People fighting robots in table tennis
The system developed by researchers ABB IRB 1100 DeepMind with an industrial robot arm called special artificial intelligence software brings it together. While an expert human player could defeat the robot, the system demonstrates the machines’ potential to master complex physical tasks that require split-second decision-making and adaptability.
During the tests carried out, the table tennis robot encountered managed to beat all beginner players. With intermediate players In the matches held, robot matches were used. won 55 percent. However high level When faced with table tennis players, the robot lost all matches. In all the games carried out within the scope of the tests, the robot He managed to win 45 percent of 29 matches.
Researchers say that this is the first robot that can perform sports at the same level as humans, showing that significant progress has been made in the learning and control processes of robots. On the other hand, researchers underline that although human levels can be reached in individual tasks, there is still more work to be done for general-purpose robots that can interact with humans skillfully and safely in the real world.
The algorithm behind the robot
DeepMind researchers have developed a technology that allows the robot to apply specific table tennis techniques and adapt its strategy to each opponent’s playing style in real time. a two-level approach developed. In other words, the robot is skilled enough to play with any amateur human player without requiring special training against the human player.
The system’s architecture combines neural networks trained to execute specific table tennis techniques, such as forehand strikes, backhand returns or serve responses, with a more complex artificial intelligence system that analyzes the game situation, adapts to the opponent’s style and chooses which low-level skill policy to activate for each incoming ball.
Leaving these fancy sentences aside, there are areas where the robot has difficulty. For example, the system has difficulty with extremely fast or high balls. He also has trouble reading the backhand and the spin on the incoming ball. Researchers say that these problems are caused by insufficient training data and the delay within the system itself.
On the other hand, researchers are not developing this system to embarrass people in table tennis. The techniques developed for this project can be applied to a wide range of robotic tasks that require rapid response and adaptation to unpredictable human behavior. However, the researchers are confident that by improving the system they can also beat professional players. After all, DeepMind is AlphaZero ve AlphaGo It managed to beat the professionals and the best in chess and Go with its artificial intelligence systems such as.
This news our mobile application Download using
You can read it whenever you want (even offline):
Source link: https://www.donanimhaber.com/google-yapay-zekali-masa-tenisi-robotu-gelistirdi–180520