Turns Out the Hardware in Self-Driving Cars Is Pretty Cheap

The full benefits of autonomous cars—a drastic drop in fatal crashes, more efficient driving, the chance to master Clash of Clans while commuting—are at least five years away. Automakers and tech companies must perfect software. Insurance companies have to recalculate premiums. Regulators are thinking up fresh rules.

But one part of the puzzle’s been solved: the hardware. The physical elements crucial for seeing and navigating the world have been developed. Some are already in production. And they’re surprisingly cheap.

The cost of self-driving cars isn’t often discussed, mostly because we’re still years from commercial production, and because there are much tougher questions to answer before we can talk money. But a new report on the market for and development of self-driving cars, by the Boston Consulting Group, offers some estimates. And partly thanks to that affordable hardware, they’re not that high.

According to the group’s research, urban and autopilot will each add about $5,500 to a car’s price tag. You can have a car that parks itself for an extra $2,000. If you want full autonomy—the ability to drive anywhere, with no human input—get ready to add $10,000 to the price tag, at least in the first 10 years the technology’s on the market.

“The technology is already there,” says Xavier Mosquet, head of Boston Consulting Group’s North America automotive division.

Hardware falls into three categories: sensors, processors, and actuators. Given the high level of electronics in today’s car, actuators—the bits that allow a computer to physically do things like brake, change gears, and steer—don’t pose a problem. Processors handle data from the car’s sensors, and direct the actuators as they see fit. They’re basically ready, though there’s room for more speed and costs to come down a bit.

Then there are the sensors, which do the critical work of reading the world around the car. The above diagram, from the Boston Consulting Group report, points out the different technologies being used. Cameras, radars, and ultrasonic sensors are cheap, ranging from as little as $15 to $200. That’s largely because they’ve been common for several years, necessary for features like lane keeping and adaptive cruise control. Just as those advanced driver assistance systems are slowly getting consumers ready for cars that drive themselves, they’re helping the industry get there.

There is work to be done, however, in two places: LIDaR (Light Detection and Ranging) and high-end GPS. The LIDaR on Google’s self-driving car uses 64 lasers to map the physical the world, collecting more than a million data points about its surroundings every second. It costs north of $50,000. The next-gen Google car will have an $8,000 version, Mosquet says, still way too expensive for consumer adoption. You can get a LIDaR for less than $100, but that’s a single-beam unit made for much simpler applications.

If you’re using GPS to help operate, not just navigate the car, it needs to be accurate within centimeters, not meters. That capability exists, but before it can be used on a production car, its cost needs to come down significantly from the current $6,000 figure. Same for LIDaR. “You definitely need more and better technology development,” Mosquet says.

The good news is that automakers don’t have to rely on any one sensor for their self-driving platforms, and there are lots of ways to combine these technologies to reach autonomous capability.

Now we just have to answer all those other questions. And you know, build the things.

No comments:

Post a Comment