Australians are heading to Italy today to take part in a robot soccer competition designed to improve the way robots interact with the environment.

The winner of the Australian RoboCup competition in May, rUNSWift (a combined team from the University of New South Wales and NICTA), is expected to place in the top three teams worldwide. Over the last four years they have won twice and come second twice.

RoboCup — in which robot teams battle it out in soccer-like games — has serious application, since the skills required to compete in a game of soccer are necessary in just about every situation that involves an autonomous robot. The capacity to move, to detect objects and to respond to changing situations are all necessary.

“There is a lot of research involved in RoboCup,” Dr William Uther, of NICTA’s Robot Learning Research Group, told ZDNet Australia  . “One important part of RoboCup is that it gets the academics out of the ivory tower – you have to produce something that really works in practice.”

“In general a robot must have three parts, sensors that detect and interpret the world, strategies for behaving in the world, and controllers and effectors that execute those behaviours,” said Uther. “The chain analogy holds very strongly here: you are only as good as your weakest link.”

The UNSW/NICTA team focus on machine learning, since a machine that can learn doesn’t need to be programmed for each separate task, only told the desired outcome.

“Other teams concentrate on different things. Some teams are more interested in co-operation amongst the robots on the team. Last year Carnegie Mellon University in Pittsburgh introduced learning during the game in the Small Size competition,” said Uther, explaining that previously all learning had occurred before the game.

The rUNSWift team will compete in the four-legged competition using Sony Aibo’s (robot dogs) with their software modified. The team consists of Uther, Prof Claude Sammut from UNSW Computer Science and Engineering, Bernhard Hengst from NICTA and students Nathan Wong, Eric Chung, Ross Edwards, Jin (Ricky) Chen, Raymond Sheh and Eileen Mak.

Other Australian teams in the four-legged division include Mi-Pal from Griffith University, Nubots from the University of Newcastle and UTS Unleashed, from the University of Technology, Sydney. The Simulation division – where software programs compete in a virtual environment – is being entered by Cyberoos2003
from the CSIRO and DSL_UNITED from the University of Wollongong.

GUROO from the University of Queensland will compete in the humanoid robot division, RoboRoos from the same University will compete in the small-size robot division and MU-Wallabies MU-Wallabies from the University of Melbourne will compete in the middle-size robot division.

The University of Auckland has a team — The Black Sheep — entered in the Rescue Simulation division.