Australia wins (robot) Soccer World Cup
* Feature image credited to RoboCup Federation
Not just fancy footwork – two sport savvy researchers from the University of Sydney and CSIRO’s Data61 with expertise spanning coding, data and complex systems have scored first place at one of the world’s most prestigious robotics competitions.
Kicking goals: this is the second time Professor Prokopenko’s team has placed first at RoboCup.
Fifteen international teams from Australia, Brazil, China, Germany, Iran, Japan and Portugal were tasked with building and programming a team of fully autonomous virtual robots, in pursuit of the coveted RoboCup Millennium Challenge: designing humanoid robots capable of winning a soccer game against the 2050 FIFA World Cup holder, while complying with official FIFA rules.
Held in Sydney for the first time, the RoboCup 2D Simulation League included teams of 11 fully autonomous virtual robots – known as “agents” – playing football in a two-dimensional virtual soccer stadium without the assistance of a remote control.
Their software – Fractals – won them the grand final against the 2018 world champion team, HELIOS2019.
“We are so delighted that our team has won the Soccer Simulation 2D League at the 2019 RoboCup. The grand final was a very tense game which went into extratime after a goalless full-time, with Fractals scoring a winning goal with less than a minute to go,” said Faculty of Engineering academic and Centre for Complex Systems Director Professor Mikhail Prokopenko.
“The key to our success was the long-term collaboration between the University of Sydney and CSIRO’s Data61, combining the strengths of complex systems and data mining.”
To determine a winner, 156 games were played over multiple rounds over four days, with Fractals playing 22 games, winning 16 times and outscoring their opponents 59 -10.
Each of the Fractals players was built with virtual, acoustic, physical and visual sensors to navigate and detect the cyber playing field, performing actions like running, turning and kicking the ball.
“The 2D Soccer Simulation League is a complex game as it requires intelligent agents to autonomously make optimal and strategic decisions in a high-stress, ‘noisy’ environment, all while facing the disruptive actions of opposing teams,” explained Professor Prokopenko.
“The main challenge is to get the agents to work together as a team, which we achieved by coordinating the individual agent actions across a multi-agent team, in order to produce coherent tactics.”
Senior research engineer at CSIRO’s Data61, Peter Wang said being part of the 2D Simulation League at the 2019 RoboCup was an incredible experience.
“We were tasked with an interesting and challenging problem, that was solved using a combination of methods and by balancing risks and rewards. Once we found the correct structure and balance, everything came together,” said Peter Wang.
Since its inception in 1996, the 2D Simulation League has made several important technological advances that achieves autonomous decision making under constraints, such as flexible tactical planning, collective behaviour and teamwork, as well as opponent modelling and adaptation, setting new benchmarks for artificial intelligence.
The winning team, Fractals, was led by University of Sydney Director of the Centre for Complex Systems, Professor Mikhail Prokopenko and included Peter Wang, a senior research engineer with CSIRO’s Data61.
To view the original article by Sydney University, click here.
Want to watch the full game? Here’s all the action: