KTEP - El Paso, Texas

Pittsburgh Offers Driving Lessons For Uber's Autonomous Cars

Apr 3, 2017
Originally published on April 4, 2017 2:13 pm

Ever wonder what it would be like to live in a laboratory? People in Pittsburgh could tell you it's not so bad. They've been sharing city streets with Uber's experimental self-driving cars since last September. Six months in, no one has been hurt and there have been no major accidents. Plus, the project is bringing in investments and boosting the city's reputation as a tech hub.

While there are only a handful of autonomous vehicles climbing Pittsburgh's hills, crossing its many bridges and boring through its tunnels, Uber says it's learning what a tough, old city like this one can teach the ride-hailing company.

Uber also tests cars in Arizona, but on terrain that's less challenging. Even so, last month a human-controlled car smashed into an Uber self-driving car and tipped it on its side. After a brief suspension, Uber's self-driving vehicles were back on the road.

"It's still very early in what we're doing, but it's going very well," says Uber's Emily Duff Bartel. One reason may be that each autonomous vehicle comes with two Uber employees in the front seat. Just in case. One is ready to grab the wheel and apply a foot to the brake pedal. The other, in the passenger seat, has a computer screen showing what the car's rooftop laser-bouncing radar is seeing.

"It's still very early in what we're doing, but it's going very well," Uber engineer Emily Duff Bartel says.
Art Silverman / NPR

"Their primary focus is 100 percent of their time to be monitoring the system," Bartel says. Last month, Recode reported that, on average, these Uber autonomous vehicles required a human operator to intervene in the driving every 0.8 mile.

When NPR took a short spin in one of the Uber vehicles, the car behaved like a very cautious driver, stopping for a full three seconds at stop signs and never going above the speed limit. The Uber humans had to take to get around a bulky, parked 18-wheeler and to get the car into and out of the Uber parking lot.

When this testing began, Mayor William Peduto showered praise on the company. Over the six months, he's become a bit more critical because he thinks Uber is not sharing what it's learning.

"If we were able to utilize what Uber has and enhance it, we could create not only a more efficient traffic system in the city of Pittsburgh, but a much, much safer system," Peduto says. He would like to have data on potholes, where pinch points are and what traffic patterns are.

Pittsburgh Mayor Wlliam Peduot, an early supporter of the project, has become a bit more critical because he thinks Uber is not sharing what it's learning.
Art Silverman / NPR

Nonetheless, he's happy for the investment the company has made in the city. City Councilman Corey O'Connor is happy, too. Uber's self-driving test facility is in his district. Its facility is located in an old industrial site.

"That's important because now that site has become a worldwide site," O'Connor says.

Another happy Pittsburgh institution is Carnegie Mellon University. At first, when Uber scooped up dozens of faculty members to work on the self-driving car project, the school had to scramble. But the benefits soon outweighed the loss.

"That fact that Uber is out here investing megabucks makes the region very attractive," says Carnegie Mellon electrical and computer engineering professor Raj Rajkumar. "It retains our graduates — a brain drain from Pittsburgh to Pittsburgh."

Professor Raj Rajkumar of Carnegie Mellon's College of Engineering believes that Uber's presence in Pittsburgh helps the city keep its graduates in the area.
Art Silverman / NPR

Another positive report comes from bicyclists. Scott Bricker, executive director of Bike Pittsburgh, says members of his group feel a lot more comfortable around autonomous vehicles than human drivers.

"They can guarantee that the [autonomous vehicle] is not inebriated, that it is not programmed to drive aggressively," he says. The autonomous vehicles in Pittsburgh are programmed to maintain a four-foot distance from bikes.

While Uber's vehicles are practicing good traffic etiquette, becoming self-reliant is a hard climb. Rajkumar, an early engineer on self-driving cars, says that a world where such cars are the norm is a long way off.

"All these predictions that people made, including myself, going back several years, it turns out we are not there yet," he says.

"It turns out that driving is a very complex activity. In fact, it may be the most complex activity that most adults on the planet engage in."

Even harder for autonomous cars to master are local quirks; things like "the Pittsburgh left." That's a custom unique to the city that allows the first driver trying to make a left turn to do so before oncoming cars pass through an intersection. It's one thing that Emily Duff Bartel says Uber's fleet of self-driving cars won't be programmed to practice. "We're following the rules on this one," she says.

That's because an autonomous vehicle is a rational machine, so you can't make it do human tricks. A real city like Pittsburgh is an irrational obstacle course, where the obstacles are sometimes at rest and sometimes in motion. We learn rules of the road that we then commonly violate. And we often use eye contact to communicate with pedestrians and other drivers.

For all the automotive industry's enthusiasm for autonomous vehicles, Uber's Pittsburgh experiment is still truly experimental.

Copyright 2020 NPR. To see more, visit https://www.npr.org.


What do Detroit and Silicon Valley have in common? Well, everyone from Ford and GM to Google and Uber is developing an autonomous vehicle - a self-driving car. That's our subject for today's All Tech Considered.



Driverless vehicles are becoming a reality. Some have been out in the streets. Uber has some that actually answer calls, although there's always one human in the driver's seat and another sitting next to him or her. Robert went to find out what people make of all this.


SIEGEL: This is a laboratory. It's called Pittsburgh, Pa. I'm standing in Market Square Park, seeing the sights of a - of a city - traffic, cars, buses, pedestrians, some bicycles, a lot of people jaywalking. In this city of three rivers, umpteenth bridges and tunnels and hills - one of them so steep, it's served by two for funiculars they call inclines here - this is where Uber is testing a handful of automated vehicles. How's the experiment working out?

UNIDENTIFIED MAN #1: Once it's safe, and, like, it's cost-effective, I think, yeah, it'll be around more.

UNIDENTIFIED MAN #2: I've seen a few, and it looked like it took two people to operate it. It looked like there's always someone sitting in the passenger seat.

SIEGEL: Would you think it would be cool to be in a self-driving car?

UNIDENTIFIED MAN #2: Kind of (laughter). I like to have control of the car.

UNIDENTIFIED MAN #2: What do you think about the idea of autonomous vehicles?

UNIDENTIFIED WOMAN #1: It seems to me to be inevitable.

UNIDENTIFIED MAN #3: Oh, yeah. Oh, yeah. See them all the time.

SIEGEL: Have you ever been in one?

UNIDENTIFIED WOMAN #2: No. I've never been in one.

UNIDENTIFIED WOMAN #3: No. I've never been in one.

UNIDENTIFIED WOMAN #4: I think it's really cool. Science is amazing.

UNIDENTIFIED WOMAN #5: It is kind of scary - a little.

UNIDENTIFIED WOMAN #6: The control aspect is a little inky.


SIEGEL: That's Pittsburgher Erin Cherry (ph), and before her, we heard from Nolan Hall (ph), Susan Yowee (ph), Josh Andrusky (ph) and Momad Tayir. None of them has actually taken an Uber AV. There are, as I said, only a handful of them in Pittsburgh. That's the company's word, and they won't say how much a handful is. Uber's AVs don't take fares that would require highway driving. And if you're offered one, you have the choice to decline and take a human-driven car instead. I was eager for the ride and took one under company supervision.

MYIN PENN: We might need you to buckle up.

SIEGEL: That's Myin Penn (ph), the mandated human in the driver's seat. His hands were poised just off the steering wheel, ready to pounce. Turns out one of the toughest environments is the parking lot at Uber headquarters - no lanes, no clear rules - so those hands are actually on the wheel until we were out onto Pittsburgh streets.

STUART KRUGER: And we're in auto right now, so the car's driving itself.

SIEGEL: That's Uber's Stuart Kruger (ph), who's in the front passenger seat observing, laptop in hand. The car follows its navigation system by beaming lasers from a spinning device mounted on the roof. The system is called LIDAR. Pittsburghers call the spinning gizmo the rotating chicken bucket. It's about the size of a KFC container. The self-driving Uber stops where it's supposed to stop, sticks to its lane, maybe ventures safely outside its lane if there's a parked truck protruding too far from the curb. And whatever else this self-driving car is, it is very cautious and law-abiding.


SIEGEL: Now, it waited very patiently for the car ahead of us to get to a length or two head. It stopped on a yellow light...


SIEGEL: ...That a more bold human driver might have tried to make a left turn on or go through.


SIEGEL: At one intersection where a right turn on red was permitted it, it waited for a green light before turning. At another, a four-way stop, there was no vehicle in sight, but the autonomous Uber paused for a perfect, legally required three seconds before proceeding at precisely the speed limit.


SIEGEL: We just slowed for pedestrians.

UNIDENTIFIED MAN #5: Yeah. We just slowed down for a pedestrian right there.

SIEGEL: Literally a jaywalker - he was crossing in the middle of the street. A car turned into the street ahead of us, and we slowed again very politely to keep our distance behind the car ahead.


SIEGEL: Emily Duff Bartell is working on an AV project. She's developed the iPad screen view that shows a passenger in the back seat the image of the car's surroundings that the lasers are producing.

BARTEL: This is a representation of what the LIDAR is returning right now.

SIEGEL: Well, I'm seeing an image of our car going down the street. All of the parked cars are represented and some of the structures up ahead. We just changed lanes. At that point, the car changed lanes, not on its own, but by action of the human driver. How often do human operators intervene? Well, Uber won't say, but there was a leak to the tech website Recode last month that said that it happens on average nationwide in Uber AVs every eight-tenths of a mile. One of Pittsburgh's main attractions to Uber and other companies is Carnegie Mellon University, which is a leading center for robotics. Uber's Emily Duff Bartel is a CMU graduate. In fact, when the company came to town, it hired 42 Carnegie Mellon roboticists, depleting the department.

RAJ RAJKUMAR: This has been a very synergistic relationship between CMU-slash-Pittsburgh and Uber.

SIEGEL: That's Professor Raj Rajkumar of Carnegie Mellon's electrical and computer engineering department. He developed an autonomous Cadillac in 2011. The takeaway about making AVs - he says it's a lot harder than computer scientists think.

RAJKUMAR: All these predictions that people have made, including myself, going back several years - we don't have that. We are not there yet.

SIEGEL: An autonomous vehicle is a rational machine. A real city is an irrational obstacle course where the obstacles are sometimes at rest and sometimes in motion. We learn rules of the road, and then we commonly violate them. And we often use eye contact to communicate with pedestrians and other drivers.

RAJKUMAR: It turns out that driving is a very complex activity. In fact, it may be the most complex activity that most adults on the planet engage in. When you are the driver of a vehicle, you are consuming tons of information that is constantly changing. We are processing all that information, and we are making decisions.

SIEGEL: Back at Uber's Pittsburgh headquarters, I told Emily Duff Bartel about a complaint that I'd heard from a human Pittsburgh driver. He had seen two uber AVs doing 25 on a road where he said no one does 25, even though that's the limit.

BARTEL: We are law-abiding citizens. Yes.


SIEGEL: You're the only car on the road. That's the law-abiding citizen.

BARTEL: We do go the speed limit. Yes.

SIEGEL: Coming to Pittsburgh, I've learned of something that I didn't know existed. It's called the Pittsburgh left.

BARTEL: Oh, yeah. It's when you're sitting at a red light, and you are trying to make a left turn, the light turns green, and you basically have two ways of traffic. The first car in the left lane gets to kind of cut off the oncoming traffic to make that left turn.

SIEGEL: And everybody kind of waves and salutes as the turning car makes the left.

BARTEL: Yeah. And learning how to drive here, I didn't know you - that wasn't a thing that was permissible other places.

SIEGEL: Do autonomous vehicles in Pittsburgh get schooled to how to be a Pittsburgh driver?

BARTEL: No. No. Again, we follow the rules on this one (laughter).

SIEGEL: The great hope for autonomous vehicles is that they will improve highway safety. They don't get drunk or fall asleep at the wheel. They are immune to road rage, and they scrupulously obey the rules. After a recent collision in Arizona involving an Uber vehicle, the other car was found to be at fault.

For now, though, vehicles that cost hundreds of thousands of dollars to outfit are riding Pittsburgh streets with a backup human driver who's required to be poised to take the wheel. For all the automotive industry's enthusiasm for autonomous vehicles, what struck me most about Uber's Pittsburgh experiment is how truly experimental it is.

(SOUNDBITE OF BENGALA SONG, "JAZA") Transcript provided by NPR, Copyright NPR.