Robot Cars - the DARPA Urban Challenge

11 November 2007

Interview with

Chris Vallance

Chris - Now it's time to take a look at the world of technology. This month Chris Vallance is going to introduce us to the DARPA Urban Challenge. This is a competition where people develop cars that can drive themselves around town. The Final took place on November 3rd. The competing vehicles had to compete on a sixty mile journey, obeying all the rules of the road and avoiding other drivers. This didn't happen on public roads because there were one or two accidents. They used an urban military training facility that was in California. It was a big success and the reason they're trying to do this is to develop driverless vehicles that can be used in war zones. There are likely to be some other big spin-offs for the average driver like you and me.

An Urban Challenge VehicleChris Vallance - I'm obsessed with robot cars. It was the DARPA Urban Challenge. The reason I like this is that I saw the first challenge that DARPA did. DARPA is the Defence Advanced Research Project Agency. It's, if you like, the 'brains trust' of the US military. It's their geek squad. They ponied up two million dollars to see if someone could produce a robot car to drive around a simulation of an urban environment on its own. These aren't remote controlled cars. These are cars with computers on and laser scanners so they can see what is ahead of them. They intelligently drive themselves from point A to point B.

Chris - Have they had much success Chris?

Chris Vallance - Yes, it was won by a Chevvy Taho by the name of Boss produced by a team from Carnegie Mellon University in Pittsburgh. It's quite a technological feat. I remember back in 2005 they first tried this trick with what's called the DARPA Grand Challenge. That was out in the Mohave Desert. There was absolutely nothing for miles around. There were dozens of teams trying to compete and driving round the kind of desert terrain, that I think if you or I were driving, we'd come a cropper in 5-10 minutes.  These cars actually drove successfully around that track.

Chris - So what are the challenges that the people who are doing these challenges are going to overcome? It doesn't sound like rocket science to make a car follow a pre-programmed route.

Chris Vallance - It's not pre-programmed. They have GPS way points. They know roughly where they've got to go but there's not a pre-programmed route. They've got to navigate obstacles for themselves. If the first DARPA challenge that was just navigating rocks and bits of rough road - not driving off cliffs or hitting bridges. In the latest challenge, the Urban Challenge, they had to navigate around other vehicles. They had to stay in the right lane in the road and deal with an urban environment.

Chris - That's a whole new ballgame. Sounds like a recipe for disaster. Were there any spectacular wipe-outs?

Chris Vallance - Well, there was a spectacular wipe-out. One of the really interesting vehicles from the Defence Agency's point of view was a vehicle called Terramax. It's a big, big truck that tried to drive into a disused shop. The cars do get a bit confused. It's not a perfect science by any means and it is quite a difficult computational and object-detection problem that these teams have to face. If you look at these vehicles they are absolutely covered in scanners. They have these things called lidars, laser detection systems and radars. They have all kinds of detections systems to help them figure out what's the road and what isn't. If you think about an environment where there are lots of drivers about that becomes really, really crucial.

Chris - How do we think that this information that goes into making these cars able to do this on their own could be translated to making my life, getting to work in Cambridge, through a horrible rush hour a lot easier?

Chris Vallance - Well, funnily enough the DARPA are primarily interested in the military implications of this stuff. If you think about the casualties in places like Iraq there are people in convoys hit by improvised explosive devices and there's actually a mandate from congress for a certain number of US military vehicles to be robotic by a certain date. In terms of the kind of cars that your or I drive, they are very keen to have this technology work its way into regular vehicles. This is Chris Earnson from the winning team of Carnegie Mellon University talking about how he sees some of this tech eventually working its way into the regular vehicles that you or I might drive.

Chris Earnson - We're hoping that you'd get to a point where the car that you're driving won't crash. You could always drive it if you want to but if there's something going wrong it might help you and prevent the collision. In the morning when you're commuting to work you wouldn't have to be behind the wheel. You could be having a nap or checking your email or on the phone and doing that safely. You'd be trusting the vehicle to get you where you're going.

Chris - He's talking about avoiding crashes, Chris. What about getting into tight parking spaces because we have a nightmare time when someone's parked and they've left a just slightly bigger space than my car. I'm not brave enough to go into those spaces sometimes. It technology able to help?

Chris Vallance - I think that's one of the things that we'll see this technology help with in parking assist. There was a Japanese motor show very recently where we had all kinds of vehicles that could park themselves. That's definitely one of the priorities, parking. Who wants a ding on your car when you're trying to get into a tight space?

Chris - Who indeed and we've heard from Paula in Lowestoft. She says, "cars using sat-nav these days often get lost. Just look at that lorry that got jammed in a country lane for three days recently. Will we end up with robot cars driving around getting lost everywhere?" Probably, is the answer to that one.

Add a comment