Most automotive LiDAR (Light Detection and Ranging) systems for self-driving cars now being tested on public roads are of a mechanical design, wherein an environment-surveying laser atop the vehicle constantly rotates 360 degrees. The devices are effective, but bulky and expensive. Emerging solid state LiDAR technology’s low profile and price promises a better fit for mass-produced autonomous vehicles.
Innoviz is a two-year-old Israeli provider of LiDAR remote sensing solutions. The company took a leap forward in the solid state LiDAR field recently by scoring a couple of important partnerships: BMW chose solid-state Innoviz LiDAR sensors and computer vision tech to enable its Level 3-5 autonomous vehicles in April; and last month Innoviz teamed up with Chinese automotive supplier HiRain Technologies to enter China’s lucrative autonomous vehicle market.
Innoviz Founder and CEO Omer Keilaf is an Israeli specialist in electrical optics and micro-electromechanical systems (MEMs). After graduating from Tel Aviv University and developing military technologies for the Israeli Defense Forces, he moved into an R&D role with startups. In 2007 he joined fledgling semiconductor company Anobit, which was acquired by Apple for US$400 million in 2011. He also served as R&D manager at STMicroelectronics, now valued at US$20 billion.
In 2016, Keilaf saw new opportunities for LiDAR, a sensing method that generates 3D-representations of objects by illuminating them with pulsed laser light. The technology has been used in terrestrial and airborne measurement tasks for many years, and now plays a critical role in autonomous driving systems by building a point cloud map of a vehicle’s surroundings.
The 2016 LiDAR landscape was all about mechanical design. Silicon Valley-based LiDAR technology provider Velodyne has dominated the LiDAR market since that time, providing cutting edge mechanical LiDAR systems for Google, Baidu, Microsoft, Tomtom and others. The only hiccup with Velodyne’s products is price. The high-end HDL-64 sensor unit costs some US$75,000 — twice the price of a Tesla Model 3. Velodyne halved the price of its popular VLP-16 LiDAR last year, but even that still costs US$4,000.
Keilaf believed cheaper and smaller solid-state LiDAR systems could be integrated into mass-produced self-driving vehicles. In 2016, he founded InnoViz to do just that, aiming to enable mass commercialization of autonomous vehicles.
“If you look at all of the requirements for Level 3-5 mass production autonomous cars, you need something that is one hundred times better, one hundred times smaller, one thousand times cheaper,” says Keilaf.
Innoviz’s LiDAR is based on MEMS. From Keilaf’s perspective, OPA and Flash are still early stage and have many limitations, while MEMS LiDAR can fill the market gap to more quickly enable full autonomous capabilities.
Innoviz currently has two LiDAR solutions: InnovizPro, its first, standalone solid-state roof-mounted LiDAR; and InnovizOne, a new Automotive Grade LiDAR that can be embedded inside vehicles. (“Automotive Grade” systems have much higher reliability and performance standards than regular commercial grade systems.)
InnovizPro was designed for testing and R&D of autonomous driving technologies, and was not even in Keilaf’s original company blueprint. That changed at the 2017 Consumer Electronics Show (CES), where Innoviz intended to show OEM and Tier1 representatives the Automotive Grade LiDAR they were developing.
“We were telling them about Automotive Grade, but they were saying like: ‘What are you talking about? This [InnovizPro] is really good enough. I want to buy hundreds of these now, and I don’t care today about Automotive Grade, sell me this!” recalls Keilaf.
The team quickly froze the design and took it into mass production when they returned to Israel. They unveiled InnovizPro four months later, teaming up with leading product solutions company Jabil for manufacturing. The product’s commercial launch came earlier this year and it is now available for orders.
InnovizPro’s high-definition intelligent 3D sensing provides an accurate and reliable scan of a vehicle’s surroundings, with up to 150 meters range, 0.15°x0.3° angular resolution, 73°x20° field of view, and 20 FPS frame rate. A field-programmable gate array (FPGA) provides the computing power.
If InnovizPro is an appetizer, InnovizOne will be a full course meal. Using a newer version of MEMS and upgraded detector technology, InnovizOne gets a significantly higher optical power budget that enables increases in resolution (0.1°x0.1°), range (up to 250 meters), frame rate (25 FPS), etc. The built-in FPGA is replaced by an application-specific integrated circuit (ASIC).
Components are now in production and Innoviz says it will start building InnovizOne around September. The company hopes to showcase the new product by year’s end, with a commercial launch target of 2019.
It was InnovizOne that helped the company win its BMW contract. (Innoviz also provides BMW with object detection and classification, and simultaneous localization and mapping (SLAM) computer vision technologies). The leading luxury carmaker has announced plans to bring a Level 3 car to consumers in 2021, while delivering Level 4 and 5 ride-hailing vehicles to urban pilot programs the same year.
Other automakers have also set Level 3 and 4 self-driving cars mass production goals for 2021 and 2022, and there is an industry-wide sense that companies will have to finalize their vehicle tech and designs as soon as possible.
“There is a real demand. There is a really thirst for LiDAR. Today, companies are waiting on a list to get LiDAR,” says Keilaf.
Startups such as Shenzhen-based Robosense and Israel’s Quanergy have also developed solid-state LiDARs in recent years, while industry leader Velodyne introduced its own last year to consolidate its dominance in the automotive LiDAR industry.
Many factors must be taken into account to design a high-quality solid-state LiDAR with a further range, higher spatial resolution, greater frame rate, etc. Heat dissipation for example raises challenges for manufacturers because LiDAR range and accuracy can be affected by operating temperature. Size matters if a solid-state LiDAR is mounted behind the windshield or in the headlamps. And of course there is cost. While the unit price of LiDAR varies greatly according to order size, Keilaf says he expects InnovizOne to retail at hundreds of dollars per unit.
One big challenge for Innoviz is to meet the market demand and ramp up production quickly. Its production lines are set up to reach an initial volume of 1000/month by end of the year. A second site will be established and running as of the second half of 2019 to support volumes of the mass market.
Innoviz is on the right track. The company has raised US$82 million in funding and has over 100 talents on board. Keilaf tells Synced they have already begun development of next-generation InnovizTwo, with no details disclosed yet.
Keilaf laughs when asked about the secret to Innoviz’s success. He says he’s always been fascinated by the 2001 American heist film Ocean’s Eleven, and believes setting a goal, assembling the right talents, and putting in the hard work are the keys to success — whether you’re robbing a casino or building a LiDAR company.
Journalist: Tony Peng | Editor: Michael Sarazen