The artificial intelligence robot Curly in March 2018 prepares its stone for a test match at the Icheon Training Center of the Korea Paralympic Committee. At the time, Curly lost a two-end game against a high school team, 3-0. (Korea.net DB)

By Kim Young Deok and Lee Jihae 
Video = Korea.net 

An artificial intelligence (AI) robot developed by Korean researchers has won three out of four games against a curling team comprising top female players and members of the national wheelchair team.

A research team led by professor Lee Seong-whan from Korea University's Department of Artificial Intelligence told the global journal Science Robotics on Sept. 24 that the robot, named Curly, achieved this result through "deep reinforcement learning."

Deep reinforcement enables the AI robot to adapt in real time without relearning to its exposed environment, which has a variety of variables and high uncertainty. Through real-time adaptation, the robot stably fulfills its task.

The report said Curly set over three to four days pitching strategies and internalized the strength and direction for pitching and methods to control the swirls of the stone. As a result, it came to possess skills close to that of a curling professional.  

Curling is an official event of the Winter Olympics. The ice field can be irregular depending on the venue's temperature and humidity and ice maintenance. Thus a player needs to consider these elements.

Professor Lee said his team developed the core technology to develop an AI curling robot with the performance level of a skilled athlete. He said the achievement is notable as AI technology based on machine learning can lead to performance at the level of a skilled competitor.

The professor said, "We note that the insights obtained within our framework on how to alleviate challenges such as strong temporal variability, uncertainties, and continuousness are readily transferable for contributing to other real-world applications of comparable complexity in robotics and beyond."

He first developed Curly back in March 2018 at the time of the PyeongChang Winter Olympics. At the time, the robot lost, 3-0, in its two-end game against a high-school curling team. 



By Jung Joo-ri and Yoon Jihye
Video = Park Jang-bhin
Icheon l March 8, 2018

A high-tech curling match recently took place between a human team and a specially designed curling robot.

The Ministry of Science and ICT hosted the match between an up-and-coming men’s curling team and the robot, appropriately named Curly, at the Curling Center of Korea Paralympic Committee (KPC) Icheon Training Center on March 8. The robot was developed by some 60 researchers from eight institutions, including Korea University and the Ulsan National Institute of Science andTechnology (UNIST). 


A stone thrown by Curly the curling robot heads into the house during a match against a human team, at a national curling center in Icheon on March 8. (Ministry of Science and ICT)

In the first ever curling match in the world against a robot, Curly played two games against a high school team from the Chuncheon Mechanical and Technological High School in Gangwon-do Province.

In the first game, Curly won 1:0. The robot figured out the rules in general, just as a human would, and threw a series of great take outs to score.

In the second game, the high school team won 3:0. The human team only used its sweeping skills and scored one point in the first end, and won the second end as well with two points, in which both teams only threw their stones without sweeping. 


The engineers in charge of Curly check the stadium and ice conditions during a preliminary round for 30 minutes before the main game. An engineer examines the practice throws using the software program CurlBrain. (Park Jang-bhin)

However, the robot showed better performance than expected. This was possible as it has collected data and studied it all in-depth. Curly used the 1,321 records of the strategy used in international curling games and built its game data base of 160,000 throws based on date from between 2014 and 2017. Nevertheless, uncertainty arose from the subtle difference in temperature and humidity at the venue, giving the high school team victory.

Professor Lee Seong-whan of the Department of Brain and Cognitive Engineering at Korea University, part of the research team, said, “A perfect throw is practically impossible because of various problems like ice conditions. Once a sweeper robot is developed to reduce errors as much as possible, in autumn this year, accuracy concerning draws, take-outs and other strategy building plans is expected to be much greater.”

Kim Jae-won, who played on the high school team, said, “I thought the game against a robot would be more boring than against humans, but it was rather exciting. Practicing with robots like Curly will help improve our skills.”

Curly was developed with support from the Ministry of Science and ICT in April last year. The research team aims to build a sweeping robot and raise its throwing skills to a professional level by December.


Newsletter Sign Up

By clicking "submit," you agree to receive emails from Bron2Global and accept our web terms of use and privacy and cookie policy*Terms apply.