Telling robots what to do

David McKinnon of the School of Computer Science and Engineering, UNSW, points to a small plastic humanoid robot that stands atop a desk.

DAVID MCKINNON: So to get the robots to do what we want them to do, we program them with a language called C++, and basically the way we do it is we get input from the cameras and the microphones, and there's also an internal gyroscopic sensor that senses whether you're leaning one way or another. We get all that information, then analyse it, so we'll look through the camera images and go, 'Is there a ball in here? Can we see goal posts? Can we see other robots? What parts of the field can we see?' Then we'll make decisions based on that information, so we'll have previously looked around the field and go, 'Well, I think I'm in this corner, there's a ball right in front of me and the goals are just over there. I'm gonna kick it over that way.'

Two small plastic humanoid robots move around a green scaled-down soccer pitch, following a small orange ball and kicking it.

DAVID MCKINNON: Then we'll send a command to the leg motors that take it through a kicking motion, so it leans up on one foot, swings its leg back, forwards, then steps down and keeps walking. It'll also send information to the other team members so, 'I've just seen the ball in front of me, everyone else, did you hear that?' Then they'll go, 'Yep, we know where the ball is now. We're watching up there and we're ready to move in case it comes near us.'

David McKinnon speaks to interviewer.

DAVID MCKINNON: So we've written it all beforehand basically in this huge decision process, essentially, that goes, 'What's our input? What's in this input? What are we gonna do based on it?' Sorta the way humans work. I see you, you ask me a question, I hear that, I talk to you. We've written all that beforehand and then we load that onto them, hit play and it just keeps ticking through that the whole match, running through that program, through all those decisions and thus they play.