in less than a single second.

The algorithm enables a robot to figure out how to adjust its grip on an object by using a surface to push against. One example given is a robot grasping a plastic 'T' and changing how it grasps the large letter by using a flat surface to reposition its gripper. This capability would be particularly useful for industrial robots that may one day work by sorting items in warehouses.

The work is made possible by using what the team calls 'motion cones,' which are described as maps of friction that are shaped like cones. Each motion cone is calculated to factor in the different potential configurations involving the object, the robotic gripper, and the surface the robot uses to manipulate the object.

Rachel Holladay, one of the study's authors, explained, "Seemingly simple variations, such as how hard robot grasps the object, can significantly change how the object moves in the grasp when pushed. Based on how hard you're grasping, there will be a different motion. And that's part of the physical reasoning that the algorithm handles."