Researchers from the University of Tokyo taught a robot how to peel a banana using deep imitation learning. This method is meant to teach machines how to become more human by performing tasks.
How Do You Train A Robot To Peel A Banana?
The methods used to teach a robot how to peel a banana aren’t much different from training a human. It all comes down to practice and having a good teacher.
In imitation learning, experts model desired behaviors in order to help their subjects learn the behaviors more quickly.
After 13 hours of training with a human operator, a machine learning algorithm was able to develop the motor skills needed for a robot to peel a banana. The researchers taught the robot’s artificial intelligence how to peel a banana by showing it what to do over and over and over again.
Peeling A Banana Isn’t Easy… If You Have Robot Hands
Even though handling soft fruit is difficult for machines, the two-handed robot is now able to successfully peel a banana 57% of the time.
The results are far from perfect, but it’s a big start. The long-term goal of this project is to give robots a more delicate touch when engaging with living creatures or delicate objects.
It’s also a good example of how deep imitation learning can be used to improve machine performance.
Why Delicate Robot Tasks Are So Important
This deep imitation learning initiative for robots shows how machines can be taught to become more human.
The ability to have robots with artificial intelligence perform delicate tasks will be a major part of humanity’s future. Everything from changing a baby’s diaper to performing surgery will need a delicate touch.