BionicSoftHand determines how to perform tasks by trial and error, but without tedious and tedious manual processes. Once the engineers have given a goal to the robot, 3D cameras and depth sensors create a virtual copy of an object. The artificial intelligence of the hand then runs a myriad of simulations to help him quickly understand what to do. In a demonstration video, the robot determines how to roll a dodecahedron between his fingers without dropping it. Without AI, the robot could take months to learn to do.
Artificial intelligence is only part of the project. The robot has a soft hand, which means that it can interact safely with humans.
The fingers are made of a 3D textile knit (using elastic and high strength fibers), with tactile sensors and a flexible printed circuit board. There is a bellows on the inside to move the seals when they fill up and empty themselves with air. The pneumatic swivels at the base of the thumb and forefinger allow them to move from one side to the other, while there are two dozen proportional piezo valves in which is essentially the wrist for everything to work. According to Festo, the construction of the hand allows him to have 12 degrees of freedom, which allows him to manipulate objects with dexterity.
We are probably far from seeing this technology in factories and classrooms. The automatic learning means that it will improve over time and that as the AI and the hardware progress, robots such as BionicSoftHand have to determine how to manage even more agile tasks with autonomy.