found the dogs had better speed and ball control when
walking on their forearms.
years ago, Alan Alda and the FRONTIERS team attended RoboCup
1999, a competition among computer scientists the world over
and their teams of soccer-playing robots. At this contest
in Stockholm, we discovered small, speedy robots that were
not only physically able to dribble and kick. These machines
were programmed to make the decisions necessary to play the
game, independent of their programmers. For professors Manuela
Veloso of Carnegie Mellon and Raffaello
D'Andrea of Cornell, RoboCup was the perfect challenge
for their talented students. After many exciting matches,
Cornell's robots were able to win the day.
forward to RoboCup 2001 and the level of play is decidedly
more advanced. Alan joins Raffaello in the stands to watch
the Cornell team fight to retain their title. As in the 1999
contest, each team's robots are controlled by a central computer
system which uses a camera mounted above the playing surface.
The students teach their computer to see the ball and anticipate
certain scenarios for the robots to carry out, all in a split
second. Each year, the most advanced teams are not just quick
and agile, they're able to work together effectively on the
field. To Raffaello's disappointment, Cornell finds a team
from Singapore a formidable foe in a close semi-final, and
the robosoccer champion falls, 1-0.
and Raffaello D'Andrea get caught up in a tense match.
in another category, teams of soccer-playing robotic dogs
have also made incredible progress since their debut in 1999.
The dogs (who, unlike the wheeled robots, function independently
of each other as well as their human programmers) are now
faster and better able to pass and shoot with precision. In
this category, and in every other, the teams share their software
secrets at the end of the competition. So each year, entirely
new breakthroughs are achieved, building on the collective
innovations of the past.
more on this topic, see the web features:
The Future of AI