Robots on the loose in the world are clumsy and sometimes make bad decisions – the trip on sidewalks and fall downstairs and into water fountains.
A team of researchers in machine learning, cognitive engineering, and computer science from Korea University and the Berlin Institute of Technology has developed a robot capable of curling. Named Curly, a new AI-based curling-playing robot is able to compete at a professional level. It has a better handle on those complexities, thanks to an artificially intelligent brain that can quickly assess and map the icy environment, the state of play, and optimal strategies for winning.
Curling is a team sport played on ice, where two teams take it in turns to slide stones made of granite towards a target – known as a House – which is segmented into four concentric circles. It is an Olympic and Paralympic winter sport with medal disciplines for Women’s, Men’s, Mixed Doubles, and mixed Wheelchair teams.
The robot Curly is actually a team of two robots, one of which analyzes the position of the stones on the target, calculates the best power and the trajectory to use to win, and throws the stone. And it hits the mark! In another world, the white, turtle-shaped robot is able to observe real-world conditions and act accordingly in a precise and strategic manner.
According to the researchers, the components can communicate to identify launch errors, interpret changes in the ice, and make appropriate adjustments for future launches.
To put their development to the test, they pitted the robot against a “high-level” South Korean team, and the result was a landslide three-to-one victory, thanks in part to the robot’s good adaptability. In fact, the researchers themselves conclude that these results indicate that the gap between physics-based simulators and the real world can be narrowed.