Among the most competitive component technologies for autonomous vehicles is lidar, even as the debate continues over whether such sensors, radar, or optical cameras are better. SiLC Technologies Inc. has developed its own silicon-based solution for creating 4D maps for self-driving cars.
Monrovia, Calif.-based SiLC, which stands for “Silicon Light Chip Technologies,” was founded last year and came out of stealth mode at CES 2019.
“Conventional imaging solutions are based on the use of an image sensor to convert photons to electrons directly to establish a 2D image,” the company said on its website. “This immediate conversion process ignores a wealth of information the photons carry in their phase, wavelength, and polarization.”
SiLC said its “Smart Vision” system on a chip integrates optical signal processing to extract additional data on depth and velocity from the returning photons before their conversion to electrons.
“We’re integrating optical functionalities in a silicon chip in the same way that the electronics industry integrated circuits into silicon chips,” said Mehdi Asghari, CEO of SiLC. “We’re taking advantage of the same manufacturing infrastructure but using photons instead of electrons.”
“Our solution is different from other lidar, which looks like a KFC bucket on top of a car,” he said. “That’s not the future of lidar for cars, robots, or drones. We see lidar as miniature sensors, like the tiny camera in your phone.”
Building on silicon photonics
“This is my third startup — one, Bookham, went to IPO [initial public stock offering], and the other, Kotura, was a successful commercial sale,” Asghari told The Robot Report. “Both were also in silicon photonics and commercially oriented productization.”
“In my previous startups, we made devices for the telecom industry and the data center,” he explained. “Now, SiLC is focused on 4D visual perception systems.”
“With coherent lidar in a 1,550nm wavelength, you get a continuous wave that’s immune to other lidar and sunlight,” said Asghari. “FMCW [frequency-modulated continuous wave] provides Doppler information and instantaneous velocity. It’s also more energy-efficient and safer for the human eye.”
“With time-of-flight lidar, you don’t get the same accuracy. Our system delivers two to three orders of magnitude in measuring distance and velocity,” he claimed. “You can actually control performance by changing parameters such as range and accuracy at the software level. This enables the same solution to serve a multitude of markets, like cars, robots, and drones.”
“For industrial use, we’re coming up with a very modular design,” Asghari said. “You can put the sensor head anywhere in the car and connect it by a short optical fiber link to the vision processor module or VPM, which you can put where it will be safe. You can replace a relatively low-cost sensor head without worrying about it being damaged or stolen. SiLC’s technology could go in a headlamp or a rearview mirror.”
“This architecture enables greater flexibility and discretion, and it’s equally convenient for a robot or a drone — think of integrating a sensor head to a VPM the size of a matchbox,” he added. “The VPM could share power with other electronics and integrate with other sensors.”
“With silicon on a chip, you could use our sensor heads and data from a camera to perform sensor fusion, but we don’t do that yet in our VPM,” said Asghari. “We will make the VPM available to other sensors.”
SiLC ready to move from testing to manufacturing
SiLC launched its first chip last month and is in early testing. “We’ve gotten positive feedback from potential customers, like automotive manufacturers,” Asghari said. “They can start deploying them on Day 1 and don’t need a full harness. They can retrofit existing models and eventually have sensor networks in cars.”
“Our fully integrated lidar should be ready by the end of summer, and we’ll demonstrate products at CES 2020,” he said. “We’re going to demo a short-range lidar at tens of meters, suitable for robots, and a long-range lidar, at 200 meters, at CES.”
“We’re getting inquiries from people who want to test the product,” noted Asghari. “We hope to ship to customers in the first quarter of 2020.”
“The prototypes and demo systems were made in Asia,” he said. “We’re making sure our supplier ecosystem is lined up and is ready to ramp up manufacturing.”
“We have a close relationship with a dedicated fab that has more than 20 years of experience,” he added. “We’re able to inherit its capability and knowledge.”
Lidar business grows, but costs need to drop
Shipments of automotive lidar sensors will increase from 100,000 in 2018 to 34 million units per year by 2032, predicted Woodiside Capital Partners. The solid-state lidar market will experience a high compound annual growth rate once it is commercialized in the coming year, said Global Market Insights Inc.
Numerous companies are getting into the lidar business. For instance, Waymo is offering its custom lidar sensors to customers that don’t compete directly with its autonomous taxi business.
Aurora Innovation Inc., which raised $520 million in its Series B round and another $600 million more recently, acquired Blackmore Sensors and Analytics Inc. in May. Sense Photonics, which is building 3D sensors, closed on a $26 million Series A round in June. It was joined by $38 million for Innoviz Technologies Inc. and $25 million for BrightWay Vision Ltd.
This month, Analog Devices Inc. and First Sensor AG partnered to work on sensors, and lidar maker Luminar raised $100 million. At the same time, researchers are continuing to pursue vision-based alternatives to lidar.
“Elon Musk’s main objection to lidar is based on cost,” said Asghari. “He’s setting the bar for multidimensional cameras to add safety to self-driving cars. To achieve real full autonomy, we’ll need a range of sensors and visual systems.”
“Current prices for lidar aren’t sustainable; they will have to go down,” he acknowledged. “The only companies that will survive are those with the right cost point.”
“For the automotive industry to deploy lidar in volume in Level 3 to 4 ADAS [advanced driver-assistance systems] and Level 5 autonomy, sensors need to be in the hundreds of dollars range,” Asghari added. “For consumer applications, prices need to be even lower, like single-digit dollars for phones.”
Market leader Velodyne Lidar Inc. has partnered with Sendai Nikon Corp. to mass-produce cheaper lidar sensors.
Asghari didn’t specify prices for SiLC’s systems, but he said he expects them to be competitive.