The multibillion dollar race to put autonomous cars on the road might not be won by a carmaker at all
SANTA CLARA, USA — Singapore could be the first country in the world to get a widespread transport service made up entirely of self-driving cars. That’s the opinion of Glen De Vos, the chief technology officer of Delphi Corporation, a major automotive supplier that is working to run such a robot taxi business here.
Delphi already has one self-driving car on the road here, and will introduce two more this month to gather more data as it prepares to launch an on-demand taxi service without the taxi driver.
The company has been working with the Land Transport Authority since August last year on an autonomous taxi trial. The project is designed to figure out what is needed on the infrastructure side — in terms of the road network, data centres and so on — to make robot cabs viable here.
Delphi CTO Glen De Vos says Singapore is at the top of the list of cities that could get self-driving taxis
If all goes to plan, by 2020 you should be able to summon a driverless car with your smartphone to take you wherever you want.
“We’re not doing it as a science experiment. We want to work with the LTA to actually launch a commercial service,” says De Vos. “I would put Singapore at the very top of the list of cities that could deploy this quickly.”
But if the thought of sharing the ECP with a self-driving taxi fills you with dread, it shouldn’t.
The Delphi car is similar to another prototype that drove across America in 2014. The engineers sitting in it took the controls for only one percent of the 5,000km journey, meaning if you had a car like that, your 20km drive to work would need you to steer it yourself for 200m.
And while you and I have just two eyes (well, four between us, but you know what I mean) to study our surroundings with, the Delphi Audi has 26 cameras and sensors that scan the world around it constantly.
But what will make robot cars a reality isn’t eyes. It’s brains.
This car has two blue ovals, but it’s not the one with the Ford logo that counts…
That’s where Intel jumps in. The chip-making giant’s microprocessors already power everything with a keyboard, and it’s scrambling to have a presence that’s as ubiquitous on the road as it is in the office.
Delphi’s Audi is controlled by Intel chips — more than 100 autonomous prototypes around the world are, actually — and the Santa Clara-based company is building a kit that carmakers can eventually plug into their products. Makes sense to us. If you manufacture cars, why spend billions on R&D when you can buy an off-the-shelf solution and tailor it to your needs?
Eran Sandhaus, vice-president of software and services for Delphi, says the company’s autonomous driving kit should be ready by 2019, and cost “thousands of dollars”.
BMW, which is also partnering with Intel, has a more modest target. The iNext, its first fully self-driving car will go on sale in 2021. Meanwhile, it is putting 40 prototypes on the road to figure out how best to integrate the technology with its cars’ basic architecture.
A BMW that comes with a built-in chauffuer? That will be a reality by 2021.
While it’s obviously early days in the race to autonomy, a potentially huge prize waits at the finish line. Goldman Sachs, an investment bank, estimates that the market for autonomous cars could be worth US$96 billion (S$135 billion) a year by 2025, and US$290 billion by 2035.
Until then, it almost seems like the cars themselves will be the easy part. Intel says an autonomous vehicle will generate four terabytes of data a day — and you thought your monthly 6GB cellphone plan was plenty.
Making sense of that digital tsunami is going to take serious computing power, says Jack Weast (below), the chief system architect for Intel’s autonomous driving division. “We see this as a data challenge,” he says.
Not all of the computing power has to be in the cars themselves, he says. Once a 5G mobile network is up and running here, it’s data centres that will keep autonomous cars running properly. “If you connect to the cloud, you can download high-definition map information, for example,” says Weast. That ensures the cars would never be confused by the sudden appearance of a new road.
Connected autonomous cars could also effectively talk, keeping up a constant flow of digital chatter that would warn them about road hazards long before a human driver would encounter them.
They could avoid jams, or drive in ultra-close formation to cut fuel consumption by as much as 15 percent by sharing their wind resistance.
There’s some chance that those Intel-brained cars will drive better than most of us, too. On a short ride in Delphi’s Audi revealed how the car accelerated smoothly but decisively, signalled religiously, and lined itself up in good time for its next move. Better, in short, than ComfortDelgro’s woeful norm.
Intel’s Weast predicts a huge drop in traffic injury or fatality once autonomous cars become the norm, along with a corresponding fall in insurance costs. The elderly and the very young would gain mobility, and an enormous amount of land could be reclaimed from today’s parking lots, he says.
“Autonomous vehicles are going to be very, very valuable to society at large,” he says.
To what extent that comes to pass, and how soon the kind of driverless future imagined by Weast and other engineers comes to fruition remain to be seen.
But given how much computing power is going to be needed to pull it off, one thing seems certain: there’s a high chance that whichever carmakers end up leading the autonomous race, their cars will have an Intel inside.