Meet the Kinema Pick. It's a self-taught robot capable of breaking down large pallets of cardboard boxes, and it's probably going to steal a lot of jobs someday. It's the brainchild of Kinema Systems, a robotics startup from Menlo Park, California. According to the company, the nifty machine is the very first "self-training robot picking solution". Designed for use in warehouses and stores, it uses a combination of 2D and 3D vision, and computer learning to dismantle a shipping pallet.
Watch it in action:
It uses a suction system to grab hold of the boxes, which is simple enough for a human but considerably more complicated for a robot. Because the boxes are often different sizes and shapes, the robot must correctly judge where to grab on to, or else it might drop it.
Sachin Chitta, Kinema Systems' CEO, wouldn't explain to MIT Technology Review exactly how the bot figures out how to grab the first box -- he referred to it as the "secret sauce" -- but said that it only takes a few seconds. After it learns the shape of that particular box, it takes less than a second to pick up the same box in the future. (Humans, on the other hand, take about six seconds.)
Robotic random picking -- the act of selecting different parts from a given area during the production process -- is a known problem in industrial robotics; Kinema refers to it as the "holy grail". And now a robot, of course, is here to save the day. On the bright side, it will likely be gentler with your stuff than humans are.