Emerging Technologies

Robot football: How the 'beautiful game' has advanced AI and automation

Published · Updated
Programming a robot at a RoboCup tournament in Germany.

"Important real world applications can start with creating robot footballers," said the World Economic Forum's head of artificial intelligence. Image: Photo by C M on Unsplash

Spencer Feingold
Digital Editor, World Economic Forum
  • The complexities of football make it a challenging feat for artificial intelligence and autonomous technologies.
  • The International RoboCup Federation organises robot football matches and publishes scientific papers on machine learning.
  • Experts note that advances in robotic football have significant implications for other forms of automation.

In 1997, a computer programme named Deep Blue beat the world’s reigning chess grandmaster, Garry Kasparov, in a highly publicised, multi-day chess match. Enthusiasts have debated the fairness of the game, but the computer’s victory was a major milestone for machine learning.

Today, researchers have been using another game to advance artificial intelligence (AI) and automation: football.

The agility, strategy and teamwork required in football makes the beautiful game a challenging and fruitful feat for AI and autonomous technologies. The International RoboCup Federation, which organises robotic football tournaments around the world, has long been leading such efforts.

“RoboCup is a very special, uniquely inspiring event,” said Peter Stone, the former RoboCup president and the executive director of Sony AI America, an AI development and research company. “It merges long-term, ambitious thinking about how AI and robotics will change the world with practical, immediate considerations regarding how to design, build and program prototype, custom-built robots to sense, decide and act.”


How is the World Economic Forum ensuring the responsible use of technology?

Under RoboCup rules, each team must consist of fully autonomous robot footballers; once the match begins, the human coaches are not allowed to interact with their robot players. Stone, who is also a professor of computer science at the University of Texas at Austin in the United States and the director of the university's Learning Agents Research Group, noted that achievements in robotic football have advanced a range of AI and machine learning fields. This includes human-robot interaction, computer vision and multiagent systems, among others.

Such advancements are shared in the RoboCup Symposium, a publishing arm of the organisation that has so far produced hundreds of scientific papers detailing machine learning breakthroughs that emerge from the competitions.

RoboCup is an enduring grand challenge for AI and robotics and a training ground for the next generation

Peter Stone, professor of computer science at UT Austin, executive director of Sony AI America and former RoboCup president

Football was also used recently to advance AI by researchers who successfully created digital footballers that learned the game as they faced off against each other in a computer simulation. The project—published in August in Science Robotics, a peer-reviewed scientific journal—used human motion data and other machine learning systems to enable the players to perform and improve football skills.

“We tackled the integration of motor control and long-horizon decision-making in the context of simulated humanoid football, which requires agile motor control and multiagent coordination,” the research team wrote. “Our work constitutes a complete demonstration of learned integrated decision-making at multiple scales in a multiagent setting.”


Experts note that advances in robotic football have significant implications for other forms of AI and automation such as self-driving cars and care robots.

“In many countries, we have populations with insufficient young people to care for their elders. Robots will be the way in which that care can be provided to keep people active and in their own homes longer. Autonomous vehicles will provide mobility to the older generations," said Kay Firth-Butterfield, the World Economic Forum's head of AI and machine learning. "These important real world applications can start with creating robot footballers where tests of agility, machine vision and machine learning are vital to playing the game.”

RoboCup, for instance, has expanded its efforts to include new competitive leagues that feature autonomous service robots. This includes the RoboCup@Home league, which focuses on the development of robotic personal and domestic service capabilities. The organisation also hosts the RoboCup Rescue league where teams compete to develop advanced robotic tools for emergency responders. The most recent winner in this tournament was a research team from Kyoto University in Japan whose search and rescue robot demonstrated novel dexterity.

The complexities inherent to football continue to make the RoboCup football tournament a leading source of inspiration and challenge for robotics and AI researchers.

“Will soccer go the way of chess, checker, poker, Jeopardy, Starcraft, Gran Turismo, and other heretofore human endeavours that have been conquered by AI? Or will the requirements for real-world perception and human-like speed and agility keep it out of reach?” Stone asked. “This question remains a source of debate and uncertainty within the community.”

Nonetheless, RoboCup organisers maintain an ambitious goal: by the middle of the 21st century, a team of fully autonomous robot footballers will face off against the most recent winner of the World Cup and win.

Have you read?

About Us



Partners & Members

  • Sign in
  • Join Us

Language Editions

Privacy Policy & Terms of Service

© 2024 World Economic Forum