As the real-world deployment of driverless cars draws near, the race is on to make them even remotely affordable. The systems that currently drive robot cars cost upward of $100,000 per vehicle – not counting the cost of the car itself.
Until that price comes down, there will be no mass market for driverless cars.
As many as 50 start-ups are out to get costs way down, along with the research and development departments at automakers and auto industry suppliers, and Silicon Valley giants such as Google and Apple. Much of their focus is on the most expensive item in the computerized system that controls a robot car: laser-based sensors, or lidar.
The latest competitor is a stealth start-up that comes out of hiding today. San Francisco-based Ouster opened its website to begin selling a laser-based sensor system for self-driving cars to driverless vehicle makers and companies that supply them.
Today, the highest-end automotive lidar systems available cost $75,000, made by a company named Velodyne. Ouster claims its new system, the OS1, boasts near-top performance for less than a fifth of that cost, or $12,000.
“A Corvette is 90% as fast as a Ferrari and it’s 10% of the cost,” said Angus Pacala, Ouster chief executive and co-founder.
Velodyne pioneered automotive lidar, which in turn enabled Waymo, the driverless car division of Google/Alphabet, to begin its driverless car project in 2009. Waymo’s success has fueled radical transformation of the automobile industry as traditional automakers rush to keep current.
Lidar units shoot invisible beams of laser light off objects and measure how long they take to bounce back. Knitting millions of these signals together creates images called “point clouds,” which identify the material world around the car.
The technology can identify building and traffic signals and can discriminate people from animals, bicycles from motorcycles, rocks from soccer balls by shape and how fast and in which direction they’re moving.
Lidar is only one of several sensor systems that make up the robot system in a driverless car. Each has strengths and weaknesses. Video cameras and radar units are relatively cheap, but don’t see the world with the detail of a lidar system. Lidar has problems with rain, snow and fog, because the beams can bounce off snowflakes and water droplets.
Virtually all automakers believe all three sensor systems are critical to the safety and performance of driverless cars.
San Jose-based Velodyne is deep into its own cost-cutting drive, and recently built a second factory to meet heavy demand. Executives are quick to point out that its systems remain the most powerful on the market.
“We’ve been doing this for 10 years, we think we know a lot about what a car needs to see,” said John Eggert, Velodyne director of automotive sales and marketing.
But existing systems “remain very, very niche,” said Arunprasad Nandakumar, market analyst at Frost & Sullivan. And expensive. The price of a lidar system needs to plummet to a few hundred dollars for any chance at mass deployment, he said.
He believes that will happen. He forecasts sales of 6 million lidar units in 2025 — half of them complex lidars needed for driverless cars, for a $2-billion market. (The rest are simpler, single-beam lidars used to detect pedestrians and other obstructions.)
The reason for his optimism: Lidar systems are undergoing the same smaller, faster, cheaper evolution that has put the power of a 1990s supercomputer into a smartphone.
Similar to the way handheld digital cameras were shrunk to fit into an iPhone, today’s mechanical lidar systems and their many moving parts will increasingly become solid-state.
At least 50 lidar startups have been out raising money. Venture investors have sunk $678.4 million into lidar startups over the last four years, more than half of that going to 18 startups in 2017 alone, according to CB Insights.
Traditional auto suppliers such as Continental and Bosch are working on lidar systems. Some, such as Delphi (recently renamed Aptiv) and Magna, have struck partnerships with smaller lidar companies.
Major automakers are working on their own systems while hedging their bets with outside investments or acquisitions. Ford and China’s Baidu have sunk $150 million into Velodyne. Ford acquired lidar startup Princeton Lightwave. General Motors recently bought Pasadena-based startup Strobe. Toyota is working with Luminar.
There won’t be room for most of these systems, of course. “Right now there are six or seven different ways on how to do lidar,” said Oren Rosenzweig of Innoviz, an Israeli lidar startup working with Delphi and Magna. “There are not seven different ways to make a DVD player.” Ultimately, one or two approaches wlll prevail, he said.
Waymo and ride-hailing service Uber are also working on their own lidar systems. Uber currently is defending itself in federal court on Waymo’s charge that it stole lidar trade secrets.
The only driverless player that sees no need for lidar is Tesla, whose CEO, Elon Musk, believes his company’s software can make cameras, radars and ultrasound work together well enough to do without. As with just about everything at Tesla right now, proof of that concept is off in the future.
The children of Generation Z might look at pictures of today’s driverless cars and laugh at the ungainly gadgets that resemble spinning coffee cans or kitchen blenders rigged atop the car roof. The newer models often look like oversized hockey pucks.
The Velodyne technology is known as “mechanical” even though it’s packed with sophisticated electronics. With 64 laser channels, the cylindrical unit spins for a panoramic 360-degree look at the world.
Velodyne offers a much smaller unit fitted with 16 lasers. Fewer lasers means lower performance on range, radius, refresh rates and so on. But it costs $8,000, nearly a tenth the cost of the company’s most capable system.
Start-ups are exploring different methods. Innoviz uses stationary lasers that reflect off a tiny mirror that oscillates at absurdly high speeds on a single axis. Ouster is starting out with a “hybrid” system that spins like Velodyne’s but with far fewer semiconductors and other electronic parts inside.
“We looked at over 50 companies in the lidar space, and found 14 of them to be reputable,” said Chris Thomas, founder and partner at Fontinalis Partners, a Detroit-based venture capital firm and early Ouster investor. “Are some of them going to make it? Yeah. Are all of them? No way.”