When you handle objects, your eyes are only responsible for part of the dexterity that keeps you from dropping or damaging them. The rest depends on your sense of touch, and your brain is able to merge the information from both senses in order to learn how best to grasp a specific item. Robots struggle with that process because they often rely solely on visuals, or even just explicitly programmed routines. FAMULA self-learning robot hands mimic the way our own hands work in order to make robots like Floka more versatile.
Floka is composed of a pair of industrial robotic arms, a head with moving eyes, and the FAMULA hands developed by a team at Germany’s Bielefeld University Cluster of Excellence Cognitive Interaction Technology (CITEC). The hands each have five fingers, and have movement capability that is similar to a human’s. And, like a human, they can feel what they’re touching and then use that information to figure out the best way to hold any given item.
Using tactile sensors build into the fingers, the FAMULA system can build a three-dimensional point cloud model of whatever it’s touching so that it knows the exact size and shape. It can also use those sensors to determine how much pressure it’s putting on an object. It can detect if something is slipping from its grasp, and then use more pressure as necessary. The system has a lot of potential for general purpose robots that need to interact with unknown objects, such as household robots that have to be able to handle whatever you happen to have in your home.