news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Xinhua News Agency, Beijing, August 11 (Kong Zixuan) DeepMind, a subsidiary of Google, recently announced that its research and development team has developed a table tennis robot that can reach the level of human amateur table tennis players in competitions.

The research and development team published an article on the preprint website arXiv, saying that this is the first learning robot agent that can reach the level of amateur human players in table tennis. Its main body is a six-axis robotic arm that can move forward, backward, left and right through the bottom slide. In the 29 games against humans, the robot won 13 games, with a winning rate of 45%. The opponents were all human players that the robot had never seen before, and their skill levels ranged from beginners to advanced players.

The bot lost all of its matches against advanced players, but won all of its matches against beginners and 55 percent of its matches against intermediate players, the researchers said.

To achieve human-level speed and performance, the R&D team adopted a hierarchical and modular strategy architecture, which enables the robot to not only master "low-level skills" such as forehand topspin and backhand push, but also formulate strategies through the "high-level controller" equivalent to the brain. During the game, the "high-level controller" can formulate the best skill plan based on the actual game situation, the robot's own capabilities and the opponent's capabilities. After the game, the robot can also analyze the battle data and continuously improve its skills.

The researchers said that the robot still has many shortcomings, such as weak backhand play and poor response to fast balls, balls that are too high or too low, or balls that spin strongly. They will try to further improve the robot's performance by improving control algorithms and optimizing hardware. (End)