Sport
Humanoid robots face tougher football challenge at World Robot Games
This year’s competition features football and table tennis. In medium-sized robot football, teams have expanded from three players to five, while the pitch is almost twice the size of those used last year.
Humanoid robots are set to face a tougher football challenge at the second World Humanoid Robot Games in Beijing, with larger playing fields, more teams and improved technology testing the limits of the machines.
The qualifying round began on August 11, with the matches also serving as a technical test ahead of the main event, which opens on August 22.
The games, jointly hosted by the Beijing municipal government and China Media Group, are the world’s first multi-sport competition dedicated to humanoid robots.
The inaugural event in 2025 attracted 280 teams from 16 countries, with more than 500 robots taking part.
This year’s competition features football and table tennis. In medium-sized robot football, teams have expanded from three players to five, while the pitch is almost twice the size of those used last year.
The larger field is expected to test the robots’ ability to perceive their surroundings, track their position and navigate in real time.
Under the rules, the robots must operate autonomously without direct human control. The larger playing area reduces their effective scanning range and increases the risk of positioning errors as matches become more competitive.
To compensate, some teams have programmed their robots to turn their heads more frequently to scan the field and identify reference points.
“When the robot scores a goal or performs a goal kick, an indirect or direct free kick, it will turn its head to re-search for feature points,” said Cheng Junming, a participant from Beijing Information Science and Technology University.
He said the technique provides a second form of positioning to help correct the robot’s location.
The robots have also been designed to move in increasingly human-like ways, reflecting advances in both performance and entertainment.
Teams are using reinforcement learning to train robots to control and dribble the ball, while depth cameras help them detect the ball and establish its position relative to the robot.
These developments are allowing the machines to move with greater coordination and play football in a way that more closely resembles human players, Cheng said.
