With 691 horsepower, my black Tesla Model S could easily accelerate from zero to 60 miles per hour in less than three seconds. Lacking a combustion engine, the all-electric car is noiseless and nonpolluting. But on this test drive along the highway that parallels the Hudson River just north of New York City, Im not concerned with speed or eco-friendliness. Its the glimpse of a driverless future that really excites me.
Tapping twice on a left-side lever, I switch on the autopilot, place my hands on my lap and let the Tesla drive itself. The vehicle stays in its lane and takes the curves smoothly. When cars ahead slow down, so does the Tesla. I briefly take back the steering wheel to exit at a Yonkers shopping center and head to the parking lot. Another tap on the controls and the vehicle begins to park, interrupting the maneuver when its sensors detect a mother and her toddler walking too close. Once they pass, the Tesla finishes parking itself, perfectly equidistant between two cars.
The day when most new vehicles will be driverless is still at least a couple of decades away, according to industry experts and research studies. But the first commercial, fully autonomous driving, or AD, cars will debut by 2020, probably along designated corridors in urban communities and on commuter routes.
Semiautonomous and AD cars (see the Autonomous Vehicles index) are the latest chapter in the Internet-era disruptions that have hammered brick-and-mortar giants in retailing, publishing and broadcasting. But no industry offers a potentially richer prize, or one with greater social and scientific impact, than motor vehicles. The auto industry, parts and other aftermarket products and services account for a global annual market worth $3.5 trillion, according to a January report by McKinsey & Co. Autonomous driving is likely to be the most disruptive event we have seen in the marketplace ever, says Joseph Fath, who focuses on disruptive technologies as portfolio manager of two T. Rowe Price funds the Growth Stock Fund and the U.S. Growth Stock Strategy with more than $67 billion in assets under management.
Driving is etched deeply into the American psyche as a coming-of-age passage and a quick escape from daily drudgery. In emerging markets car ownership is the ultimate symbol of middle-class status. But will new generations be interested in driving their own cars, or will they prefer AD vehicles summoned on smartphones without the hassle of parking? In my lifetime it will be considered strange that human beings were once allowed to drive cars, asserts Gavin Baker, 40, portfolio manager of the Fidelity OTC Portfolio, whose $12.4 billion in assets under management include companies involved in the development of AD vehicles.
Meanwhile, traditional carmakers face the same kind of quandaries roiling incumbents in other industries under threat by disruptive technologies: How much and how fast should they embrace autonomous driving? Does it make sense to reduce their focus on very profitable traditional models, such as sport utility vehicles and pickup trucks, and instead funnel more capital, engineering and management resources into AD? Is it possible to thrive in a future where many fewer vehicles autonomously driven and operating longer hours than conventional cars and trucks will be needed to shuttle around people and goods? And because human error is responsible for 90 percent of accidents, wont pressure grow on traditional carmakers to accelerate the development of driverless cars that are expected to sharply cut down on crashes?
For now, motor vehicle companies have sought to straddle both the old proved and brave new worlds. The high-tech devices that will control AD cars sensors, high-definition maps and artificial-intelligence software are becoming standard features in new, safer models put out by traditional automakers like Fiat Chrysler Automobiles, Ford Motor Co., General Motors Co., Renault-Nissan Alliance, Toyota Motor Corp. and Volkswagen, as well as by upstarts like Tesla Motors. These active systems are already cutting down accidents in our cars by 20 to 30 percent, says Marcus Rothoff, director of Volvos autonomous driving program at the carmakers headquarters in Gothenburg, Sweden.
Autonomous driving is also drawing massive investments of time and money from Internet-era disrupters such as Alphabets Google, iPhone maker Apple and Chinese web services company Baidu, either as possible vehicle manufacturers or software partners of existing carmakers. And the likelihood that driverless fleets will someday dominate car-hailing and car-sharing services in urban areas has helped send valuations soaring for privately held Uber Technologies and Lyft.
But the biggest beneficiaries of the capital pouring into semiautonomous and fully autonomous driving over the next few years are likely to be parts suppliers like the U.K.s Delphi Automotive, Israels Mobileye and Germanys Continental. We are seeing the payoff now in semiautonomous and AD in terms of revenue growth, says Jeffrey Owens, Delphis Troy, Michiganbased chief technology officer. The industry views it as the hottest, fastest-growing space in the automotive business, bar none.
Less clear is the impact on the $700 billion-a-year global auto insurance industry (see Auto Insurers Hope to Avoid Crashing in a Driverless Future). A 2015 research study by accounting giant KPMG predicts Armageddon for auto insurers 25 years hence as human error is virtually eliminated from driving and accident rates plummet: Within 25 years, our models suggest a scenario where the personal auto insurance sector could shrink to 40 percent of current size. But insurance experts insist that, at least in the near future, there will be higher profits for the leading insurers as underwriting results rise and claims fall. And there is plenty of time, they add, for the industry to adjust and create products to cover new contingencies, such as cyberterrorism and manufacturers liability.
The impact of AD is being felt far beyond the streets and highways. So complex are the scientific problems posed by the vehicles avoiding collisions at high speeds, knowing every constantly changing inch of road in the world, recognizing instantly any object or living being in a cars path that their solutions will provide quantum leaps forward for artificial intelligence. The AI you develop for cars can migrate to other areas of technology, says Amnon Shashua, co-founder and CTO at Jerusalem-based Mobileye. This is why so many companies, not just carmakers and their suppliers, are interested in having a role in developing autonomous cars.
The ad movements Woodstock took place in Californias Mojave Desert in 2004. On March 13 of that year, the U.S. Defense Department organized the so-called DARPA (Defense Advanced Research Projects Agency) Grand Challenge, the first long-distance world competition for driverless vehicles.
The aim was to spur the technology needed to develop autonomous ground transportation for U.S. military forces. A $1 million prize drew more than 100 teams, including participants or sponsors from Ford, GM and other global automakers, as well as entrants from leading universities, such as Stanford and Carnegie Mellon.
No vehicle finished the 150-mile route. In fact, none got farther than 7.3 miles before crashing or sputtering off the desert highway into sands and gullies. But with futility came the epiphany that three major ingredients were essential for success in AD vehicle development: sensors, artificial intelligence and high-definition maps.
The primary sensors are cameras with computer vision that can see where highway and avenue lanes are painted, whether traffic lights are green or red, where roads curve and curbs begin, and the locations of other vehicles, people and animals. Redundancy is required. Besides multiple cameras, every vehicle must be equipped with radar and lasers that can overcome camera vision impairment caused by heavy rain, snow, fog or reflected sunlight. You need constant 360-degree awareness, says Mobileyes Shashua.
Artificial-intelligence programs are necessary to interpret the enormous data detected by the sensors and answer basic questions: When should the vehicle hit the brakes? When should it merge into traffic at a roundabout? When can it overtake another vehicle? The map for AD vehicles has to be unlike any other in terms of sophistication and data. This is far beyond GPS, says Rod Lache, a New Yorkbased automotive analyst for Deutsche Bank Securities. It involves a very detailed 3-D scan and analysis of what the vehicle is driving past every bridge and tree and fire hydrant. Complicating such map-making are changes in the terrain as routes and roadside structures come under construction and repair. And then there is the daunting geographic scope: more than 4 million miles of roads in the U.S. alone.
No company has tackled all three elements of the AD conundrum more ambitiously than Google. Beginning in 2009 the company launched a self-driving car project aimed at accruing 100,000 miles of autonomous driving on public roads over 18 months. That mileage and data have been vastly expanded through the use of Google vehicles. Some are modified versions of Toyota SUVs. Others look like cartoonish versions of the VW Beetle, and all carry a cluster of cameras and sensors on their roofs. In May, Google took another leap forward by announcing it will install its technology in a fleet of 100 Chrysler Pacifica minivans to begin adapting mass-market vehicles to fully autonomous driving.
Currently, Google performs more than 3 million miles of testing simulation every day. We have teams of people whose job it is to make the cars life difficult, said Chris Urmson, director of Googles self-driving-car project, during a March presentation at the South by Southwest Interactive Festival in Austin, Texas, where numerous companies showed off new digital technologies. They come up with all sorts of crazy things that you may or may never see in a lifetime of driving.
Sometimes reality is beyond the imagination even of the research teams. Urmson played a video of an incident 18 months before, when a Google AD car abruptly stopped on the main street of Mountain View, California, to avoid hitting an elderly woman in a battery-powered wheelchair as she pirouetted across traffic lanes while wielding a broom to herd several ducks back to her property.
Another, less humorous incident, on Valentines Day of this year, received widespread media coverage: A Google vehicle near the companys Mountain View headquarters hit the side of a bus it had erroneously expected would brake to allow the driverless car to pull into a slow-moving lane of traffic. There were no casualties, and the only damage was to a few sensors knocked off the Google car. But the episode dented some of the optimism about driverless vehicles.
Jerry Kaplan, a Stanford, Californiabased technology entrepreneur and author of Humans Need Not Apply: A Guide to Wealth and Work in the Age of Artificial Intelligence, doesnt doubt that Google and other companies involved in AD development will eventually resolve the technological obstacles. The real impediments are going to be social, he says. How does a cop stop an autonomous vehicle? Will a self-driven car observe parking time limits? How does it trade information with other people in an accident?
These are some of the reasons Urmson hedges when asked how soon self-driving cars will become widely available. If you read the papers, it will take maybe three years, maybe 30 years, he said at the Austin conference. Im here to tell you honestly its a bit of both.
No such ambiguity clutters the thinking of Tesla founder and CEO Elon Musk. He predicts that fully AD vehicles will be on the roads by 2020 if federal, state and municipal regulators give the green light.
The South Africanborn, Canadian-American inventor and entrepreneur, now 44, launched Tesla in 2003 at a Fremont, California, factory that hed purchased from GM and Toyota for about $41 million. Five years later he introduced the Roadster, the worlds first electric sports car. It was followed by the Model S luxury sedan and the Model X SUV, both also fully electric. But Teslas $80,000 to $130,000 price tags mean that only 120,000 of them are on the roads worldwide and have made the company seem like a fringe carmaker for the affluent.
With the marketing this year of the $35,000 Model 3, an electric vehicle, or EV, with a 215-mile range between charges, Tesla gained instant mass appeal and industrywide credibility. The car wont be available until late 2017, but in the first week after its March 31 unveiling, Musks company took in 325,000 orders totaling $14 billion. Thats a world record for any industry in any country for any product, notes Trip Chowdhry, an analyst at Redwood Shores, Californiabased Global Equities Research. By the first week of May, some 400,000 people had each made the required $1,000 payment to reserve a Model 3.
Meanwhile, anticipation of the Model 3 drove Teslas market capitalization to $28 billion in mid-May. Thats not far behind Fords $53.6 billion and GMs $48 billion, though its only a fraction of global leader Toyotas $152.5 billion valuation. But those traditional carmakers produce millions of vehicles annually.
The hoopla surrounding the novelty of a battery-powered car has obscured Teslas emergence as the clear leader in the development of semiautonomous and fully AD vehicles. At T. Rowe Price, one of Teslas largest shareholders, portfolio manager Fath initially invested in the company as an EV maker but soon grasped its even greater attraction as a pioneer in automation and new safety devices. When Tesla saw the opportunity and realized it had the technological capability, it moved rapidly, he says.
Tesla has convinced many investors, analysts and consumers that an all-electric vehicle is the best platform for autonomous driving because it depends on electronic braking and steering, and high-precision drive control. All that investment is already done in an electric vehicle, says Jeff Evanson, Teslas head of investor relations.
Thats a notion hotly contested by traditional carmakers. They point out that all those functions, plus the driving itself, place too many demands on a battery-powered, all-electric vehicle and limit its travel radius between charges. Hybrids make more sense, says Raj Nair, chief technical officer at Ford, the No. 2 seller of battery and combustion-engine hybrids in the U.S., after Nissan. With a hybrid you get the battery capability and also the range.
Critics fault Tesla for its vertical integration business model. The company not only assembles its cars but also manufactures the chassis, batteries, engines, most parts and much of the software. It is building a global network of more than 3,500 roadside stations to recharge for free the batteries of Tesla vehicles that are on long-distance trips or werent recharged overnight in their garages. And it has opened a so-called gigafactory near Reno, Nevada, to mass-produce its own lithium-ion batteries. This would be the equivalent of a traditional combustion-engine car company with its own oil fields, refineries and fuel pumps.
Tesla defends its in-house approach by pleading an inability to find reliable suppliers. Many of the parts and products we needed didnt exist we had to invent them, says Evanson, pointing as an example to the Model Xs doors, which open and fold like falcon wings.
Thus far in the companys 13-year existence, Teslas capital spending has totaled only about $4 billion, less than a third of VWs annual expenditures. Its a common misperception that we have a capital-intensive business model, Evanson says. But it is managerially distracting.
Both management problems and capital expenditures are bound to grow as demand for Teslas soars and the company strains to meet production orders for the Model 3. The company recently lost three top executives. In April, Chris Porritt, vice president for vehicle engineering, left to join Apples secretive project on AD electric cars. In May, Greg Reichow, vice president of production, and Josh Ensign, vice president of manufacturing, also announced they were leaving Tesla, without indicating their future plans.
But at the first-quarter conference call with analysts, on May 4, CEO Musk insisted that despite the executive losses Tesla would produce an annual total of 500,000 vehicles by 2018, two years earlier than he had previously announced. I have a sleeping bag and a conference room adjacent to the production line, which I use quite frequently, he said. The whole team is superfocused.
Whether or not it meets that output forecast, the company seems certain to expand its lead in semiautonomous and AD driving devices. Tesla boosters liken the cars to iPhones in the way the vehicles react to users needs. Software updates are beamed over the air to Tesla cars once or twice a month, just as Apple does with its wireless devices.
The feature received widespread attention in 2013, when two Teslas caught fire as a result of road debris damaging the lithium batteries that cover the undercarriage; the company sent overnight software updates that raised the chassis and increased ground clearance on all Teslas by a few inches. No more on-the-road battery fires were reported. In a more permanent solution, approved by the National Highway Traffic Safety Administration (NHTSA) in 2014, Tesla installed a titanium undercarriage shield and aluminum deflector plates on new vehicles.
But high-tech wizardry hasnt made Tesla immune to lower-tech headaches. On April 11 it announced recalls of 2,700 Model X SUVs after testing revealed that the third-row seats could collapse in crashes. And last year all 90,000 Model S vehicles had to be recalled to check for possible safety belt defects.
Still, Teslas ability to instantly offer over-the-air improvements keeps it far ahead of the rest of the field. Such software updates activate preinstalled sensors and advanced driver assistance technology as soon as they receive approval from the NHTSA and other regulators.
Last October, for example, regulators endorsed the autopilot feature that allows Teslas to drive themselves on highways, and the vehicles immediately received over-the-air activation. None of the traditional car companies can do this because they dont have the capability yet, T. Rowe Prices Fath says. Instead, owners of traditional cars must await the next model for a built-in set of new semiautonomous and AD features.
But futurist research suggests that even Tesla will find it increasingly difficult to adapt to an AD world. According to a December 2015 report by Barclays, by 2040 half of all cars in the U.S. will be autonomously driven and in most cases will be shared rather than individually owned. Each shared vehicle could replace as many as seven traditional cars. But its hard to shake the good cheer currently sweeping the U.S. car industry.
Detroit is back could easily be the slogan of traditional carmakers, both domestic and foreign, in the U.S. market. The bankruptcies and bailouts of the 200809 financial crisis are receding in the rearview mirror. With auto sales in the U.S. reaching a record 17.5 million units in 2015, there is no industrywide sense of panic over falling behind in AD development.
In fact, this springs New York International Auto Show was a celebration of traditional combustion-engine vehicles. More than 1.1 million people waded into the glass-sheathed Jacob Javits Convention Center, which sprawls over six square blocks where midtown Manhattan abuts the Hudson River. These enthusiasts werent interested in Google AD cars or experimental driverless shared vehicles.
Among the more than 1,000 vehicles on display, sport utility vehicles were the star attractions. In 2015, SUVs, pickups and crossovers accounted for more than three quarters of GMs $9.7 billion in net profits and Fords $7.4 billion. With U.S. gasoline prices in the neighborhood of $2.00 a gallon, fuel is no longer a drag on sales of big guzzlers. But cheap gas doesnt explain the adoption of SUVs by superluxury exotics like Maserati and Lamborghini. At the auto show families leaned over a waist-high metal-and-rope barrier to gawk at Bentleys Bentayga (starting price $229,100).
Vehicles with autonomous capabilities were barely visible at the Javits Center, reflecting the bias of the old-fashioned driving enthusiasts in the crowds. According to the auto show catalogue, in a worldwide survey of 10,000 people taken by Volvo, 88 percent of respondents thought AD vehicles should respect the love of driving. The most prominent quote in the catalogue under the headline Turned-Off by Driverless Cars? was from Porsche CEO Oliver Blume, who said his company had no plans to develop AD vehicles. One wants to drive a Porsche by oneself, he explained.
Outside the auto shows thats an increasingly rarefied view among traditional carmakers. They envision a continued focus on combustion-engine cars that incrementally incorporate new safety and driver assistance technology, but they accept the inevitability that AD vehicles will claim a growing portion of the market. Its not going to be the whole fleet changing over to AD vehicles, says Ford CTO Nair. There will be a need for other vehicles as well. So we anticipate they will coexist.
Nonetheless, Ford is trying hard to spruce up its Rust Belt image. In 2015 it opened a research center in Palo Alto, California, to stake out a presence in Silicon Valley. It is assembling a small fleet of experimental AD cars while loading up new mainstream models with mostly in-house technology. In March the company launched a separate Ford Smart Mobility division to more efficiently pursue semiautonomous and fully autonomous driving projects. And in April it announced plans for a ten-year transformation of its Dearborn, Michigan, headquarters and research center at an undisclosed price into a campus setting similar to the Silicon Valley headquarters of Google and Apple.
Rumors abound among automotive analysts that Ford may seek a joint venture with Google to spur AD vehicle development. In late April both companies agreed to lead a coalition of auto-related businesses to press for more-rapid federal approval of AD vehicles. We have never ruled out a partnership that makes sense, says Nair.
General Motors is moving into the AD arena even faster than Ford. In January, GM announced a $500 million investment in Lyft, a San Franciscobased ride-hailing start-up. That capital injection valued Lyft at $5.5 billion, a fraction of market leader Ubers recent $62.5 billion valuation. Then in February, GM spent more than $1 billion in stock and cash to acquire Cruise Automation, a small Silicon Valley AD technology start-up that had been valued at less than $100 million in a venture capitalraising round just six months before. By investing in Lyft, GM is saying it believes in a future of driverless fleets, says Egil Juliussen, Minnetonka, Minnesotabased director of research at consulting firm IHS Automotive. Time will tell whether GM paid too much for Cruise. But they essentially got Silicon Valleytype software expertise and quickness to develop new products.
The Lyft investment has garnered the most attention. The alliance will allow customers whether professional chauffeurs or ordinary drivers to rent vehicles from GM through the Lyft app. GM will continue to own the cars. But now, instead of being a one-time sale to a rental agency, GM will receive ongoing revenue, says David Rust, director of operations strategy at Lyft.
The program was launched in Chicago on March 28 and will be extended to other major urban markets, including Boston and Washington, by the end of this year. In 2017, GM and Lyft plan to test a fleet of driverless taxis, using Chevrolet Bolt electric cars. Once AD technology and regulatory approvals are in place, perhaps before 2020, drivers will cede their jobs to driverless vehicles that will be allowed on specified routes. Eventually, we might see 50-mile commutes with shared autonomous vehicles for people who now drive every day from, say, San Jose to San Francisco, Rust says.
Theoretically, GMs investment in an AD-vehicle-sharing future implies its acceptance of lower car sales. In fact, its part of the same market segmentation envisioned by rivals like Ford, Toyota and VW.
AD networks will be most prevalent in cities like New York and Los Angeles, where GMs profits from compacts and sedans are low compared with its earnings from larger vehicles in rural and suburban zones. If I manage a car company that is making outsize profits in SUVs and pickup trucks but low margins in other vehicles in urban areas, then I view driverless mobility networks as a chance to start again in cities, says Itay Michaeli, a New Yorkbased automotive industry analyst at Citigroup.
What if a traditional carmaker isnt large enough to make the high-tech investments and cross-industry deals of a GM? Volvo, for example, sold only 503,000 units last year, compared with GMs 9.8 million. Yet size has never dissuaded the Swedish company from being on the cutting edge of safety and other automotive technology.
Over the past four decades, Volvo has kept a forensics team on round-the-clock watch at its headquarters, on Swedens west coast. They race to the scene of any Volvo accident within 100 kilometers of Gothenburg to gauge the damage and injuries, and then figure out how to reduce casualties and property losses for future models.
In 2010 the company announced that by 2020 it would be producing cars loaded with enough sensors, automatic braking and other driver assistance systems that nobody would be killed or injured in a new Volvo driven in Sweden. Thats still our aim, and were getting closer to it, says Rothoff, director of Volvos AD program, known as Drive Me.
Also by 2020, a limited number of AD Volvos will accommodate Swedish commuters on some city and suburban routes, and on designated highways that have received expensive upgrades. But expansion abroad will have to wait for regulatory approvals and large-scale infrastructure investments.
Whether they are behemoths like GM or smaller fry like Volvo, all traditional carmakers suffer the legacy burdens of being incumbents in an age of high-tech disrupters like Tesla and Google. Traditional automaking is a capital-intensive business, says Mark Wakefield, Detroit-based head of automotive affairs at consulting firm AlixPartners. It takes a lot of cash to feed the continuous cycle of models, and normal car development budgets run into the billions of dollars.
Traditional carmakers are under much greater pressure from investors than high-tech upstarts are. Tesla showed $889 million in net losses on $4 billion in revenue last year, compared with a $294 million net loss on $3.2 billion in revenue in 2014. Shareholders view Tesla as a growth stock and are willing to forgo returns on investment for years, but their criteria for traditional carmakers are profits and dividends every quarter. That further reduces the capital left over for new AD investments.
The dealership network is another huge legacy issue. Manufacturers are often forced to guess what customers want in terms of vehicle features and colors. Unsold cars sit at dealers lots for months before being sold at steep discounts. According to automotive news service WardsAuto, U.S. dealers currently have 3.8 million new cars in inventory, the most in a decade.
Tesla, however, is still small enough to customize cars to client specifications. Despite complaints from dealership lobbying groups, it refuses to use dealers and instead sells directly to consumers.
The autonomous driving horizon is rosier and less complicated for auto-parts suppliers than for traditional dealers and carmakers. For an investor the most compelling companies and clearest winners are the suppliers, says Deutsche Banks Lache.
McKinsey predicts that by 2030 semiautonomous vehicles will account for 35 percent of all new cars and fully AD vehicles will take an additional 15 percent of the market. At the very least, these vehicles will have autopilots to keep them in highway lanes and at safe distances from the cars ahead. They will also have a combination of cameras, radar and lasers to warn the driver that the car is converging on another vehicle or person; if the driver doesnt react immediately, the car will slam on the brakes.
Regulators and consumers are spurring automakers to load up their new vehicles with the latest semiautonomous and AD equipment. Every year the NHTSA rates new car models on safety, with the highest four- and five-star ratings for vehicles that adopt autonomous emergency braking, lane-change warnings and the like. Last year 96 percent of the cars purchased in the U.S. had four- or five-star ratings. When consumers are aware of the technology, they want it, says Delphi CTO Owens.
Most of this technology will be developed and manufactured by the suppliers. That could mean a $120 billion annual market for suppliers that doesnt exist today, Lache says.
Meanwhile, a mergers and acquisitions frenzy has overtaken auto-parts suppliers. According to a PricewaterhouseCoopers report in January, M&A activity by suppliers topped $48 billion last year, a whopping 340 percent increase over 2014 and well above the previous record high of $35 billion in 2007. And when suppliers arent acquiring one another, they are often partnering to meet the semiautomatic and AD demands of carmakers.
Examples of this symbiosis are joint projects undertaken by Delphi, which has supplied parts to traditional automakers for a generation, and Mobileye, a technology company focused on machine-learning equipment, along with algorithms to interpret data and computer systems for semiautonomous and fully AD vehicles. Delphi is by far the larger company, with $15.2 billion in revenue and net profits of $1.5 billion last year and a current market cap of $18.8 billion, compared with Mobileyes $241 million in revenue, net profits of $68.5 million and $8.1 billion market cap. The two suppliers have collaborated on the sensors and automatic braking used by Tesla and Volvo as well as other carmakers.
Typically, technology companies are very good at whatever technology they focus on, but the ability to transfer that to the automotive environment isnt part of their repertoire, says Delphis Owens. So we try to be the most effective bridge between the tech sector and the automotive space.
Cooperation with Delphi hasnt precluded Mobileye from alliances with other companies in pursuit of the Holy Grail of automated driving: the development of a high-definition map that will guide AD vehicles over every road on the planet. This shifting map must keep the driverless car aware of every highway curve and lane, every traffic light and sign, every road repair and construction activity for at least a half mile ahead. What makes it even trickier is the fact that you need to maintain the vehicles location inside this map with an accuracy of ten centimeters [four inches], says Mobileyes Shashua.
Among a half dozen serious competitors, Googles mapping effort has received the most attention and takes the most extreme approach. The Google map constantly measures the shifting three-dimensional environment around the moving vehicle. The company surveys the same route twice to eliminate moving objects from the map. This map must be repeatedly updated on a nationwide scale, across those 4 million miles of U.S. roads.
The amount of data is enormous about 1.6 gigabytes per mile, or for every minute or two putting a strain on the computing capacity and battery of any driverless vehicle. (By comparison, the average iPhone subscriber uses 1.6 gigabytes a month.)
Mobileye opted for a less detailed high-definition map, known as the Road Experience Management (REM) map, with a more limited use of 3-D measurements. The amount of data about 16 kilobytes per mile is far more manageable for an AD vehicles sensors and computers. It still enables a continuous real-time update of the map, Shashua explains.
The company scored a huge victory in January and February by signing up GM, VW and Renault-Nissan, which together account for one third of global car production, for its REM maps. Beginning late this year GM will mount REM sensors and artificial-intelligence equipment on a million of its new cars to collect data for the map. By 2018, VW and Renault-Nissan will do the same for large, unspecified numbers of their new vehicles. The plan is to bring all other carmakers into the fold to contribute to this map, which would then be made available to everyone, Shashua says.
Once such maps gain widespread use, the last major technological hurdle will be overcome and the first commercial AD vehicles will become available. A global expansion of driverless cars is sure to quickly follow. Its happening much more rapidly than was initially thought possible, says T. Rowe Prices Fath.
The AD tsunami will be felt first in cities, with pooled driverless vehicles carrying one to three passengers, summoned Uber-style on smartphones or other portable wireless devices. Larger driverless vans will transport commuters dozens of miles between home and work. In suburbs the multicar household will become obsolete: A single AD vehicle will drop the kids at school, pick them up in the afternoon and, in between, transport their parents to work or on errands. And in rural, less inhabited communities, there will be dwindling numbers of human-driven, mostly electric automobiles and utility vehicles.
At a Tesla earnings result conference in November, CEO Musk predicted that in 20 years possessing a nonautonomous car will be as rare as keeping a horse. You will only own it for sentimental reasons, he said.
He might be off the mark. But probably not by more than a decade or so.