A robot designed to play air hockey can learn a human opponent's moves and use them to win.
Computers have been playing games against humans almost since computers existed — but not many are programmed to learn as they play. Just look at any enemy AI in a video game, or the robot goalieearlier this year.
An air hockey robot designed by researchers at Japan's Namiki Lab at Chiba University, though, can do just that, adapting to the style of every individual player to cross its table.
The robot itself is pretty minimal, consisting of an air hockey table, a four-axis robotic arm, two high-speed cameras and an external PC. The cameras track the speed and position of the opponent's paddle in relation to the hockey puck at a super-high rate of 500 frames per second.
This data is then interpreted by the PC in real time, compared against reference patterns, so that the robot can determine whether the opponent is playing aggressively or defensively, and react accordingly. This keeps the game from getting repetitive and boring. "To avoid this, the robot should be offensive when the opponent is offensive, and should be defensive when the opponent is defensive," the researchers wrote in a paper.
This means that, by reading its opponents' strategies, the robot is able to force the human player to adapt and change the way they play in order to try and beat it. Somehow, we don't think that's going to be easy.
The paper, "Hierarchical Processing Architecture for an Air-Hockey Robot System", was presented last month at the IEEE International Conference on Robotics and Automation in Germany.