A team of UNSW software engineers has won the bronze medal in the annual RoboCup competition, where autonomous humanoid robots compete for supremacy on the soccer pitch.
Although they came up just shy against the eventual tournament champions from the University of Texas in their semi-final match, losing 7-6, UNSW's robots tallied more goals than any of the 24 teams in the competition. (A video of the semi-final match is viewable here).
The squad - known as rUNSWift - competed in the standard platform division of the competition and racked up 62 goals while limiting their opponents to just 13 scores.
Brad Hall from the School of Computer Science and Engineering at UNSW says the objective of the international competition is the transfer of machine learning technology to things like rescue robots and the overall advancement of artificial intelligence.
Sean Harris, a first-year PhD student who is investigating machine learning, travelled to Mexico for the 2012 competition - his second RoboCup with UNSW. He says the competition is progressing quickly each year: "In 2008, the first year using the Nao platform, none of the robots could walk properly and nearly all of the games were 0-0 draws."
"One of the only goals in the tournament was because a robot ran out of batteries, fell over and knocked the ball into it's own net," he says. "This year we saw games with as many as 13 goals being scored, so it's definitely come along way."
Harris says there are four main areas of focus that programmers must work on to develop a winning team: vision, localisation, mobility and behaviour.
The robots must be able to detect different things in their environment such as other robots, the boundary lines on the field, the ball and the goal. They do this by looking for edges and colours, says Harris.
By using their vision, the robots can work out where on the field they are in relation to the objects they detect - something called localisation. This is important because they can communicate over wireless and share this information with teammates.
"The robots are essentially talking to each other and can make informed group decisions about things. They have a belief model about where the ball should be, and where their teammates and opponents are positioned, and they move based on this," he says.
Researchers also look at how they can enable faster, more stable movements, and how they can program the robot's behaviour so it can make decisions based on information about the current state of the game.
It is the 13th year that UNSW has participated in the international competition and the 10th time they have placed in the top three.
Related Links:
http://www.facebook.com/media/set/?set=a.10151208455452738.543688.607287737&type=3&l=31d99963bc