Illustrations by Léonard Dupond
Two years ago, former Uber Chief Executive Officer Travis Kalanick took a trip to Carnegie Mellon University, in Pittsburgh. The reason, he later told Bloomberg Businessweek, was to recruit the best engineers and roboticists to build a whole autonomous driving system capable of eventually replacing the ride-sharing giant’s more than 1 million human drivers and, as he described in a blog post, making transportation safer and more reliable. What’s more, he wanted this done as quickly as possible, even though analysts were insisting that fully functioning self-driving cars could be decades away.
Kalanick is also the co-founder of Uber, a company worth an estimated $62.5 billion, of which hundreds of millions go to paying humans to drive. He sees the business opportunity in leveraging ride sharing as a catalyst for the driverless revolution. And he seems eager to capitalize on it right now.
He and other executives in the ride-sharing space, along with automotive-manufacturing leaders, are betting that transportation as a service is a better launch pad for autonomous vehicles than personal ownership. Creating services centered on intelligent, self-driving cars allows for a smoother, more cost-effective rollout of the technology that fits with current shifts within the traditional auto industry—which is also ripe for disruption.
For Kalanick, turning to Carnegie Mellon for robotics experts was obvious since the university had already trained some of the strongest leaders in the autonomous driving field, including Sebastian Thrun and Chris Urmson, both of whom worked on Google’s self-driving car project that tested autonomous vehicles throughout the San Francisco Bay area. At Carnegie Mellon, Kalanick must have liked what he saw because the following spring, Uber hired 40 researchers and scientists from the university. At the time of his visit, automakers such as Audi, Ford, GM, and Mercedes-Benz had begun testing their own self-driving technologies on highways and byways all over the globe. Tesla had started previewing its Autopilot feature on surface roads within cities and towns, including the high-pressure streets of New York City. But none of those industry players was close to delivering a completely autonomous car, or what’s known as an SAE International standard Level 5 autonomous vehicle.
Still, Kalanick seemed determined to do something that none of the automotive or technology behemoths had achieved: monetize self-driving technology.
Transportation as a Service
It’s easy to think of a self-driving car as a magical machine that pulls up one day and solves all of our transportation woes overnight. But that isn’t going to happen. Instead, the transition from driver to completely driverless will likely occur incrementally. As seen in early pilot programs from Google and Uber, the vehicles will have limited capabilities and travel along set routes for a defined period of time. Eventually they’ll be ready to go anywhere at any time without restrictions.
As such, the next autonomous driving systems to emerge need to achieve SAE International standard Level 4 high automation—vehicles that can operate without falling back on a human driver but can’t quite handle all conditions and every scenario. We’re seeing Level 3 conditional automation vehicles that still have a minder and can only operate in specific geographic areas, sort of like licensed taxi drivers.
“How can we impact mobility for the greatest number of people in the shortest period of time?”
Randy Visintainer Director, Autonomous Vehicles and Controls, Ford
With this in mind, ride-sharing operations such as Uber and Lyft can combine the convenience of not having to drive with the flexibility of a self-driving system that should improve and become more robust over time and whose software will allow vehicle manufacturers to respond quickly when the physical environment changes.
“How can we impact mobility for the greatest number of people in the shortest period of time?” asks Randy Visintainer, Ford’s director for autonomous vehicles and controls. For one, ride sharing eliminates major cost issues. It puts the technology in the hands of the masses as opposed to putting a very expensive system in luxury vehicles only available to few, he explains. “You can start it in an urban environment, and then as the technology evolves and proves out—we’re looking at new technology that gets less dependent on the maps—we can start expanding from those regions,” he says. “It’s going to be a way for people to get comfortable with the technology, and when there’s a better business brief for personal use, the technology will have already been out there.” Consumers can then decide whether they want to invest in it or not.
Autonomous Cars Are Here
Last September, Uber launched its commercial autonomous ride-sharing pilot program in Pittsburgh—rides in an autonomous Uber started out as free, but they won’t be after testing is complete. This was a first for the industry, and it started a new chapter in transportation.
Unlike companies, including Google, Tesla, and Ford, that are talking about launching ride-sharing services using their own self-driving technologies, Uber doesn’t plan to manufacture autonomous cars. Instead, the company aims to build the software that will make the cars work. It is also forging partnerships with automakers such as Volvo to source the vehicles.
Uber used a small fleet of Ford Fusion Hybrid sedans to test its robotic livery cab technology in Pittsburgh last fall. The company is also testing a new fleet of 100 specially modified Volvo XC90 sport-utility vehicles. Going forward, the company says it will work with many different carmakers.
Although Volvo is working on a proprietary self-driving solution, called Drive Me, at its Göteborg headquarters, Uber will not be using it. Instead the company is outfitting each Volvo XC90 with its own unique autonomous system designed to handle acceleration, braking, steering, and point-to-point navigation.
Uber executives are tight-lipped about the inner workings of their self-driving system and have declined to be interviewed in-depth about their autonomous vehicle technology or the Pittsburgh pilot. However, in a post on Uber’s website, Kalanick has indicated that aspects of the company’s system will be sourced from technology created by Otto, a San Francisco self-driving truck startup that Uber purchased last summer for $700 million.
Founded by Google veterans that included self-driving project engineer Anthony Levandowski, Otto has created a self-driving software suite for big rigs. The company successfully tested the technology this past October by making the first ever autonomous truck delivery: a 120-mile Budweiser beer run from Fort Collins, Colorado, to Colorado Springs.
A human driver maneuvered the 18-wheeler, equipped with $30,000 worth of hardware and software, from the brewery onto Interstate 25. Then, in autonomous mode, the truck piloted itself along the highway. The driver was able to rest in the back of the cab while the truck handled the heavy load.
In discussions around self-driving capabilities, trucks represent practicality. They are also desperately needed. The American Trucking Association reported that trucks haul 70% of the nation’s freight—about 10.5 billion tons annually—but the industry is currently experiencing a massive driver shortage that the association predicted could reach nearly 175,000 by 2024.
Otto’s big rig test bodes well for Uber’s acquisition of the company. “Self-driving is core to Uber’s mission: reliable transportation everywhere for everyone,” Levandowski, who is now Uber’s vice president of engineering and the head of its autonomous vehicle initiative, told a group of journalists gathered for the launch of the Pittsburgh pilot, as reported by the San Francisco Chronicle. “For me, it is the most important thing computers are going to do in the next 10 years.”
Uber’s Volvo XC90s are difficult to miss. Each is adorned with Uber’s logo and topped with a roof-top array containing radar and laser tech, as well as GPS receivers, cameras, and other sensors. The system is capable of generating more than a million data points each second, which the car then uses to create a picture of the environment it operates in. All that data also allows Uber to rapidly upgrade the company’s software for mapping and navigation.
In the lead-up to the launch of the autonomous vehicle pilot in Pittsburgh, Uber generated detailed maps of the city that included not just roads, traffic lights, and lanes, but buildings, parking spaces, fire hydrants, even trees and potholes. As the Volvo XC90 moves, it compares what it detects on the road with the preloaded maps to sense and avoid hitting new potential hazards such as bicyclists, pedestrians, or animals.
“Self-driving is core to Uber’s mission: reliable transportation everywhere for everyone.”
Anthony Levandowski VP, Engineering, Uber
Not Quite Driverless Yet
The basics of hailing an Uber remain the same despite the robotics. Customers request a car through the Uber app. They enter the pickup and destination locations, then wait to get paired with the nearest ride.
Since there are a limited number of robotic vehicles, not every passenger who hails an Uber in Pittsburgh gets one. Selection is random. The pilot’s autonomous Ubers are only available between the hours of 7 a.m. and 10 p.m., and only to Pittsburgh residents.
The Uber users who do receive a robotic vehicle in Pittsburgh will notice that the entire experience is centered around a tablet computer facing the backseat passengers. The device tells riders that they’re in an autonomous car and walks them through what will happen. Riders are prompted to confirm the destination, tap the screen when ready, and off they go. The screen is meant to put riders at ease, displaying what the car sees in blue for the road and in red for objects. The visuals help passengers get used to not relying on a human driver—who, it should be noted, is still in the front seat for now just in case.
So, yes, Uber’s autonomous vehicles do have human minders: two vehicle operators who sit up front and make sure everything goes smoothly. Plus, Pennsylvania law requires a human being to be able to take control of the vehicle if the driving system gets confused and doesn’t know what to do.
Consequently, one of the operators sits with his or her hands lightly holding the wheel, ready to take over if needed. That’s a good thing, too, because riders have reported that when the driverless liveries encountered issues such as a double-parked or stalled car, the Uber cars didn’t know what to do and just waited for the vehicle in front to move instead of going around. In such cases, the human driver took control, maneuvered around the unanticipated obstacle, then reactivated the autonomous system. The other operator monitors any incoming data and takes notes on a laptop.
For those who have used advanced driver assist systems such as auto-parking, the human-machine interface will look familiar. If the livery encounters an obstacle and must drop out of autonomous operation, a chime sounds inside the car so that the driver knows to take over. The driver can also do so by simply grabbing the steering wheel, or stepping on the brake or accelerator pedals.
Uber is not releasing data on the efficacy of its self-driving system, but riders have reported that the human operator was in control at least 30% of the time. That means that the self-driving system tended to be in control approximately 70% of the time.
The autonomous vehicles are also limited to specific neighborhoods in Pittsburgh—an approach called “geo-fencing”—in certain geographic areas that have been well mapped. Operating in such a confined area can help foster public trust in the technology faster. Because the vehicle won’t encounter unfamiliar terrain, the result is a smoother and more reliable ride than a customer might expect. Uber’s vehicles are also programmed to stay under 35 miles per hour and obey traffic laws.
The revenue trajectory for popular ride-sharing services such as Lyft and Uber is becoming clearer. Removing the driver from the equation eliminates one of the biggest costs of running a modern livery service. After all, lower operating expenses can help give way to more profit.
Uber’s mission to mitigate the most expensive and riskiest element of its ride-sharing equation—drivers—is spelled out right in its marketing materials: “Self-driving technology will mean fewer accidents, less congestion, and radical improvements to the economics of transportation over time,” a company brochure reads. “These goals are core to Uber’s mission of bringing reliable, affordable transportation to everyone in the world.” But the challenge is that Uber does not own the nonautonomous cars out there now. It contracts its fleets through its driver networks. Therefore, eliminating human drivers would also, in theory, cause Uber to lose access to the vehicles it needs for business. To go completely driverless, Uber first needs to build or buy a fleet of new cars, which would amount to a fundamental shift in the company’s business model. What previously helped the company scale up so quickly was providing the platform—and not having to manage, service, finance, or house the vehicles. The prospect of creating a fully autonomous fleet of cars for the service may prove to be much more difficult—and more expensive—than relying on human car owners.
The idea that SAE Level 5 must be achieved before any of this is even possible is a bone of contention for many. Some automakers like Ford argue that it’s all about when, where, and how you implement these services as to whether Level 4 is a sufficient driverless technology. Ford believes that it’s possible, but doubtful, developers will solve the ability to drive anywhere at any time in any weather before Uber or any other service gets fully autonomous vehicles out on the road full-time. Meanwhile, Uber executives are making moves to transform their business model as quickly as possible by phasing out human drivers.
Still, fleets of shared autonomous vehicles make good economic sense. Research from Morgan Stanley indicated that they could contribute $1.3 trillion in annual savings to the U.S. economy and boost the country’s productivity by $507 billion annually. Ford predicts that over a quarter of trips previously made in private vehicles will move to a shared fleet as the price per mile falls to $1. Analysts have estimated that the cost of operating a shared self-driving taxi fleet could be as low as 35 cents per mile, which is less than a tenth of the average per-mile cost of taking a taxi in the U.S. and roughly half that of owning a car.
But it’s the automakers building cars and developing advanced autonomous technologies for years—like Audi, BMW, Ford, GM, Google, Mercedes-Benz, Tesla, and Toyota—that might have an even better opportunity. Not only could each start their own ride-sharing businesses, but they could supply the driving systems, platforms, and integrated end-to-end fleet management. In order to do this, however, automakers have to be willing to change their traditional business models, too
Between 2000 and 2013, driver’s license ownership among people ages 16 to 24 has dropped from 76% to 71%.
Mobility is a tough concept to grasp, says David Cole, co-founder and chairman of the board for AutoHarvest, a nonprofit working to bring automakers together with entrepreneurs and inventors: “I think everybody is struggling.” Figuring out mobility, Cole explains, is like throwing darts at a board and hoping you hit the cork. No one actually knows which areas of the pie will be profitable and have the best return on investment. You should be prepared to play whatever game is going to be the most profitable, he says, whether that is making cars or transportation as a service or something else entirely.
Ford, a Flex partner, sees great business potential in the transportation services sector. That’s where there’s a big growth opportunity, says Alan Hall, communications manager for Ford Technology, Research, and Innovation. The market for selling new cars is worth $2.3 trillion. But Ford’s research shows that the transportation services market—taxis, buses, ride sharing, trains, and route-planning apps—is worth $5.4 trillion. At the same time, Hall says that trends like rising populations and a growing global middle class point to potentially explosive growth in this segment over the next few decades.
“Transportation services—that’s double the size of the marketplace in which we [OEMs] currently play,” says Hall. And it’s untapped by traditional car builders. Consequently, there is huge potential for an OEM to expand its business. How much so, depends on what services it offers.
Ford, the automaker that put the world on wheels more than a century ago and one of the leaders in adding automatic features to vehicles, believes that the indicators of success are changing. Instead of just tracking car sales, the company is paying attention to the number of miles traveled.
“We’ve always looked at it as the number of cars sold because that’s our business,” says Hall. “Once we sell the vehicle, we’re there to continue servicing it. What you do with the car, that’s your thing. You could put 100 miles on it or 100,000 miles on it. It doesn’t impact us that much.”
Riders using transportation as a service are paying for miles traveled rather than paying to own the car. Car-sharing services also mean more wear and tear on the vehicles since they’re on the road so much, adds Cole. Hence, Ford could make money more than one way in this emerging business model.
Last fall, Ford CEO Mark Fields forecasted that autonomous vehicles will account for up to one in 10 miles traveled by 2025 and 30% of vehicle sales by 2030. With Level 4 autonomous vehicle technology, Ford could potentially reduce the transportation-as-a-service price per mile and see larger profits.
Pittsburgh is the first pilot, but driverless fleets are coming to other areas soon. As this story was going to press, Uber was attempting a driverless car rollout in San Francisco but facing some challenges from state regulators—California’s DMV called Uber’s cars “illegal.” Meanwhile, the parts manufacturing company Delphi Automotive is forming a consortium to introduce a ride-sharing service with self-driving cars in Singapore this year. BMW and Ford each announced that they plan to have fully autonomous vehicles for ride sharing delivered by 2021. General Motors has invested $500 million in plans to deploy its own self-driving vehicles for Lyft’s service.
The convenience of on-demand transportation and the lower hassle of a “no-car” lifestyle is appealing to many of today’s youth. But is the market cap sustainable on a ride-sharing company like Uber? “We don’t know,” says Cole. That said, he adds, one of the beauties of the service business is that the cost to step away is not very great compared to the high cost of walking away from making cars. Even so, who wants to be left out of a low-cost, high-margin enterprise?