Germany may have won this year’s World Cup but Australia’s robot researchers have emerged victorious from a subtly different competition.
The last time Australia brought home the trophy was in 2008, when the University of Newcastle’s NUbots team defeated the US 1-0 in a nail-biting final.
Although this latest result is great news for Australia, it’s not the only success that we have had in Brazil this year. Let me start by explaining that RoboCup is not all about the robots playing soccer.
RoboCup’s 2050 Millennium Challenge
RoboCup is an international robotics competition formed in 1997 with the official aim that:
[…] by mid-21st century, a team of fully autonomous humanoid soccer players shall win the soccer game, complying with the official rule of the FIFA, against the winner of the most recent World Cup.
RoboCup has grown massively in popularity over the years and last year’s competition in Eindhoven, in the Netherlands, attracting participants from 45 countries and more than 40,000 visitors, including Queen Máxima of the Netherlands.
When most people think of RoboCup, they usually think of the Standard Platform League competition. In this league, all teams are required to use the Aldebaran NAO humanoid robots – the competition is all about writing the best software possible.
While teams such as the UNSW one are busy writing hundreds of thousands of lines of soccer-playing code, others are hard at work designing the next generation of robot hardware.
This is the domain of the Humanoid League - the second-largest RoboCup competition – of which the University of Newcastle’s NUbots have been annual competitors since 2012.
It’s not all about robots
As anyone who has worked with robots will testify to, there are a huge number of issues that need to be overcome in order for them to do the simplest things.
Hardware limitations still prevent the implementation of team behaviour, such as accurate passing, or even the usage of the full set of FIFA rules with full-size fields, throw-ins and slide tackles.
This is where the RoboCup Simulation League comes in. This allows the ambitious 2050 Millennium Challenge to be decomposed into parallel fields of development:
- research using physical robots as in the Standard Platform and Humanoid Leagues
- research into complex team strategies and behaviour using artificial players in a server-simulated environment.
Although simulated games displayed on a computer screen may not sound exciting as actual robots playing, the research emerging from this competition certainly is.
While UNSW were busy serving cold revenge to the Germans in the Standard Platform League (they beat us 4-0 in the 2010 World Cup, remember!), another Australian team, the Gliders, was making a name for itself in the 2D Simulation League.
Australia’s newest sporting heroes
Alright – so we didn’t quite bring home the gold, but second place ain’t bad!
In an incredibly tight match, the Gliders lost 2-3 against five-time world champions, WrightEagle from China.
To reach the finals, Gliders defeated two-time world champions HELIOS from Japan - an incredible improvement on previous year’s performance in Eindhoven, where Gliders finished 5th overall.
These matches can be viewed here.
During my time at CSIRO in 2012-13, I was responsible for the development of the self-localisation code utilised in this year’s competition.
Making a difference in the real world
Most RoboCup research extends naturally to the wider field of industrial robotics and automation. This is a multi-billion dollar industry where the ability for a robot to “see”, “navigate” and make “intelligent” decisions is critical to both profit and the safety of their human co-workers.
But the real excitement is in the less-obvious applications. One of the beauties of research is the incredible results that can emerge through the union of two seemingly unrelated disciplines (such as Shakespearean literature and cancer diagnoses!)
RoboCup is no exception. The same information-theoretic modeling that Gliders use to measure the performance of a soccer player has been used to better understand swarming of insects, birds and fish, as well as modeling the structure and connectivity of the human brain.
The nature of a simulated-based competition also provides a powerful platform for researching statistically significant phenomena that emerge in sporting scenarios.
As an example, we recently demonstrated a new tournament structure to yield fairer results than those used in previous RoboCup competitions. We are now investigating ways of showing that this structure (which combines a classical round-robin with a series of classification games) yields fairer results in human sporting competitions.
Our proposed structure was successfully adopted in the 2014 RoboCup Simulation League competition.
The year 2050 is still a while away, and many people still believe the Millennium Challenge to be impossible. Impossible or not, Australian researchers are changing the world in its pursuit.