My ride in a self-driving Uber; or how I learned to stop worrying and trust the algorithm
Uber announced that passengers in Pittsburgh be able to hail a self-driving car from the high-tech, ride-hailing service.
Parked outside a warehouse by the Allegheny River was Uber’s vision of the future: 14 Ford Fusions, each mounted with conspicuous cameras, antennae and sensors. A lumpy lidar unit, which uses light to map its surroundings, spun atop each car’s roof like a high-tech propeller hat.
Stylish? Not quite. But I was about to get into a car that drives itself. Style was the least of my worries.
This fleet of geekmobiles, clearly marked with Uber’s logo, will be deployed Wednesday as part of a test that will let Pittsburgh customers hail a self-driving car. They will roll out of Uber’s Advanced Technologies Center, less than three miles from Carnegie Mellon University, from which Uber poached some of the school’s top robotics experts.
When it opened the facility 18 months ago, the San Francisco ride-hailing start-up, which has a valuation of $62.5 billion, threw its hat into an increasingly crowded ring of automakers and tech firms racing to be first to market with a self-driving car.
Uber Technologies Inc. will be the first company in the U.S. to offer commercial self-driving rides to passengers.
“Self-driving is core to Uber’s mission of providing reliable transportation everywhere to everyone,” Anthony Levandowski, vice president of engineering at Uber, told the media at the company’s robotics center Tuesday.
While Uber gave the impression last month that it was ahead of the game, boldly announcing that it would soon offer self-driving car rides to its customers, this week it outlined a more conservative plan.
The company will invite loyal Pittsburgh customers to opt into the program. Only some will get one of Uber’s self-driving Ford Fusion vehicles when they hail a ride.
The company didn’t detail what kind of rides it will offer but said it chose Pittsburgh because it’s an “ideal environment” with winding roads, traffic congestion and weather extremes. Much of the testing has so far been done on highways and during rush hour.
Each ride is conducted by two Uber engineers. The engineer in the driver’s seat is ready to take over the car at a moment’s notice. If the car ventures into areas Uber has yet to map (less than half of Pittsburgh neighborhoods have been mapped so far), the driver will assume control.
In the passenger’s seat, another engineer with a laptop takes notes, logging unusual events and incidents for further analysis.
In the back seat, the passenger can see on an iPad what the car sees. There’s the option of seeing the route. Another view shows the speed, whether the car is in autonomous mode and the turns of the steering wheel. There’s a camera button too. Touch that, and it’ll snap a selfie.
Despite the tablet display, sitting in the back seat of one of Uber’s autonomous Ford Fusions feels a lot like sitting in the back seat for an ordinary Uber ride.
While the whirring of the lidar sensor was audible from the outside, on the inside it was easy to forget that the car was driving itself.
This was my second time in a self-driving car. The first time, I rode in Google’s autonomous vehicle. On both occasions, I went in thinking I’d be able to detect every fault. I was ready for the algorithm to screw up – to brake too forcefully, to accelerate too quickly, to drive too slowly.
None of those things happened. On both occasions, I got bored and started admiring the view out the window instead.
We drove through the city’s industrial Strip District. The engineers pointed out what the sensors were detecting and how the car was responding to its environment. We followed the river to the edge of Downtown. We sat at a traffic light. We navigated through busy intersections. We swerved to avoid a truck that had ventured into our lane.
The driver took over only on one occasion — when another vehicle ran a stop sign. And it was only obvious that he’d taken over because I could see his arms move. The rest of the time, they remained still, his hands barely touching the steering wheel as it turned back and forth on its own.
The engineer in the passenger’s seat said the vehicle’s sensors detected that the other car was running the stop sign. Our car would have stopped, he said, but the driver intervened anyway, just to be safe. He noted this on his laptop.
Automakers such as Ford, which are also developing their own self-driving cars, don’t expect to deploy this kind of technology without drivers behind the wheel until 2021 at the earliest. Lyft, which is also in the game with GM, says it’s probably 10 years away. Volvo is working with Uber, but it’s also developing its own self-driving technology that it expects to be ready in 2020.
Even Uber doesn’t know when its technology will be ready to hit the road without an engineer behind the wheel. There’s also the matter of autonomous vehicle regulation in the U.S., which remains up in the air as it awaits regulatory guidelines from the National Highway Traffic Safety Administration. So why is Uber doing a pilot now?
“It’s an opportunity for real-world testing,” said Emily Bartel, a product manager at Uber’s Advanced Technologies Center.
The company isn’t just testing the technology’s efficacy — it’s trying to gauge how customers will react once they’re in the cars. Maybe they’ll freak out. Maybe they’ll have valuable feedback.
Or maybe, like me, they’ll quickly forget that they’re putting their lives in the hands of a computer.