The World Cup, the pinnacle of soccer, starts this June in Brazil. NPR science correspondent Joe Palca will be one of those obsessed, screaming fans. It's not often Joe gets to do a story that mixes science and soccer, but as part of his new project, Joe's Big Idea, he found a computer scientist who actually studies soccer using robots as players. So Joe felt compelled to investigate.
I love watching soccer matches. I really do. I get it why soccer is called "the beautiful game". It's played
with a mixture of speed, skill, and cunning. Robot soccer, on the other hand is not quite so beautiful. Alison is a two foot tall robot. She’s made of white plastic and looks like a robot. By robot standards, she's a scoring machine. There's a right foot kick, and the ball is heading, and goal. Oh, that was exciting. As I watched, Alison scored several times into an empty net. But by human standards, well, how do I put this gently, I've seen toddlers do better. She got up, she fell, she tripped over, she took oh, she's got a little balance problem. Despite the clunkiness, Professor Peter Stone thinks robot soccer is also a beautiful game. Professor Stone is a computer scientist at the University of Texas at Austin. He has built an indoor soccer field in his lab where he puts his robot players through their paces.
You shouldn't be surprised that a serious scientist is studying a game as part of his serious research. Computer scientists have been doing that for a long time, soccer's just the latest. In many ways, soccer raises things up to a completely different level compared to chess and jeopardy and other things because it happens in the physical world. It's not just an information competition. There are two huge problems for anyone trying to get robots to play soccer.
One is the robots themselves. Professor Stone isn't working with the most agile humanoid robots ever created. Those are generally research prototypes that are much larger, expensive to design and build, and hard to come by. But even the newest robot wouldn't be mistaken for a Brazilian soccer star. The other problem is computational. "If you gave me a robot that is as agile as a person right now, I don't think that we could just, you know, immediately put it on a field and it would be able to beat people yet".
Professor Stone says he and his colleagues are still learning how to turn the rules humans use in soccer - rules they seem to know instinctively - into a set of instructions that will inform a robot what to do. "The robot has to make the right decision at all times, not just some of the time and we spend a lot of our time when the robot does something that we're not pleased with we want to go and figure out what were the conditions that caused that, and that really forces us to analyze the decision-making part of the code very, very closely".
Judging by the extremely clunky robotic players today, it's hard to believe a team of humanoid robots could ever beat a team of human players, but some computer scientists believe that really could happen as soon as the middle of this century. Stone says don't bet against it. "From the Wright Brothers first flying an airplane to landing a man on the moon was on the order of 60 or 70 years; from discovering the double helix to sequencing the whole human genome is about 50 years; from the first computer to beating Gary Kasparov, the world chess champion, is 50 years". So maybe a robot soccer team will beat a human team by 2050, or 2060 at the latest "but one of the lessons you learn when making time predictions is, you know, the safe thing to do is predict a date after your career is likely to be over. Because otherwise people will come back and say, you said it was going to happen by now". Yes, I predict NPR will cover it when it happens. I just don't have to do it. I'll be sitting in a rocking chair some place. I hope. Joe Palca, NPR News. There's a right-foot kick and the ball is heading for the goal but didn't quite make it. Alison follows up, again, lines it up, and goal.