An Australian team won the international RoboCup competition by having its robots learn how to walk faster.
RoboCup is a robotic soccer competition designed to create a competitive, real-life environment where teams can pitch their theories and software against 24 opponents from around the world.
The combined University of NSW and National ICT Australia team, rUNSWift, took out top spot in the four-legged competition -- which uses standard Sony Aibos -- after using machine learning technology to have the Aibos learn how to walk faster. They were the first team in the competition to use that technology.
"Our software was self-organising and it learned to walk faster," team supervisor Dr William Uther told ZDNet Australia . "The dogs spent hours walking back and forth across the field figuring out how to walk faster."
The average speed of the other teams' dogs was 23 cm per second. By the end of the competition the rUNSWift dogs -- which use exactly the same hardware -- were walking at 27 cm per second, or 15 percent faster, according to Uther.
The other difference between rUNSWift and its opponents was the tactics used. Most teams spread out across the field and had a long-distance passing strategy, while rUNSWift had two dogs attack the ball and work closely with each other. This required spending a lot of time working on the dogs cooperating when they were close together, according to Uther.
The team who came third (behind Upennalisers from the University of Pennsylvania) was NUBots from the University of Newcastle. Uther also expressed admiration for the efforts of UTS Unleashed, from the University of Technology Sydney, who entered the contest for the first time.
"Australia in general does well because of the Australian attitude to research, in that it's important to have something that works rather than sounds good," said Uther. He added it was easy to get Australians interested in the competition because it is such a sports-mad nation.
The soccer competition is accompanied by an International Symposium on robotics, and Uther was particularly impressed by a paper that outlined some applications that had transferred from RoboCup to the "real world".
One discipline that stands to gain from technology developed for RoboCup is Developmental Biology, which studies the way cells grow and differentiate as an organism ages. A model organism in this discipline, Caenorhabditis elegans, is a worm that develops exactly the same way every time, and biologists have spent decades mapping out the movement and development of every cell in the organism.
Scientists used ball-tracking technology from the small-sized league of RoboCup and applied it to tracking cells of C. elegans, reproducing the work on the development of the worm in nine hours.
Another example is the localisation technology used by the robots, which was used [in conjunction with mapping technology] to safely map disused mines in Pennsylvania in order to avoid a repeat of last year's accident, where miners hit an abandoned mine shaft that had been flooded and became trapped.
Let the editors know what you think in the Mailroom.