In a breakthrough development, a team of researchers from Deep Mind London Project of Google have taught animated players how to play a realistic version of soccer on a computer screen.
A paper published in the journal Science Robotics describes how the animated players were taught to play solo as well as in teams.
In fact, for many years, robot engineers have been engaged to develop robots that are capable of playing soccer. Such efforts have resulted in competition between different groups to see who can create the best robot players.
The creation of RoboCup is a result of this, which has several leagues, both in simulation and real world. Meanwhile, for the new effort, researchers used a new degree of artificial intelligence and learning networks for teaching simulated robots how to play soccer without any rules.
Importantly, the idea behind the model is to make simulated players to learn the game in the same manner humans do – by watching other players. The players started to learn from pretty much level zero. In the first step, the simulated players had to learn how to walk, then to run and kick a ball in the ground.
At every new level of training, players in the AI system were made to view videos of real-world soccer players, which allowed them to learn the basics of playing soccer, and also mimic the manner professional athletes move as they enter in high level sporting spectacles.
Once robots received training to play as a solo player, they were first made to oppose a single player. With improvement in skills, more players were added to the team. Eventually, the researchers could create small teams playing against each other, such as two against two.