IT PAYS to field heavies and rough up your opponents a bit, even if it costs
you a few yellow cards. At least, that鈥檚 partly how a team from Cornell
University in New York beat 18 other teams to win the Robot Soccer World
Championships in Stockholm last week.
Robots tend not to play football very well, because they have trouble dealing
with unpredictable events. 鈥淲hen a robot bumps the ball or kicks the ball, it
can go in any direction,鈥� says Bart Selman, a computer scientist at Cornell who
helped students design their robotic stars.
Not only must the robots process
pictures of the playing field, search for the ball within the picture and
extract its location and heading, they must also decide how they are going to
move to intercept the ball and execute that decision, all in a matter of
milliseconds. That鈥檚 quite a task. And if the ball鈥檚 trajectory deviates from
the expected position, the calculations have to start from scratch again.
Advertisement
Cornell鈥檚 soccerbots鈥攅ach rectangular, 18 centimetres high and weighing
about 2 kilograms鈥攃oncentrated on defence rather than attack and the
strategy worked. 鈥淭he Singapore team constantly attacked, with lots of shots on
goal, but being aggressive can get you into trouble,鈥� says Selman. 鈥淵ou are at
risk of not knowing where you are or of losing the ball.鈥� Cornell beat Singapore
6-2 and went on to beat the Free University of Berlin 15-0 in the final. The
German strategy鈥攈oofing balls straight at the goal at 6 metres per
second鈥攚as stymied by the Cornell team鈥檚 ability to learn and adapt to
different styles of play.
But several of the Cornell machines received yellow cards for bullying their
opponents. 鈥淲e had to take out one of the more aggressive robots to prevent a
red card,鈥� admits Selman.