A newly invented robot hand can teach itself to manipulate objects just like a real human hand through trial and error.
Present-day robots can impressively mimic human motions and perform pre-programmed actions. They are, however, junk when it comes to performing tasks such as catching objects or repositioning a pencil in their fingers.
A Self-Learning Robot
A new amazing robot, however, can perform, hold, and manipulate an object like a human hand. Festo's robot hand called the BionicSoftHand can autonomously teach itself to handle objects such as rotating a 12-sided cube.
The BionicSoftHand could perform tasks through trial and error with the use of artificial intelligence. If it is tasked to assemble a toy, it can keep trying until it understands what will work and what will not. It uses learning algorithms to master tasks without giving new codes.
The BionicSoftHand is Festo's solution to avoid injuries caused by robots in a factory, for example. The robots in factories are made of steel and high-pressure components, making them strong and fast in moving objects and perform tasks they are programmed to do. However, an accident with these robots could cause serious injuries to humans.
Soft robots, on the other hand, are made of flexible materials such as smart fabrics and inflatable bladders. Thus, if a human accidentally made contact with a "soft" robot, unlike the steel robots, they would not cause serious injury.
Festo's new BionicSoftHand is made of inflatable bellows wrapped in a fabric skin made of knitted elastic fibers. The bellows move and flex along with the hand's motions whenever the air is pumped in to make a movement. It is like a human hand's muscle and tendon system, but the soft parts mean that it is safer for a human to interact with the robot.
The BionicSoftHand determines how to perform tasks beginning with a trial and error, minus the long manual process of tests. With the help of AI, the robot can learn quickly how to roll a cube between its fingers without dropping it. Without AI, the robot could take months to learn the process.
The new invention is expected to improve over time, and experts are hopeful to see this new technology in high-tech factories and classrooms as well.