There was a way to teach a robot a new action in just 10 minutes

Researchers at the Massachusetts Institute of Technology have developed a new algorithm for which

you need a person to show how to complete the task several times: this is enough to reprogram the robot.

A new machine learning method allows the robot to pick up and manipulate previously unknown objects that are in random poses. In 10-15 minutes, the robot will be ready to perform a new task.

According to the authors, in their approach they usedneural network, which is specially designed to restore the shapes of 3D objects. With just a few demonstrations, the system recognizes the object's geometry.

The authors set up a simulation usingreal robotic arm. They showed that their system could efficiently manipulate mugs, bowls, and bottles in random positions. The authors used a total of 10 demonstrations to train the robot.

“Our work is a contribution to the development of systems,that can work in more unstructured environments,” said Anthony Simeonov, PhD student in the Department of Electrical Engineering and Computer Science (EECS) and co-author of the paper.

Read more:

It has been hunted for centuries: what do we know about the planet Vulcan next to the Sun

Astronomers have found a planet near the Earth: it has a very strange orbit

Scientists 'resurrect' an ancient enzyme to feed 9 billion people by 2050