Pages Menu
Categories Menu

Posted by on Oct 28, 2012 in Technology |

Artificial Intelligence: Playing table tennis robots

Robotics is present in more and more areas, something we can see materialized with autonomous vehicles, in the military with robots and exoskeletons or AlphaDog and even some of them are able to dance steps marked the Gagnam Style . Within the wide range of robots that we know through the media and other groups, we can find different projects robots capable of solving complex problems and act, on the fly, based on the knowledge acquired through a process learning. Learning processes based robots are projects like Macgyver robot or robotic arms at MIT we met yesterday and we will add a curious player robot developed by the Max Planck Institute for Systems Development Smart.

This project takes time to develop within the Max Planck Institute for Intelligent Systems Development, located in Tübingen (Germany), where the investigators have taken the basis of a commercial robot arm and has focused on developing a learning algorithm for the arm to be able to play table tennis, that is, locating the ball (using a vision system arficial) and return it to the opposite side of the table with one blow.

And how can a robot learn to play table tennis? The answer may be simple: programmatically, but really is more complicated since one of the members of the research team grabbed the robot arm and led him into a conventional starting with the idea of getting a set of guidelines and basic movements. This tutoring, the team extracted some 25 basic movement patterns and, by combining them, the robot is able to respond (with more or less success) and hit the ball with your paddle.

When the robot receives the ball and must respond, perform a combination of the 25 patterns by a weighting system, ie each of these 25 basic movements are assigned a weight and the end result is the blow to be made. And whence comes the weights? This is where comes into play again, the learning and training process , a process in which the system under test (he does play) and was fed with the successes and failures in his punches, recalibrating the weights and, through practice, obtaining better results rate.

The old adage that practice makes perfect applies not only to the human case. While this project has some time, and the last of the videos you’ve uploaded one of the researchers on the team is a few months ago, we must recognize that the result is magnificent.

Tags: , , , ,