Self-driving Cars Heading Our Way
Smart vehicles could make the roads safer and ease congestion.
Susan Hassler: Another benefit of life in 2030 will be “autonomous,” also known as “self-driving” vehicles. Many major automotive manufacturers have begun testing driverless car systems. And, Google says it’s suite of self-driving Priuses have logged 300 000 miles, 50 000 of them without any intervention from human drivers.
Phil Ross: Of course, Google is the major driver in changing state laws to accommodate autonomous vehicles. California has passed a law mandating regulations for self-driving cars by 2015. Florida and Nevada have passed similar laws. So, it’s on our radar.
Susan Hassler: Our next story is from reporter Laurie Howell who says she has always fantasized about having a chauffeur, so why not a self-driving car?
Laurie Howell: Hi there. Well, you know, whenever you’re working with life-altering technologies, one of the biggest challenges is how to play well with humans.
Ümit Özgüner: All those other human beings will not necessarily be driving by the rules, and if your car is going to survive, it has to accommodate what strange behavior, what oops factors that you have to deal with.
Laurie Howell: Electrical Engineer Ümit Özgüner and his team at the Ohio State University are working on integrating self-driving cars with human drivers in urban environments—from small communities to big cities.
Ümit Özgüner: Can we coordinate everything and be able to send overall directives and guidance and sensor data reliably all the time to, let‘s say, tens of thousands or a million different mobile entities.
Keith Redmill: So, this is our indoor lab where we do some initial testing of software that we design to control automated vehicles…
Laurie Howell: Electrical engineer Keith Redmill gives me a tour of the team’s mock urban environment. We’re in a room about 1200 square feet—with a painted road that winds around boxes representing buildings. There are intersections, stop signs, traffic lights—even an overpass.
Keith Redmill: We have these small robots that—we‘ve programmed them to act sort of like cars in the sense of how they move and how they turn and how they stop and start. Even though they don‘t really look like cars, they act like cars and they run software that‘s compatible with the control software on the real cars
Laurie Howell: Researcher Scott Biddlestone starts them up.
Scott Biddlestone: You give them a list of points you want them to go through and they just try to do it however they can. You‘re always going to get a little bit of variation in how they move around.
Keith Redmill: What we‘re trying to test here is the decision-making and the planning and the reactions that are built into the software.
Keith Redmill: Variability is what we gain by running an experimental setting as opposed to running a simulation. And, as Scott says, he can drive these around and interfere however he wants to with whatever experiment is running and see what the software does.
Laurie Howell: Biddlestone is working on what they call “convoying.”
Scott Biddlestone: If you were driving around a road, how would you make a group real quickly and then convoy so you can get through a traffic light together faster—and hopefully make a smaller gap, keep people closer together, and get more efficient at a traffic light or on highway.
Laurie Howell: If convoying sounds far-fetched, have you heard about one automaker’s new “traffic jam assistance function”? It allows a car in stop-and-go traffic to maintain a set distance from the vehicle in front.
Laurie Howell: After lab testing, the software and sensors are tried on actual cars at the university’s Center for Automotive Research.
Laurie Howell: Over the noisy engine exhaust fans from a research project next door, Keith Redmill is walking me around the vehicles. There are a few parked in and just outside their two-car garage. First up is the star player—an SUV loaded with sensors, radar, lidar, laser range finders, and computers.
Keith Redmill: Right now, we are using this car to test vehicle-to-vehicle communication. So we have this car—we have another car that‘s partially automated. It doesn’t have steering capability but it has throttle and brake. And then we have a couple of just manually driven cars that we can have people drive around us to interact with this car and do maneuvers, cut it off.
Laurie Howell: And, just as in the lab demonstration, I saw earlier, they have a traffic light that communicates—this one life size. So, we’re not just talking about the cars of the future.
Keith Redmill: You know, traffic lights are called infrastructure. We expect that in the future, maybe not in 5 years, but in 15 or 20 years, things like traffic lights will be broadcasting possibly all kinds of information.
Laurie Howell: In the urban environment of the future—traffic lights and other aspects of the transportation infrastructure will be communicating with smart cars, which will be communicating with each other. There will be less risk for driver error—and it’s driver error that is responsible for 90 percent of crashes. That’s according to the National Highway Traffic Safety Administration, which also reports that more than 32 000 people died in vehicle crashes in 2011.
Ashok Krishnamurthy: And I believe that increasing automation in vehicles is going to be driven from safety considerations more than anything else, not just because the technology exists but because of safety.
Laurie Howell: At Ohio State’s Supercomputer Center, electrical engineer Ashok Krishnamurthy deals with the signal processing
Ashok Krishnamurthy: I look at the signals and then I produce estimates of what the behavior of the vehicle is going to be or based on the data that I‘m observing.
Laurie Howell: Krishnamurthy runs a key aspect of the research—modeling the behavior of human drivers with sensors that record things like how long it takes them to slow down, when they signal.
Ashok Krishnamurthy: You take a variety of subjects and make them go through these maneuvers multiple times. You collect the data and then you then have a machine-learning algorithm—and use that to train these models which then understand. And then what you do then is once you know that, then you can say, “Oh, in this particular case, I don‘t know what they’re going to do but I‘m going to collect the data and I‘m going to use this model to make a prediction on what they‘re going to do.”
Laurie Howell: Everyone on the team agrees that safety will be the No. 1 issue driving us toward autonomous vehicles. Other major benefits are convenience, fuel efficiency, and better traffic flow. But, no one is quite ready to say that self-driving cars will be fully integrated with human drivers by 2030.
Ashok Krishnamurthy: I can certainly see that in 2030 that we will do less driving, that the vehicle will drive itself a lot more, and perhaps we will be in a much more supervisory role in terms of how we travel the vehicle.
Ümit Özgüner: Right now, what we can more think about is sort of specialized lanes and highways or small closed environments like a park or Disney World or some research center with closed roadways and so on or an island. These can all be fully automated and I do imagine that by year 2030 some of those we‘ll see.
Laurie Howell: I’m Laurie Howell
Male Car Voice: Ms. Howell.
Laurie Howell: Uh huh.
Male Car Voice: You will arrive at your destination in 3 minutes, I am pausing your iPod and shutting down the Internet.
Laurie Howell: Uh huh, Okay.
Male Car Voice: Unless, of course, you’d like me to park for you as well?
Laurie Howell: Right.