Robots learn to grasp objects by practicing and teaching each other

Even for robots, practice makes perfect. At least that's how it works for the technique developed by Brown University assistant professor Stefanie Tellex, which teaches robots how to pick up objects so they can relay the info to other robots. Tellex has been working on the technique with the help of an industrial machine called Baxter, which has two hands and a touchscreen face. These slightly human-shaped automatons use cameras and infrared sensors to examine an object -- they then pick it up from various angles using different grasps in order to find the most secure way to hold it. Once they do determine the perfect grip, the information is encoded in a format that can be shared online and uploaded onto other robots' brains.

According to MIT's Technology Review, the robots picked up objects 75 percent "more reliably" using Tellex's method than using its preloaded software. Someday, companies could use it to teach hundreds, if not thousands or millions of industrial robots how to handle their products in factories or warehouses. Amazon, for instance, is already exploring the idea of using robotic warehouse personnel, and we're sure other companies are mulling it over, as well.