Robot that can play Table Tennis just only after 90 minute of training

A table tennis- playing robot can keep up a rally against humans, but like numerous amateur players, it struggles when trying fancier shots. 

Robot that can play Table Tennis just only after 90 minute of training

  Yapeng Gao, Jonas Tebbe and Andreas Zell at the University of Tübingen in Germany began by designing a computer simulation in which a virtual robot arm equipped with a table tennis chatter tried to return clunk pong balls across a virtual table tennis table. 
The experimenters ran this simulation so that a machine learning algorithm could learn how the haste and exposure of the chatter affects the path the ball takes. 

 Once this algorithm, which learns by trial and error, could reliably return the ball, the experimenters set it up to control the movement of a real robot arm deposited next to a real table ( pictured). 
The system used two cameras to track the position of the real ball every 7 milliseconds, and the algorithm reused the signals and decided where to move the robotic arm to hit and return the ball. 
 The signals that the algorithm transferred allowed the robot arm to directly play shots to within an normal of24.9 centimetres of the intended position. This delicacy position was slightly worse than when the algorithm was working with a simulation – a common circumstance, says Tebbe, as computer simulations can’t directly represent everything in real life. 
 The entire process – including training in the virtual simulation and in the real world – took just1.5 hours, demonstrating how fleetly algorithms can learn to perform in a new situation. 

 Still, although the robot performed well against mortal players, it was tripped up by fast shots – and, unexpectedly, by slowones.However, the robot needs to induce further speed,” says Tebbe, “ If a ball is slow. Floundering to do that, the ball frequently drooped off the chatter. 
 “ By training the system for a fairly short period of time the robot is suitable to manage well with differences in serve, and able of returning using a arbitrary policy,” says Jonathan Aitken at the University of Sheffield in the UK, who was n’t involved in the study. 

 Aitken was surprised the algorithm missed returning slow shots. He also finds it intriguing that it occasionally plodded with making shots because of the mechanical limitations of the robot system, rather than because of failings with the algorithm. 
 The robot arm has other limitations. For case, it struggles to play backspin shots, says Zell, because the robot arm is unfit to hold the chatter at the needed angle demanded to perform similar shots. But despite these issues, he believes the robot is a good player. 

 “ It’s not worse than a regular mortal player,” he says. “ It’s formerly on par with me.”