On Monday, Mercedes Benz, Audi and BMW announced an unlikely partnership. The three car companies dropped a collective $3.1 billion to buy Nokia's digital mapping tech, dubbed HERE.
“HERE will play a key role in the digital revolution of mobility, combining high definition maps and data from vehicles to make travel safer and easier for everyone,” said BMW honcho Harald Krüger, in a press release that promised that HERE's services and products would remain available to "all customers from the automotive industry and other sectors."
The deal, which is expected to close in early 2016, is a clear signal that the auto industry plans to compete head-to-head with Google and other tech companies developing self-driving cars. And it's a big, and very telling, step: individual car companies are willing to join forces with competitors to have a better shot at winning the robocar race. Because when it comes to driverless cars, the best manufacturers will be those with the best, most up-to-date maps. And that honor currently goes to Google and its 2013 acquisition Waze.
The thing about autonomous cars is that they will need to know where they're going, at a very granular level, in order to be safe and reliable. The HERE acquisition likely reflects car companies' realization that they seriously lag behind Google when it comes to artificial intelligence and mapping.
"Carmakers have some catchup to do," said Chris Nicholson, CEO of AI company Skymind. "It's a huge transition for them to make."
The search giant, which stealthily created a robocar company in 2011, has become a clear leader in autonomous cars, and has a wealth of mapping data already. It started its mapping project 10 years ago. Since then, it's revved up its efforts to map every single corner of the earth—including far-flung places like the Grand Canyon where digital maps traditionally had blind spots—with its Google Street View fleet, which is made up of both cars and pedestrians. It augmented that when it bought Waze, a company that crowdsources navigation data, for $1 billion.
"The auto industry is understanding that data is very, very valuable, whether it's consumer or mapping data. They clearly felt that they wanted to be able to own it," said Daniel Shapiro, the senior director for automotive at Nvidia, which provides many major car companies with computer chips. "The move to autonomous vehicles emphasizes how critical owning that information is. The more accurate the map, the more precise the vehicle can be."
Car companies aren't the only ones racing to catch up with Google. In 2012, Apple, which is rumored to be developing its own autonomous vehicle, started its own mapping initiative. And last month Microsoft, which just invested $100 million in Uber, sold part of its mapping platform to the on-demand cab company. Earlier this year, Uber also poached an entire self-driving car lab from Carnegie Mellon.
Car companies do have one huge advantage over Google: they already have the infrastructure to build and market cars. Google is coming at it from the opposite side. It has spent more than a decade building an arsenal of data and expertise in artificial intelligence that can be baked into autonomous vehicles. Initially, it looked like Google would simply retrofit already-existing cars, like Lexuses and Priuses, with its software, but now it's clear that they actually want to compete with car companies, building their own friendly-looking smart cars.
A fully autonomous commercially available car is still years away, but consumers can expect to sit in vehicles that have autonomous capabilities much sooner. Audi, for instance, says that some of its next-generation cars will have traffic jam assist, a feature that lets the car drive itself on highways when going at less than 37 mph. To work well, these efforts will require very detailed maps.
As Katelin Jabbari, who handles communications for Google’s self-driving cars, told Gizmodo in May, "the first thing we have to do before we can drive autonomously is map the roads.”
I recently rode in a Mercedes Benz autonomous concept car. The aardvark-looking thing just drove around in one big loop, but even that required engineers to map out the exact route, its turns and stops, in minute detail. In January, an autonomous Audi A7 drove 500 miles from Palo Alto, California to Las Vegas. Again, that entailed a tremendous investment in mapping out speed limits, road signs, the number of lanes, and which ones are turn-only or exit-only lanes. For fully autonomous cars to hit the road, they'll need that, and more.
HERE starts to fill in some of the gaps car companies face. Before the acquisition, HERE engineers were working on developing "self-healing maps" that could be updated in real-time. That's important because, as you know from driving around, road conditions constantly change. The ability to access maps that are updated instantly is a huge boon for autonomous driving. The company had also tasked about 6,000 employees spread across 200 offices with driving mapping vehicles, processing licensing requests, and doing aerial photography, according to USA Today. That's made HERE's technology super granular, with resolution down to a few inches.
"There's a lot of data on a map you don't think about when you pull up a map on the street. You don't know lane by lane what's going on," said Nvidia's Shapiro. "A car has to have very precise software and hardware to understand all that."
In addition to mapping data, robocars have to be able to sense their environment. That's where all the sensors in the car come into play. Prototypes for self-driving cars are pimped out with cameras, GPS and lidar sensors that use lasers to survey their environment. These serve as digital "eyes" that feed into computer vision systems that allow the cars to "perceive" their surroundings and make smarter calls. That data also feeds back to the mapping platform the car uses to navigate, theoretically allowing cars to maneuver in dynamic situations.
"Being able to compare the data you're sensing to known [map] data can really improve the reliability of the system," Shapiro said. "The more types of information you have to overlay builds redundancies. It's like having a second opinion. Having the mapping data, plus other sensor input will help a car make better decisions."
In order to do this, cars need to have powerful computers on board to process that information in close to real-time. And when it comes to working with powerful computers, Google again has an advantage.
Daniela Hernandez is a senior writer at Fusion. She likes science, robots, pugs, and coffee.