For those fractions of a second before youIf you take an object in your hand, your brain will calculate all the possible combinations of movements necessary to grab and move the object safely and securely. Such a subconscious brain function develops in a person from childhood in the learning process. The result of a similar approach to training was demonstrated by the German company Festo, which manufactures industrial equipment and automation systems. The company introduced the new robotic arm BionicSoftHand, which not only has amazing flexibility, but also uses artificial intelligence to calculate the correct algorithm for manipulating objects until it takes them.
BionicSoftHand is another example of “soft”approach in robotics. The robots used in manufacturing to move heavy objects are made of steel and work on the basis of powerful pneumatic systems that make them strong and fast, but completely indelicate. The contact of a machine with a person at the time of its operation most often ends in disastrous consequences for the latter. In turn, the so-called soft robots use mostly more flexible materials in their design, for example, “smart” fabrics or inflatable parts. Therefore, interaction with them is significantly less traumatic.
For maximum security, Festo does not usein a new robotic arm a solid skeletal structure. Instead, it is equipped with inflatable bellows, around which is located a three-dimensional fabric made of elastic fibers created using 3D printing technology, which bends along with hand movements when air is blown into the bellows. In other words, if you simplify, it looks like all this is more on the work of our muscles and tendons.
Soft fingers of robotic armsBionicSoftHand are equipped with inertia (movement) and force sensors, also created on the basis of flexible materials that can bend at the moment of movement. Sensors provide feedback to the hand control systems, reporting on the movements of the arm and fingers, as well as contact with the object. In addition, they convey information about whether the object is properly located in the hand for its manipulation. In other words, a complex of sensors gives the hand a contact perception, imitating our sense of touch.
Usually production roboticmanipulators are programmed in such a way that they can perform an infinite number of specific movements. Of course, in their systems a certain proportion of the variant execution of the task is allowed, but basically all their work is reduced to the same set of manipulations with objects, for example, moving them from one place to another.
BionicSoftHand is more “similar tohuman ”: before the beginning of the manipulation of an object, its artificial intelligence system independently enumerates all possible combinations of how to correctly take an object, how to turn it and put it back. To do this, it uses a complex virtual simulation in which, by trial and error, selects the correct option for further action. As a child, you probably 100 times dropped a cup with some drink, until you learned how to hold it properly. So, BionicSoftHand performs the same thing, only in a virtual environment, going through billions of erroneous options and eventually finding the most correct and optimal one. In a split second.
In the demo video, the robot was given the taskmove a 12-sided object placed in a hand so that a certain face looks up. How exactly to do it - the machine is not explained. First, using a depth-perception 3D camera, the machine creates a digital copy of the object in its hand, and then places it in a simulation with many virtual hands that manipulate the object until the right combination is selected. Then the physical hand of BionicSoftHand comes into play. With this method of learning the machine is able to learn much faster than a child. At the same time, when performing new tasks, it can use the manipulation algorithms already learned before.
Subscribe to our Yandex.Dzen to keep abreast of the latest interesting events from the world of science and technology.