Blink and you might miss the moment automated vehicles go mainstream. At some point in 2017, a fully autonomous Tesla will blast across the country en route from Los Angeles to New York. The person sitting in the front left seat — let’s no longer call her the driver — will be free to watch a movie, drink a latte, or wave to locals as she zips past. If Elon Musk has his way, the tech will then roll out to drivers in 2018.
That timeline might be optimistic, but at least one thing is clear: 2017 is the year in which self-driving cars of numerous makes will embark on ambitious rounds of testing. Both Ford and BMW will send their autonomous cars through their paces in Europe. VW just launched its smart mobility brand Moia, with a focus on self-driving shuttles, and Volvo is pushing ahead with plans to give 100 self-driving XC90 crossovers to consumers in Gothenburg, Sweden.
Not to be outdone, the big technology firms are also sending their cars out into the world. Baidu has started offering public rides along a two-mile stretch of road in China, while Uber will likely expand its passenger trials beyond Pittsburgh and San Francisco. Lyft and even Apple are likely to make self-driving news in the coming year, as well.
This rush to deployment has caught some industry watchers by surprise. For years, a single company — Google — had hogged the headlines for tinkering with robotic vehicles and quietly wooing regulators. Then, in 2015, a horde of startups and newcomers burst onto the scene, eager to explore the boundaries of self-driving technologies—and the rules surrounding them.
“2017 will show us that limited deployments are technically, legally, and socially possible, even under today’s laws,” says Bryant Walker Smith, a professor at the University of South Carolina.
The cars themselves, with their distinctive sensor rigs mounted onto their rooftops, will be most visible as they putter down our roads. Behind the scenes, their makers are hurrying to iron out the technical, regulatory, and economic details needed so that one day somewhat soon, most of us will get to take our stupendously fallible meat gloves off the wheel.
From the tech side, this reality seems imminent. The prices of these cars’ most important sensors are plummeting. Laser-ranging lidar units — the things that build up a 3D image of the car’s surroundings — used to cost multiples of the cars on which they were mounted. Now they are priced in the thousands or even hundreds of dollars. Radars, which help pinpoint other road users, are also getting smaller and cheaper, and today’s high-definition video cameras are so cheap they are almost free.
New approaches are around the corner, too. Israeli startup Oryx Vision is working on what it calls coherent optical radar, a system that uses a terahertz infrared laser and clever microscopic antennas to scan the roadway further ahead and in more detail than lidar — and without the danger of being blinded by sunlight or fog. It hopes to build a demo unit by the end of 2017 with thousands of tiny nanoantennas on a chip, eventually costing just a few dollars.
The price drops and gains in performance amount to “a continuous stream of significant advancements in all perception-related technologies,” says Bobby Hambrick, CEO of AutonomouStuff, a supplier of automotive technologies.
The computing power to handle this firehose of sensor data is also becoming more widely accessible. Nvidia now offers an off-the-shelf Drive PX 2 self-driving computing platform, which is likely to feature in robocars from Tesla, Baidu, and others. But it faces stiff competition from automotive suppliers Bosch and Delphi, as well as two Bay Area startups founded by ex-Googlers — Nuro.ai and an as-yet-unnamed company being run by the company’s well-respected former self-driving car lead, Chris Urmson.
And then of course there is Google itself, which has finally spun out its 58 autonomous vehicles, two million miles of self-driving experience, and legion of engineers into a standalone company called Waymo. Waymo seems to be focusing on partnerships with car makers rather than developing its own self-driving prototype.
For this is 2017’s real prize: to develop an operating system that can scale beyond a few experimental autonomous cars to power tomorrow’s mass-produced self-driving fleets. Waymo’s first step in that direction is a partnership with Fiat Chrysler that will evolve next year into a ridesharing scheme using 100 Pacifica minivans.
Waymo has so far shown no inclination to hand over its platform for free, as Google did with its Android smartphone software. But there are others who believe that collaboration and even open sourcing are the future of autonomous transportation.
The notoriously secretive Apple recently suggested to the National Highway Traffic Safety Administration that companies should share de-identified data from crashes and near misses, while Comma.ai founder George Hotz has released open-source software and hardware for lane keeping assist and adaptive cruise control.
Hotz thinks that Comma.ai’s software could become the Android of self-driving cars, used by a network of car makers to which it can later provide profitable services. At the very least, making code and data freely available should prime the next generation of autonomous engineers.
Udacity, an online education provider that offers rapid qualifications in high-tech subjects, has released 238GB of driving data to self-driving self-starters, while the Japanese Autoware project offers a full autonomous software stack for free. “We’ll continue to see more momentum around the open-sourcing of self-driving car tech, and more releases of open driving datasets,” says Oliver Cameron, head of Udacity’s self-driving program.
The first to benefit from cheaper, turnkey self-driving solutions could be the youngest car makers. Startups Lucid Motors, LeEco, and NextEV have recently unveiled luxury cars with self-driving hardware, to be followed at January’s CES show by Faraday Future. All are controlled by or have significant investment from Chinese firms, and all have been subject to questions about funding their prototype vehicles through to production.
Meanwhile, Silicon Valley’s home-grown autonomous taxi company Zoox will surely have to peek out of stealth mode in 2017 to prop up its near $1 billion valuation. On the east coast, NuTonomy and Optimus Ride are scaling up testing in Massachusetts.
“A few startups will breakthrough the way Tesla has, to become major players in a surprisingly short time,” predicts Cameron. Just as likely is a flurry of mergers or acquisitions in the near future, if not the outright failure of one or more.
If the financial outlook is uncertain, the regulatory environment is no less so. Though Trump’s incoming administration seems uninterested in tightening federal rules, that might change if his Rust Belt supporters’ fear of immigration spreads to automation. They have reason to worry: self-driving truck startup Peloton says its fuel-saving platooning technology will reach commercial trucking fleets in 2017, and feisty rival Otto (now owned by Uber) surely won’t be far behind with its own self-trucking technology.
Most states are playing a wait-and-see game on regulation. Notable exceptions are Nevada, which says it will introduce penalties for breaking its existing laws in 2017, and California, which should finally deliver much-delayed rules for public operation. California also looks set to pursue legal action against Uber over its decision not to apply for an AV testing permit before offering public rides in its self-driving cars in San Francisco.
The first self-driving car experiments are over, and the results are clear. The technology is feasible, becoming increasing affordable, and has a multi-billion dollar potential market. But in 2017 the tough work of scaling and commercializing begins. Concerns remain about the safety and reliability of the autonomous vehicles, especially in their interactions with their human operators and other road users. Truly high-definition mapping is still in its infancy and expanding the zones of AV operation will be a slow, painstaking process.
Startups and tech companies need to tackle the staggering capital requirements and vertiginous learning curves of automobile production, while traditional car makers will continue their struggle to reinvent themselves as agile, digital innovators. 2017 may see the first fully autonomous car dart across the nation, but filling in the gaps will take decades.