Why Delphi and Mobileye think they have the secret sauce for self-driving cars

Delphi and Mobileye, two leading auto-part suppliers, aren’t as well-known in the self-driving space as their much bigger rivals Google and Tesla. But Delphi, a GM spinoff now based in the UK, has been showing off its autonomous technology for over three years, and Mobileye, one of Israel’s biggest tech companies, is a leader in crash-avoidance and vision technology. Together, the two companies think they have the secret formula to beat their competitors to market, and it all comes down to two ingredients: computing power and crowdsourced maps.

Delphi and Mobileye joined forces earlier this year to announce their plan to create a mass-market, off-the-shelf autonomous driving system that can be plugged into a variety of vehicle types, from smaller cars to pickup trucks to SUVs. Yesterday, they announced that Intel will provide specialized chips for their system that when combined with Mobileye’s EyeQ®5 chip will be capable of computing about 20 trillion mathematical operations a second. Later versions may have two to three times that processing power.

The companies say they hope to start contracting their self-driving kit to original equipment manufacturers (OEM) like Volvo and GM by 2019. The goal is to have a fully developed package available for automakers to start integrating into new vehicles by 2019.

Once Delphi is awarded a production contract from an OEM, the package has to be integrated into the vehicle, calibrated, and validated before it goes on sale, said Sam Abuelsamid, senior research analyst for Navigant. “That’s a process that takes about 18-24 months so you won’t see production models with this system until about 2021, the same time frame as most other companies,” he said. Commercial fleet operators, like airports and big-box stores, are being eyed as potential customers.

I rode in one of Delphi and Mobileye’s self-driving cars in Pittsburgh earlier this week. The ride was unremarkable, as most driverless vehicle trips these days are. The operators tend err on the side of caution, programing the car to take safe routes at below average speeds.

Still, their car, an Audi SUV, looked noticeably different than those self-driving cars operated by Google, Ford, and Uber. It lacked the spinning-top lidar detector on the roof. No bulky array of cameras. No sensors sticking out like a sore thumb. Glen DeVos, Delphi’s vice president of engineering and services, told me the goal was to produce a driverless car that looked just like any other vehicle on the road.

“We don’t have a big mass array of sensors on the top,” he said. “They’re actually integrated very nicely with the vehicle, because that’s what a production vehicle would look like.”

The ride lasted less than 10 minutes, a quick trip from the parking lot of Ottomatika, an automated driving startup acquired by Delphi last year, through a series of office parks, along a major road, and then back. A safety driver sat behind the wheel, though the car required no intervention along the way.

I asked DeVos how his self-driving system compared to those produced by Google, Uber, or Tesla. “Google certainly has the longest history,” he said. “Where we think Mobileye and Delphi have a real advantage, we have the greatest history with respect to automotive. We understand how to put those technologies into the car.”

We approached a stop sign, and the car came to a halt. It inched forward, then came to another jerking stop. “That’s exactly what we’re trying to fine tune,” DeVos said. “It had a hesitation there. And it always errs on the side of safety. If it sees something that it isn’t sure what it is, it’ll stop.”

So what’s fueling Delphi and Mobileye’s cockiness to declare it will begin selling automated driving kits in just two years? The companies’ executives say that it’s a combination of computing power, as enabled by their partnership with Intel, and a unique approach to digital mapping that gives them an edge over their more recognizable competitors.

Rather than relying on a fleet of self-driving cars to map a particular city, like Google’s doing in Mountain View, Austin, and Kirkland, Washington, and Uber’s doing in Pittsburgh, Delphi and Mobileye are essentially outsourcing that mapping activity to millions of other traditional vehicles equipped with Mobileye’s vision technology.

The Israeli company supplies many of the top OEMs, or original equipment manufacturers, in the automotive world. Mobileye says that over 14 million cars with its cameras and vision technology are currently on the road. The data captured by those vehicles can be sent to the companies’ self-driving cars to better predict constantly evolving road conditions, such as construction zones and double-parked vehicles.

“It’s these driver-assist normal vehicles that are creating these maps,” said Dan Galves, senior vice president at Mobileye. “Meanwhile, the autonomous vehicle, as it’s driving down the road, can say, ‘Dear map, please tell me where the last million cars drove.’”

Mobileye call this technology Road Experience Management, or REM, which uses real-time data to constantly update a digital map to help guide self-driving cars. This type of data is easier for vehicles to maintain and store because it’s not as large as other digital maps, and it still works in areas of little-to-no wireless connectivity.

Of course, Delphi and Mobileye aren’t the only ones using crowdsourced maps to power its self-driving cars. In July, Ford invested in a company called Civil Maps that is doing the same thing, although that company can use multiple sensors whereas Mobileye is focused on vision sensors. Google is also using crowdsourcing with its cars. “Pretty much everyone is going this direction,” Navigant’s Abuelsamid said.

“Delphi is actually a far bigger company than either Tesla or Uber in terms of revenue and the company has supplier relationships with most of the world’s top automakers,” he added. “As an experienced supplier, Delphi has the advantage of understanding what it takes to design, validate, manufacture and support automotive grade systems and is actually far better positioned to supply integrated [advanced driver assistance systems] and autonomous systems to automakers that don’t have the resources to develop their own.”

Since both companies are auto-part suppliers, they rely on partnerships with big car makers to fuel their self-driving dreams. In July, BMW announced it was teaming up with Mobileye and Intel to bring a fully autonomous vehicle to the road by 2021. Mobileye’s sensors are also used by GM and Volkswagen. And last August, Delphi launched a fleet of self-driving cars in the city-state of Singapore for a mobility-on-demand program.

However, one partnership that won’t be going forward is the one between Mobileye and Tesla, which previously had been using the Israeli company’s technology to recognize images for its semi-autonomous Autopilot system. That break came after the two companies expressed disagreement over what caused a fatal accident in May involving a Tesla driver using Autopilot.

Mobileye and Delphi face formidable competition from all sides. Chip-maker Nvidia is outpacing Intel in this space, producing extremely powerful processing units used by Audi and Tesla. Qualcomm is also pursuing contracts.

Delphi and Mobileye say they will demonstrate the latest version of their self-driving technology at CES in 2017 in a rigorous 6.3 mile course through the heart of Las Vegas. This will include challenges such as highway merges, congested city streets with pedestrians and cyclists, and a tunnel.

Having now ridden through Pittsburgh in two self-driving cars, one operated by Uber and the other by Delphi and Mobileye, there isn’t much to distinguish the two trips. Uber let me behind the wheel, and took me for a longer trip along the Allegheny River, while Delphi and Mobileye were more restrained in what they were showing off to the media.

But Delphi and Mobileye have been working on automated driving for much longer than more recent entrants like Uber. And it showed in their prototype vehicle’s clean-looking exterior. It was the kind of car that you wouldn’t look at twice if you saw it drive by, so long as there was actually someone behind the wheel.