Robot grips have always been considered tough, but now robotic hands have been used to peel bananas with great skill and gentleness.
Interestingly, this process did not harm soft fruits like bananas, which was not possible before. It is machine learning that has resulted in the peeling of an intricately shaped soft fruit.
This was a challenge because robotic finger grips have been used to smash glassware and crush fruit before. In this experiment, the fingers of the robot were tested and on the other hand, the usefulness of this visual enabled computer vision program has also been seen. In many cases computer vision algorithms play an important role and are also called the brain of the robot arm.
University of Tokyo professor Hequel Kim and colleagues have developed a machine learning system for a two-handed robot that usually grasps small objects with its two fingers.
It also shows that the process of peeling bananas is very complicated. We humans do this very easily and quickly, but training robots proved to be a difficult task.
Experts first removed banana peels from humans and made 811-minute videos. That much data was enough to train a robot. In the next phase, training of robots started.
Many people were called in for the process, who had to pick a banana from the table and peel it, but to show the robot the process, it was divided into nine main parts and stages. That is, lifting a banana from the table, peeling the stalk with the other hand, peeling it and rotating the banana in the meantime.
In this way, the robotic hands learned to move just like humans and copied like them. Within three minutes, he had a hard time peeling bananas. According to experts, the robot was trained for a total of 13 hours and thus the cost of graphic unit processing was reduced and the need for computer processing was also reduced.