Luminar sees the future of autonomous vehicles in 3-D
- Summary:
- If you find yourself in the passenger seat of an autonomous vehicle in the next few years, you can likely give a lot of the credit to a 23-year college dropout named Austin Russell.
Self-driving cars represents the moonshot for artificial intelligence (AI) and machine learning. when AVs can pilot through a city like Washington, D.C., using AI, that means AI is ready for prime time in every single industry in America. It will change everything in this country.
The report predicts autonomous vehicles will contribute some $800 billion annually “in economic and societal benefits” when AVs are fully deployed. The benefits will come from cost savings related to reduced crashes, fuel efficiency and better access to transportation.
But first
Before all this wonderfulness can happen, however, the industry has to convince a deeply skeptical public that self-driving vehicles can perform better in any highway situation than a human driver.
A recent AAA survey found that 73% of American adults were “too afraid to ride in a self-driving car,” including 64% of millennials, who have generally been more open to embracing the trend of autonomous cars.
That means the eventual winners in the industry will likely be those companies that can build the safest vehicles and establish the highest degree of trust with the public. The airline industry has developed that level of confidence over the decades, but autonomous vehicles remain unproven to a majority of consumers.
To achieve the required levels of safety, vehicles must be able to “see” and identify objects and obstacles around them and make the best possible safety decisions in the shortest possible time. Mistakes can have fatal consequences.
The 'eyes' of autonomous vehicles are something called Lidar sensors. “Lidar” is short for “Light Detection and Ranging” and is a laser range-finding system that maps out in 3-D an object's surroundings with multiple lasers and relays them to a computer for immediate analysis.
Common Lidar systems in use today look like spinning coffee cans mounted on top of autonomous vehicles. They fire 64 lasers simultaneously connected to sensors that relay information from every surface a laser 'touches' and returns it back to a computer to process and build a wireframe map of the world it sees.
Lidar’s ability to generate detailed and eerie 3-D images under all lighting conditions, working in sync with the cameras and radars, gives self-driving vehicles an awareness of their surroundings and road conditions. They are essential for cars to safely navigate alone and distinguish objects such as pedestrians and cyclists.
They are so important that they have touched an acrimonious legal battle between Uber and Alphabet’s self-driving car unit, Waymo, who says one of its engineers took secret plans for Lidar sensors to the ride-sharing company. They are so crucial that Alphabet and Uber were forced to invent their own, better-performing Lidars for their cars.
David Hall, the same inventor who originally brought us the distortion-free subwoofer back in the 70s, invented modern three-dimensional Lidar more than a decade ago for use in the DARPA Grand Challenge competitions. His company, Velodyne, has dominated the market for self-driving car Lidar ever since. Last year, Velodyne opened a factory that it said could produce a million Lidar units in 2018—far more than any other maker of high-end Lidars. But it is about to get some serious competition.
That is where Austin Russell comes into the picture. Russell dropped out of Stanford in 2013—when he was still 17—in return for a $100,000 check under a program started by venture capitalist Peter Thiel to encourage entrepreneurship. He used part of the money to launch a startup Luminar, with the goal of building a Lidar system that sees farther and in more detail than those on the market already (like Velodyne) and become the standard of the emerging autonomous vehicle industry. What he didn’t yet know was how to do it.
Russell brought on Jason Eichenholz, a serial entrepreneur and pioneer in laser, optics and photonics product development and commercialization, as co-founder and CTO. They have spent the last five years in stealth mode designing a new Lidar system from scratch in secret.
The hard work has started paying off. Luminar received $36 million in funding in April 2017 from venture capitalists including GVA Capital, Canvas Ventures, and the 1517 Fund, a venture firm backed by Thiel and was picked up by the Toyota Research Institute (TRI), the division of Toyota that deals with robots and self-driving cars. With the money, Russell also acquired one of Luminar's suppliers to lock up its supply chain from competitors.
Volvo bets on Luminar
The most significant endorsement came in June 2018 when Volvo Cars announced that not only was it buying its Lidars from Luminar but also that it Volvo will also be making a “strategic investment” in the company via its newly formed venture capital fund. Neither Volvo nor Luminar would disclose the amount, but Russell allowed that it was “substantial.” He added:
This is a partnership with Volvo Cars to power their autonomous vehicle development effort with our Lidar sensing platform at its core. We’re the first to be able to achieve the necessary performance and safety requirements to get to that better-than-human level perception and ultimately safer-than-human autonomous capability.”
What makes Luminar different?
Unlike the multi-laser spinning coffee can on the roof approach, Luminar's Lidar is box-shaped with an all-glass front and contains a single laser. The laser is then refracted multiple times to provide the same basic mapping structure as the traditional Lidar system but with considerable more range and accuracy.
The company says it is 40 times more powerful and has 50 times more resolution than the most advanced LiDAR sensors deployed in vehicles today and can recognize a dark or non-reflective object on the road at a standardized 10 percent reflectivity target at 200 meters.
By contrast, Velodyne’ sensor can see dark objects, such as pavement or a person in dark clothing, with 10 percent reflectivity at a distance of 50 meters.
What that means as a practical matter is that in a car traveling at 70 miles an hour, an extra 150 meters of vision would give the car’s software an additional four seconds to take action when it sees an obstacle. In safety terms, that’s a huge deal.
Luminar can also zoom in on a particular object by directing more laser beams in that direction, using a system of small, moving mirrors that actively steer its laser. In contrast, sensors from others use spinning mirrors that send out beams in a fixed pattern.
Luminar currently has over 150 employees, split between its headquarters in an old tank repair facility on a 25-acre "compound" in Silicon Valley and an optical engineering and production facility in Orlando, Florida. The first run of 10,000 units is scheduled for production this year.
My take
With so many companies spending billions of dollars to make autonomous vehicles an assured inevitability, the future looks very promising for the mobility industry. Dozens of auto and mobility companies are involved in AV development, ranging from the testing of small electric shuttles in cities like Las Vegas, to Ford’s move to restore and redevelop the abandoned Michigan Central Station in Detroit as the company’s campus for self-driving car research and development. At the moment, 56 companies hold autonomous vehicle testing permits in California.
Over time and with lots of miles of safety data to prove that the combination of AI and Lidar sensors are safer 'drivers' than humans, public skepticism is likely to go the way of fear of flying for most consumers. FWIW, I’m old enough to remember the skepticism and resistance when seat belts were initially made mandatory.
And, of course, there will be lots of bright new startups like Luminar to discover unexpected safety improvements and drive breakthroughs in new AI software to keep the momentum going. Mr. Robot, start your engine.
