My Ride In A Self-Driving Lyft Showed How Far Autonomous Tech Still Has To Go

If you’re looking for an example of how far self-driving technology has to go before it’s ready to be adopted en masse, take my ride in one of Lyft’s first autonomous cars: On a demo ride earlier this month, the vehicle came across a bus that had seemingly stalled out. Rather than let the car figure out how to maneuver around it, a safety driver at the steering wheel of my ride took control of the vehicle and got us out of the way.


It was a telling moment. Lyft’s car—a BMW 5 Series outfitted with self-driving tech designed by autonomous car company Aptiv—could easily handle seemingly-complex driving tasks like changing lanes and making left-hand turns. But there wasn’t enough comfort yet to show how the car would do with a basic scenarios that drivers come across on a daily basis: how to get around an idled car.

As someone who covers the fledgling autonomous driving industry, I much rather would’ve seen how the car on its own could’ve handled the task. What’s the benefit of a demo ride if not to show how the car responds to real-world scenarios?

But Lyft and Aptiv weren’t trying to show reporters the limits of their tech. They had a different objective for our ride: At this year’s Consumer Electronics Show in Las Vegas, the two companies set out to showcase autonomous technology to those who haven’t been able to experience a robot-driven car just yet, and demo a possible business case for how it could be utilized. Humans aren’t exactly comfortable with the idea of self-driving cars yet, so instead of letting their vehicles attempt harder tasks, Lyft and Aptiv made sure the safety driver—a common facet of autonomous cars on the road today as a failsafe of sorts in the event the technology fails—took over. It’s a small step; if autonomous cars are going to succeed, before the technology ever develops to the point a driver isn’t needed, people are actually going to have to like the idea of autonomous cars.

In that sense, Lyft and Aptiv succeeded. As part of its pilot program at this year’s CES, the companies aimed to let any passenger with the Lyft app take a self-driving car for free to 20 specific destinations, one of many possible scenarios for autonomous ride-hailing services of the future.

It’s unclear how many people used the cars at CES, but unlike our ultimately seamless experience, it doesn’t appear that it went as smoothly as Lyft and Aptiv had hoped, based on reports out of Vegas that emerged after we took off. Plenty of general CES attendees were eager to try out one of the cars, but they all got stuck in traffic, as Automotive News’ publisher Jason Stein pointed out. Once the roads around the Las Vegas Convention Center really jammed up, it was impossible to showcase what Lyft’s rides could do.

“CES 2018: The world’s largest parking lot for autonomous rides, which prompted a retreat to the oldest form of transportation technology—your own two feet,” Stein wrote.


Had they gotten a clean demo ride, at least, passengers would’ve came away impressed by was how damn normal the cars looked.

Aptiv, a recent spinoff of supplier Delphi Automotive, outfitted a set of eight BMW 5 Series with a suite of autonomous driving gadgets. With 10 radars and nine LIDARs (short for light detection and ranging) integrated into the frame, you almost had to squint to make out where the self-driving tech was situated. It was a seamless look, not having a spinning LIDAR dome atop the car or a massive sensor rack jutting out from the frame.


“It feels like there’s a human there,” Jada Tapley, vice president for advanced engineering at Aptiv, told Jalopnik. “It looks like a car you would normally have. This isn’t some space-age thing. We can present it in a way that builds that confidence, builds that trust and lets us take that next step towards commercialization.”

Illustration for article titled My Ride In A Self-Driving Lyft Showed How Far Autonomous Tech Still Has To Go

For its part, Lyft’s trying to launch pilot projects across the U.S. to ferry people around by a robot. Vegas is just the companies’ first “point-to-point” experience.

Before going into my drive, here’s some context as to what I mean by “self-driving” in this pilot: Lyft and Aptiv, like most companies pouring billions into autonomous car tech, had a safety driver behind the wheel, ready to assume control of the car if anything went awry. The driver guided us out of the parking lot, something Lyft said was due to local regulations. I was legit kind of bummed about this; the basic idea of a robotaxi is that it would arrive to pick you up on its own and then autonomously drop you off without any intervention. Start to finish. Sleep through the whole ride. That’s what we’re all waiting for, and it would’ve been a fun, insightful experience. Still, I suppose they have to follow what’s on the books.


Once inside, we selected one of the destinations to start a ride on a small tablet situated between the front seats. Lyft was shuttling people around to 20 spots, but for purposes of the demo, rides for the media were confined to Caesar’s Palace.

Out of the lot, the driver engaged the car’s autonomous tech, which triggered a robotic voice to shout: “autonomous mode activated.” The only other quirk of the car’s interior was a display on the dashboard that showed pixelated images of what the car saw on the road.


What’s there to say about the drive? Honestly, not ... much? The car’s ability to handle Las Vegas roads is notable. The 5 Series kept within its lane easily; there was nothing herky-jerky about it. The car handled lane changes, red lights, and a decent amount of traffic. When we arrived at Caesar’s, it maneuvered its way through the crowded traffic pickup zone without a hitch, by far one of the most impressive things of the ride. I figured that’d be the obvious point where the car’s technology would get skittish, but we made our way through just like a human would.

The use of a safety driver might raise the obvious question among passengers who don’t know the ins and outs of autonomous driving technology. What’s the point if he’s still there and can drive the car, really? But these kind of pilots are a smart way to ease people into the idea of self-driving cars. The general thinking is that ride-hailing services will be the first to get automated cars, so it’s interesting that Lyft’s already trying to work through the idea of how to implement robotaxis into their app. And once you get past the presence of the safety driver, it really is a wild feeling to watch a car handle dense traffic all on its own.


Again, what I got was an impressive demo. But even then it all illustrated just how quickly current robot-car tech can be snarled today by something as basic as a traffic jam or a stalled bus parked in your lane.

The next, logical step would be for companies to send out cars without a driver. How soon could that happen for Lyft and Aptiv? It’s hard to say, but Aptiv’s CEO is confident its tech can handle operating a car without a driver by early 2019. For now, though, it’s just baby steps.

Senior Reporter, Jalopnik/Special Projects Desk



“We can present it in a way that builds that confidence, builds that trust”

They can present it any way they like but that doesn’t mean people are going to trust it.

Item #1 out of probably 20 reasons I won’t be getting in an autonomous car any time soon : LIDAR interference.

I’ve worked with LIDAR when mapping areas for military simulation. One thing commercial LIDAR systems don’t like is to be used in the vicinity of other LIDAR systems. Imagine a pair of them on opposite sides of the street. The reflected light from unit A is much weaker than the transmitted light from unit B. Unit A starts to get bogus data returns because it’s sensor is overpowered by the transmitted beams from the second unit, and it’s own reflected beams are lost.

I’m not sure anyone has done sufficient testing of LIDAR mapping systems on autonomous cars to know whether or not they’ll suffer the same issues, but it’s fair to assume that they will.

I suppose the same might be true to a lesser scale with radar returns - if the streets are flooded with radar signals bouncing around, how will your car be able to determine the correct return for it’s own radars?

Most autonomous car testing right now is done in ones and twos. I’m not sure anyone has ever taken their entire fleet and put them all in an enclosed area, all trying to operate at once.

Might be an interesting experiment to perform.

Also - Volvo, I’m looking at you - make sure your cameras are either inside the swept area of the windshield, or have their own wiper. I had a Volvo lane-keep-assist system freak out on me last fall when a wet maple leaf stuck over a camera lens......