In the race to build autonomous vehicles, General Motors is determined not to be left behind.
Last year, it bought Cruise Automation, a startup based here, to jumpstart its self-driving car efforts. On Tuesday, GM demonstrated what it and Cruise have been working on, giving a select group of reporters a test drive in one of their autonomous vehicles.
“This is a big moment for me, personally, and for General Motors and for Cruise,” Kyle Vogt, Cruise’s CEO, said.
GM plans to mass produce its self-driving cars in a matter of “quarters, not years,” company president Dan Ammann said. He declined to specify when, exactly, the company plans to start selling them.
“Stay tuned,” he said.
In the meantime, Ammann, Vogt, and their colleagues were eager to show off their self-driving cars. Although GM and Cruise have been internally testing prototype vehicles for more than a year, this was the first time the companies had given members of the media or the general public a ride in them.
GM’s self-driving vehicles are basically modified versions of its Bolt electric vehicles. GM has added some 40 sensors to the vehicles, include an assortment of radar and lidar devices, as well as hundreds of pounds and several miles worth of cables.
The automotive giant has designed the vehicles to be mass produced in its Orion, Michigan, factory. It’s already built 180 test vehicles.
GM and Cruise rolled out their first generation test vehicles only in May 2016. The version of the car GM and Cruise demonstrated on Tuesday is their second-generation test vehicle, which debuted this past June and includes a complete collection of sensors. They already have a third generation prototype in production; that vehicle, unlike its predecessor, includes redundant systems.
GM’s autonomous car offered a safe — but not smooth — ride
My ride got off to an inauspicious start. For the event, Cruise handed out iPhones that media members could use to summon one of their cars using the company’s app. That app, which works similarly to Uber or Lyft’s ride-haling apps, is actually the same one Cruise employees use to catch a ride with one of its cars. Only in our case, Cruise allowed us to choose from one of several pre-set destinations, rather than setting one on our own.
It took several minutes for my car to arrive in front of the Dogpatch Studio here, where GM was holding the event. But instead of stopping for me, the car drove right on by and returned back to the place it had been parked at before.
So, my Cruise helper summoned another car. That car too passed us by. But after traveling around several blocks, it came back to pick me up.
The car then proceeded to take us on a drive via city streets through the Portrero Hill neighborhood in the eastern part of this city. For about 20 minutes, we went up and down the area’s steep hills before returning to the studio.
Riding in the Cruise vehicle was a little more unnerving than my recent ride in one of Waymo’s latest self-driving cars. Waymo’s demonstration was held at its testing facility in Central California, and even though its car had no driver and had to contend with people driving, walking, and biking, the ride felt staged and artificial.
In this demonstration, a Cruise employee sat behind the steering wheel, ready to take over if anything happened and another sat in the front passenger seat calling out potential obstacles and alerting the driver to the car’s upcoming path. Even so, this was a real-world test and felt like felt like it.
San Francisco is a notoriously difficult city to drive in, thanks to its steep, narrow streets, frequently heavy traffic, and often brazenly incautious bicyclists and pedestrians. Cruise’s car had to navigate around multiple double-parked trucks while contending with oncoming traffic, avoid pedestrians randomly crossing the street, be on the lookout for cars and bicycles as it crested hills, and negotiate four-way intersections where it sometimes encountered cars with impatient drivers.
The car made it through the course without incident — we never came close to an accident, and the Cruise employee behind the wheel never had to take manual control of the car. And that was impressive, far more so than Waymo’s car navigating through a staged course.
But Cruise clearly has some things to work on. The ride, while safe, was anything but smooth. In fact, it often felt herky-jerky. We would accelerate as we turned a corner, slow down abruptly right after that, speed up soon after until we came close to the next intersection, then brake fairly suddenly at a stop sign.
The car sometimes seemed hesitant
And the car seemed overly cautious, even hesitant at times. In navigating around one double-parked truck, it slowly nudged out to peer around it, then slowly pulled around it in the oncoming lane. In the same situation, a human driver likely would have gone much more quickly in order to avoid any potential oncoming traffic.
Early on in my ride, the vehicle had to negotiate a particularly challenging stretch. In front of us was a double-parked truck. Just beyond that, on the other side of the street was another double-parked truck. In between and among the trucks were construction workers. And there were cars both behind us waiting for us to go and in front of us, heading the opposite direction.
The Cruise vehicle moved slowly and prudently into the oncoming lane. It nudged out just a bit to better see the traffic heading toward us. When things were clear, it pulled out. But then, while pointed in the wrong direction in the opposite lane, it just stopped. After waiting for two cars to go around the truck on their side and pass us going the other way, it finally went around the truck on our side and proceeded on.
Of course, a human driver might also have struggled with that scenario. But a human driver, even a cautious one, would likely have navigated around the obstacles more smoothly.
Right now, Cruise is far more focused on making its cars safe than worrying about how smoothly they drive, Cruise’s Vogt said. Not only is safety crucial for autonomous vehicles, but it’s a far harder problem to solve, he said.
Vogt expects Cruise to spend more time improving smoothness and polishing other aspects of the rider experience next year as the company nails down the safe driving aspect of the cars.
Making the cars drive more smoothly “Is the easiest part of the process,” he said.
I could see more than the car’s screen showed
Like Waymo’s autonomous cars, Cruise vehicles have screens on the backs of the driver and passenger seats that show in graphical form what the car “sees” around it. Other cars on the road show up as grayish blue boxes, people as blue cylinders.
Generally, whenever cars or people were close by, they showed up on the screen. You could see rows of boxes representing parked cars and cylinders crossing the street in tandem with the people they represented.
But sometimes the screen didn’t seem to register objects immediately. When we came to an intersection at the top of one hill, it took a while for the screen to show a car parked on the opposite corner, long after I saw it out the window.
More disconcertingly, sometimes the screen didn’t display some objects at all. As we approached one intersection, the screen registered the car coming up the street on our left. But it never showed the car coming up to the intersection from our left. Similarly, the pedestrians walking on the sidewalk that we passed never showed up on the screen, even though they were walking only a few feet away from the parked cars that were displayed.
Cruise’s cars can actually see a lot more than what they show on their screens, Vogt said. The company decided to limit what they display, because it found that when it showed more objects, riders found all the information “overwhelming,” he said.
GM is focusing first on the ride-hailing market
GM sees ride-hailing services as the initial market for these vehicles, Ammann said. The automotive giant is still working out the details of how it will approach that market. GM has invested some $500 million in Lyft, but Ammann said it wouldn’t necessarily partner with that company in all cases to offer ride-hailing services.
In some cases, GM might offer automous ride-hailing services through partnerships; in others, it may offer them by itself, he said. Either way, GM plans to maintain control of the fleet of cars it deploys, he said.
The company’s focus on the ride-hailing market has played into the testing of its autonomous vehicles, Ammann said. It and Cruise are focusing on testing their cars in dense, busy cities in part because those are the biggest markets for ride-hailing services. The companies have already been testing their cars in San Francisco and plan to begin testing them in Manhattan in the first quarter.
GM is one of several major companies developing autonomous vehicles. Among the others are Uber and Google spinoff Waymo. Like GM, those companies see the ride-hailing industry as among the most promising early markets for such cars.
Cruise and GM are concentrating on cities
However, Cruise’s Vogt took pains to differentiate his company’s development effort from that of Waymo. Waymo has undertaken a years-long effort to create autonomous cars and has done much of its real-world testing either in suburban environments such as Mountain View, California, and Chandler, Arizona, or at its testing facility at a decommissioned Air Force base in California’s Central Valley. Collectively, its cars have driven some 4 million miles autonomously, the company announced on Monday.
By contrast, Cruise has gotten to the point it’s at in just 18 months, Vogt noted. And while it also tests its cars in suburban Phoenix, much of its real-world testing is being done in urban environments.
Vogt and other GM officials declined to say how many miles its cars have driven, but they argued that the city miles they’ve driven count for much more than the miles its competitors have logged in the suburbs. That’s because incidents such as having to make a left-hand turn or having to pull into the oncoming lane of traffic to route around a blocked lane occur much more frequently in cities than in the suburbs. In some cases such incidents occur more than 40 times more often in cities than in suburban areas, he said.
That difference “really puts into perspective … the value and challenge in operating in these urban environments,” Vogt said.