Aspirations to place self-driving technology in the hands of ordinary vehicle owners had largely been relegated to the back burner.
As automakers and technology companies better understood the enormity of that challenge in recent years, they’ve chosen to focus on simpler tasks in the shorter term, concentrating on building virtual drivers for niches such as geography-constrained taxis, last-mile delivery and high-way travel.
With the exception of a certain billionaire who pitches “full self-driving capability” different from the rest of the industry, most companies have considered the real deal something far off in the fuzzy future.
That’s changing. Global supplier Mobileye unwrapped plans last week to make self-driving technology available in personally owned vehicles in 2025. The Intel subsidiary made the announcement during CES.
“Robotaxi will be somewhat of a game- changer when it’s ubiquitous, because you are eliminating the driver. But having a consumer AV? That is completely disruptive; that is completely game-changing,” said Mobileye CEO Amnon Shashua.
Few competitors have even broached the subject. Waymo has an agreement to work with Fiat Chrysler Automobiles on self-driving tech for personally owned vehicles, but that remains a “longer-term opportunity,” according to a Waymo spokesperson.
Tesla’s Elon Musk has touted plans to release a “full self-driving” feature in 2021. But Tesla defines “full self-driving” as still requiring a human driver who’s responsible for vehicle operations — much to the consternation of safety advocates and competitors.
Industry analysts don’t foresee self-driving vehicles that are defined by SAE International as Level 4 and above reaching dealerships for another decade. Brian Collie, global leader of automotive and mobility at Boston Consulting Group, forecasts self-driving cars will begin to reach consumers as personally owned vehicles by the early 2030s, with 1 million sold by 2033 or 2034. Even then, he expects technical challenges will persist.
“In a Level 4 operation, that will be available where there’s very, very detailed localization and high-definition mapping, where environmental conditions are updated on a very, very, very frequent basis,” he said.
That’s exactly where Mobileye intends to succeed.
What may enable Mobileye to expand autonomy into the realm of the car-owning masses is the collective insight of that crowd itself. For the past five years, Mobileye has been harvesting data from approximately 1 million vehicles on the road that contain its cameras and computer-vision systems.
Each day, the company collects nearly 5 million miles of driving data via this method and uses it to create high-definition maps, which give self-driving vehicles precise knowledge of their location in the world and an additional layer of information with detailed knowledge about their surroundings.
It’s an unusual approach. Most companies use lidar sensors to make these maps. By using cameras, Mobileye says it can make a big impact with just a little data — about 10 kilobytes per kilometer, which lowers the cost. By harnessing the crowd, the company can move beyond geography constraints and build maps from any roads its crowd drives.
“We can build maps of the entire globe, and this is where we are going,” Shashua told Automotive News. “The issue of scale maybe is not that critical right now. You do a robotaxi in maybe Phoenix or San Francisco, and update that map when things change. … But if you want to drive everywhere, we need the high resolution to be at scale, and you can sell this kind of function.”
How much consumers will be willing to spend remains a question.
Collie estimates that a self-driving system will cost approximately $9,000 per vehicle for manufacturers to make. By the time it’s marked up, he expects it would cost consumers $15,000 for self-driving capabilities, and that may be a reach for buyers.
“We think that’s a pretty significant uptick, even for folks in the premium range,” he said. “There’s a strong economic opportunity for robotaxis and in trucking, but in personal vehicles, the overall value proposition isn’t as great in something less than a Level 5 environment.”
But Mobileye believes a much lower price point is possible, starting with a “bill of materials” cost of $3,500 for the components of a self-driving system. Savings are possible, in part, because the company is developing its own sensors. Mobileye will use Luminar’s lidars in vehicles intended for commercial robotaxi operations, which are slated to begin in Israel in 2022. But for personal vehicles, Mobileye will use Intel’s expertise in manufacturing silicon photonics products to develop its own lidar system-on-chip starting in 2025.
No automaker has yet contracted with Mobileye to fit its self-driving systems on personally owned vehicles. But last summer, Geely said it would add the supplier’s SuperVision hands-free driver-assist system. Mobileye has crafted SuperVision as a foundational block in a steppingstone approach; it’s a camera-only system capable of Level 4 operations but is utilized for driver assistance.
When Mobileye begins Level 4 operations in robotaxi and personally owned vehicles, it intends to add lidar and radar on a second, redundant subsystem that cross-checks the actions of the camera-only system. By developing them separately and running them in parallel, Shashua says, the chance of failure is reduced by an order of magnitude.
Beyond Tel Aviv and Jerusalem, the company opened test beds in Munich and Detroit in 2020. Last week, it announced that additional test deployments in Shanghai, Paris and Tokyo will commence this year. Depending on the regulatory climate, Shashua hopes New York City will be added to the list.
In each location, Mobileye wants to showcase its system’s ability to evolve from a driver-assist feature to fully autonomous driving, with operations underpinned by the crowd-source maps that make both possible.
“That ability to have the high-resolution map everywhere, simultaneously, is critical, because you cannot sell a self-driving system to a consumer that will only be activated in San Francisco,” Shashua said. “So it’s not just thinking about the scientific experiment of proving to yourself or someone else, ‘Here, I can take a territory like San Francisco and drive there without a driver.’ If you want to build a business, not only now, but three or four years from now, being able to have high-resolution maps at scale is crucial.”