Q&A: Google Self-Driving Car Head Chris Urmson on Building a Car From Scratch
Google on Tuesday revealed a project to build prototype cars that drive themselves and forgo most of what makes a car a car. No steering wheel, no accelerator pedal, no brake pedal, no mirrors. Rather, a bunch of sensors and an onboard computer that knows how to interpret them.
For the past five years Google has retrofitted existing cars to experiment with autonomous driving. They look mostly normal except for a spinning laser mounted on top. Not these cars — they look and feel totally zany. They are small and round, studded with little black sensors (and a spinning laser on top, too), have a foam front bumper and a big flexible windshield, and are controlled by a mobile app plus just two physical buttons: Go and stop.
The head of Google’s self-driving car project, Chris Urmson, spoke with Re/code about the concepts behind the evolution of its self-driving car project and when you might get to jump in one.
Re/code: Can you explain how Google got to the point that it’s building its own cars?
Chris Urmson: We’re really excited about the idea of fully self-driving vehicles. As we started down that path, we came to the realization that to do that right, you really want a vehicle built from ground up.
What can you do in a Google-built car that you can’t do in a retrofitted car?
These prototype vehicles will have new sensors, some new radars that allow them to see further, some new lasers that allow them to see further and some new cameras as well. For example, the Lexus vehicles have this wonderful 360-degree view on the top, but because of the geometry, there’s parts right up close to the side of vehicle that they can’t see. And so the prototype vehicles will have a laser sensor mounted on the side of the car that will that allow them to see right up to vehicle. They’ll also have safety built into the vehicles from the behaviors we’ve learned from the Lexus, things like waiting at a traffic light for a second after the light turns green. We call it defensive driving.
But what happens if you’re riding in a self-driving car with no steering wheel or brakes, and something goes wrong?
In a normal car there’s power steering and power brakes, and if the power steering fails, as a strong person you can use your muscles as a fallback to still steer the vehicle. In our car there is no steering wheel so we have to design really fundamental capabilities. So we have effectively two motors and they work so if one of them fails the other can steer, so the car can always control where it’s going, and similar with brakes.
So there’s two systems, but neither of them are controlled by humans.
That’s right. We call it fault-tolerant architecture. The last layer is going beyond conventional safety. We have crash protection for the occupant, but the thing we’re really excited about is protection for pedestrians and other road users. We want to make sure should something happen, there’s the best possible outcome and anybody is hurt as little as possible. So the front end is a whole new approach where it’s compressible foam and a flexible windshield that should do a much better job of protecting people if an accident should occur.
To me, that seems interesting but a little adjacent to the fact that it’s a self-driving car. Are you building the front like this because your car is supposed to be a city car, or what?
Two things. The first is, as a society we need to do a much better job at protecting road users. So if you look at worldwide statistics, there’s 1.2 million road users [killed per year]. Around 40 percent of that number is pedestrians and cyclists. So yes, there’s an awful lot of room for improvement there, just generically.
Second, we know this is a new technology we’re bringing into the world and we want to do it as safely as possible. So we took it upon ourselves to think about how can we have an independent layer of safety for people out there in the community. We’re going to do everything we can to design the software and hardware to never fail, but should it fail or should someone step out in the road at a distance where they couldn’t take an evasive action, we want to reduce the impact.
The last thing is that we’re going to limit these vehicles to driving to 25 miles per hour. What’s a really interesting statistic is if you look at the kinetic energy that happens in a crash, 25 miles per hour is half of 35 miles per hour, because it’s the square of speed. And so by going a bit slower, that’s a huge safety win.
How did you scrape this particular shape together?
This vehicle was a first cut at building a basic frame for the prototype vehicles we’re developing. The point is to get what we we call the mechatronic parts — the motors, the electronics — to allow those to talk amongst themselves and start working properly. And then we pulled together the upper body — the greenhouse — and it’s approximately the shape we think the prototype vehicles will be, and then we put some seats on it and the computers and the sensors on it.
Who is going to drive these cars? I think you said earlier there will be at least a hundred of them over the next couple years, starting this summer.
That’s right. At first they will be used like we use the Lexus vehicles to test the software and to push what we can do with that capability. And then at some point when the software is safe we’ll start to get some pilot tests with people from the community. The next thing is the opportunity to really understand how people use them. With these vehicles, the user doesn’t have to worry about the minutia. They worry about where they’re going and what they’re doing along the way — what they’re reading, who’s talking. We’re going to be using those vehicles to explore that experience and what it means for the community.
So you’ll allow people to go in this car who don’t work for Google? When?
Yes, just like you and Kara. We don’t know exactly, but as soon as possible.
What’s the reaction you hope to get when someone is walking down the street and sees one of these cars? It looks like you want people to think it’s cute.
We definitely would like the vehicle to appear friendly. When you look at the front grill of any car, there’s a lot of thought put into that shape and what kind of emotion it shows. Many of them look like faces. In our case we wanted to find something that’s very Googley. It’s friendly, it’s kind of cute. We hope it fits into neighborhoods.
Do you have any idea what something like this will cost?
We’re still figuring that out. Right now these are prototypes.
Well, at least I didn’t ask you if you’re going to put ads in it.
You just did, I think. We just don’t think that it makes sense.
There was some rule-making this week about autonomous vehicle testing in California. What’s your interpretation of what that means for a car like this, when can people sit in it and not pay attention to it driving?
Today in California, if the technology was safe, you could do that. The law that was passed almost a year and a half ago made it quite clear that effectively driverless operation of vehicles was permitted in California and in general we believe that’s true across much of the U.S. What will happen in the not-too-distant future is the California DMV will issue regulations about the operation of self-driving vehicles, and I believe in the law there will be a a clause requiring a 6-month notification period before vehicles without drivers are allowed on the road.
If a human isn’t required to drive it, what about the legality of vehicles without humans entirely?
We believe today that yes, you can do that. This comes back to what we see as the big goal. If you have someone like Steve Mahan [an early Google self-driving car beta tester] who is blind and has challenges getting around, having a vehicle where he has to jump in to help it out periodically is just not going to help him. So a vehicle that can come to him, and take him where he’s going, and let him get on with it, is really what we’re going for and what we’re excited about.
Does that mean more or fewer vehicles on the road?
My vision for this is eventually these vehicles will be shared, and it may be within a family, or it may be within a community, and that will result in less vehicles on the road, but they’ll be used much more efficiently, and that’s good for everyone.
When I told my editor I was working on a story about how Google had built a self-driving car, his reaction was, “Uh-huh, OK. I thought they already did that.” It didn’t evoke the sort of wow factor that we had gotten from actually seeing what you built. So how would you describe why this is impactful in a simple way?
For me it’s the first time we’re seeing a real physical incarnation of our vision. As Kara put it, we’ve had a vehicle with “glommed on” sensors and it’s been great, and it’s a way to learn very quickly and to push technology forward in a way we’re excited about. But now we’re taking the step of how it’s really going to work. I have spent a lot of time in self-driving vehicles, and getting into this vehicle was just freeing. For us, it’s this physical thing in the world that embodies what we’ve been working towards.