One LiDAR to Rule Them All: Solid-State LiDAR Emerges as the Flavor of Choice for Self-Driving Cars
Solid-state LiDAR is cheap, robust, and compact in size — this is why the device is seen as the LiDAR of choice for future high-volume production of level 3 and level 4 cars.
LiDAR sensors, or light detection and ranging sensors, play a key role in detecting surrounding objects and feeding the information to the car's artificial intelligence (AI) system for self-driving capabilities. Several different flavors of LiDARs exist, including mechanical-mirror, 3D flash, and optical phase array (OPA), as well as solid-state. The latter has emerged as the most popular LiDAR type thanks to falling per-unit costs, packaging appeal, and performance.
According to analyst firm Frost & Sullivan, over 90% of all driverless cars in development have solid-state LiDARs.
Solid-state LiDAR meets the technical requirements for long-distance scanning, which is expected to eventually have a range up to 200 meters compared to the 120-meter range today's solid-state devices offer. Since it has no moving parts, it is relatively robust and long-lasting. Besides good tech specs, solid state is compact and can be discreetly designed into the car's exterior, compared to the unsightly and more expensive $70,000 dome-shaped, rotating-mirror devices on Waymo's driverless cars.
Waymo's rotating-mirror LiDARs by Velodyne, for example, can sweep areas up to 200 meters from cars and offer high data-transfer rates. But while its form factor and price tag might meet the design needs of mobility service providers that offer driverless taxi fleets, it is probably less attractive for a private owner of a premium car who would probably prefer solid-state LiDAR's discrete form factor, Anirudh Venkitaraman, an autonomous driving analyst for Frost & Sullivan, told Driverless.
Solid-state offer the best practical packaging alternative since it is positioned within the periphery of the vehicle. It just doesn't disrupt the aesthetic appeal of a car's design.
Another important and obvious factor is cost, especially considering how cars will typically have four or more LiDARs embedded in the exterior. Solid-state LiDAR producers, which include Velodyne, Valeo, and Quanergy, are beginning to meet carmakers' demands for $250 per-unit pricing. Quanergy, for example, is readying production of its S3 at this price point, which the company is touting as the "world's first" affordable solid-state LiDAR.
If the suppliers figure out a way to bring the costs down further, as Velodyne and other suppliers have suggested, then solid-state LiDAR will be an ideal solution for achieving Level 4 and 5 automation.
Solid state's robustness is critical for safety reasons, Louay Eldada, CEO and co-founder of Quanergy, told Driverless. "When LiDAR is used as the primary sensor, then it must be solid state because it is a safety-critical application. It cannot be mechanical, because mechanical LiDAR can fail and that is unacceptable as a primary sensor."
However, as the industry is drawn to solid-state for its relative low cost, more testing is required to have a patent answer about its reliability, Anand Gopalan, CTO of Velodyne, told Driverless.
While there is a perception that a solid-state solution will also be more reliable, it's too early to reach this conclusion because there are many diverse implementations out there with varying levels of reliability. That said, we've been working with customers to identify the most reliable and low-cost LiDAR sensor that can achieve their desired performance.
However, solid-state's popularity has come at a price as suppliers face challenges keeping up with demand. Solid-state LiDAR production has recently suffered from production bottlenecks as suppliers struggle to deliver the device on time at a $250 per-unit price.
Given the nascent stages of driverless car demand, analysts also say it is too early too tell which solid state-LiDAR producer will become the largest volume supplier. Mass-volume production, for example, is not expected to begin before 2019, according to Frost & Sullivan.
No one can also say with certainty that driverless cars will rely on LiDAR for critical-sensing applications indefinitely, either. Tesla founder and CEO Elon Musk said in 2015, for example, Tesla's self-driving cars would not require LiDAR and that sonar, radar, and passive-camera systems suffice. Advances in microprocessor designs that can boost the computing performance of sensors and their software could also one day enable cameras to replace LiDAR.
OEMs will continue to mix and match different sensor configurations as well, Guillaume Devauchelle, director of innovation and development for Valeo, a leading driverless car tier-one supplier and LiDAR producer, told Driverless. "At least in the beginning, the OEMs will have different sensing solutions," Devauchelle said.
But at least in the near future, driverless cars will continue to rely on LiDAR and, at the very least, most level 3-ready cars that could see production as early as this year will have the solid-state flavor.