A new research project out of Japan has air hockey champions shaking in their sneakers. Researchers have developed a robotic arm that dominates the air hockey table with a killer combination of precise moves and long-term strategy.
The robot, which is mounted on a regular air-hockey table, is manufactured by Barrett Technology, a spin-off of MIT's Artificial Intelligence Lab. Barrett's robotic manipulators are the most advanced in the world and are typically used in factories, hospitals and research facilities.
But researchers at Japan's Chiba University thought it was time these robots had a little fun. Controlled from an external PC and equipped with two high speed cameras, the robotic arm "sees" at 500 frames per second. Its opponents, mere humans, see at about 20 frames per second.
And super vision and lighting quick reflexes aren't the only thing this robot has going for it. It's programmed with a three-layer control system, enabling it to "think" like humans, only better.
The first layer of control determines the robot's physical movements at the hardware level. And layers two and three control the robot's short and long term strategy. So this machine not only knows when to hit, block or stay still, it also has an overarching game plan.
By observing the speed and position of its opponent's paddle in relation to the puck, the robot can determine if someone is playing aggressively or defensively. It keeps track of these behaviors in real time, comparing them with reference patterns in its software to determine its opponent's preferred method of play.
In a series of experiments, the robot was able to detect its opponents' strategies time and again, forcing them to change their style of play or admit defeat.
Though losing to a machine might sound like a real bummer, participants in the study told researchers that the robot's relentless strategizing actually made the game more exciting.
[See also: Beer-Pouring Robot Knows When You Need a Refill ]
Email asklizzyp@gmail.com or follow her @techEpalermo. Follow us @TechNewsDaily, on Facebook or on Google+.