Education or training robots no easy task. As a rule, scientists and engineers have to resort to complex algorithms to the motion of their mechanical wards answered to certain requirements. A team of researchers from the U.S. rice University has developed a completely new method for learning robot movements. To ensure that the robot is able to learn how to move, a person just need to push it or point the owner in the right direction.
Yes, the new method is also based on machine learning algorithms, but in this case it comes in not only math, but also the operator, who is like the parent who teaches their child to take their first steps, carefully guide the robot and tells him how to perform the task correctly and without errors. Depending on how the person guides the robot changes the trajectory of its movement. Real-time person contributes new knowledge in native code, and the robot learns something new.
As a rule, if the robot is programmed to interact with the man, he will notice the physical contact, but after that I will try to return to the initially chosen trajectory. In this case, the robot is able to use this contact for their own learning. It will take into account the change in angle, trajectory, direction, and next time will act according to the received new knowledge. This is most clearly shown in the researchers published a video where they demonstrate their system in action and explain how it works.
Researchers in the US have developed a new way to teach robots
Sergey Grey