A curling match never seen before opposed the Korean Olympic champions to a robot. The inspiration is still unknown, but the future looks bright for the new machine. Curly the robot during the match against a human team.
This year Pyeong Chang hosted the Winter Olympic, and the main thing many people remember about the organization is the number of new gadgets and technology introduced. South Korea seemed to have a lot of pride in it robotics experts and knowledge and didn’t shy away from showing them off. Even months after the Olympics, the world is still talking about South Korea’s genius. This time the conversation is focused on a certain curling match: the South Korean girls curling team faced a robot in a match. Yes, some engineers in South Korea built a robot that could play curling.
Not a very popular sport in South Korea, curling was reintroduced to Koreans during the Winter Olympics that the country hosted; thanks to its women’s team. The “Garlic Girls” as they are known, surprised and impressed the organizers of the Olympics when they received the silver medal after losing to Sweden. While many foreign spectators thought it was disappointing for the hosting country’s team to miss the gold medal, for Koreans, the silver medal was a success like they had never imagined possible. In fact, the curling team had a track record of losing. In 2014, during the Sochi Games the “Garlic Girls” only managed to win three matches out of the nine for the round-robin play. For the team that came around so fast, one would think that engineers would be wiser not to test a robot against them. But maybe, the goal of the match was not to test the girls but the robot.
Named “Curly,” the robot was designed by a team of about 60 researchers from eight different institutions such as Korea University and the Ulsan National Institute of Science and Technology. The robot played 2 games against the human team and lost the second. Apparently, the “Garlic Girls” were swifter in their moves on the ice than the robot could have been; the team members were adjusting their sliding movements based on their “feel” of the ice. Curly could not make an appreciation of the ice and was very mechanical in its moves.
However, researchers are very proud of Curly for learning the sport so quickly and holding its own against a team of Olympic champions. Curly came out of brains who wanted to understand an AI ability to adapt to an ever-changing environment. To face its opponents, Curly needed special tools like a strategy-oriented software along with a curling simulator that was meant to help the robot adjust its strategy based on the environment. Curly also had video analysis to build its knowledge of curling, data communication and throwing control modules. Its data base is composed of 160,000 throws from all international curling games from 2014 to 2017, a total of 1,321 records. Yet, while humans can appreciate when to calibrate their throws, Curly had not developed that ability yet. And, that may be the fundamental difference between mankind and machines. One of the co-authors of Curly, Prof. Lee Seong-Whan confessed that he couldn’t expect Curly’s throw to be perfect because of variables the robot has no control over.
For the “Garlic Girls,” Curly showed them that they can do better. With its data base, there is no doubt that the robot showed moves that the girls could not anticipate. Some of the girls even think that training with Curly will help them improve faster. The question is how the researchers intend to improve Curly. There is talk of building a robotic team with many “Curly” to compete more often, but how will the scientific community overcome Curly’s challenge with unknown variable? How would machine learning help solve those challenges?
Read about how it works completely after this link for the paper -- Curly: An AI-based Curling Robot Successfully Competing in the Olympic Discipline of Curling
Have a story tip? Message me at: cabe(at)element14(dot)com