Fully autonomous vehicles are taking longer to become ubiquitous on the road, but there’s still plenty of potential for lidar automotive safety applications and other use cases. Meanwhile, the lidar market is seeing some consolidation among vendors, and lidar technology will have to share the road with other sensor technologies.
Research firm LightCounting’s recent update to its forecast for automotive lidar and 3D sensors in smartphones show the lidar market is set for growth but U.S. investors are losing interest. LightCounting CEO Vladimir Kozlov told Fierce Electronics that despite some high-profile shuttering of Level 4/5 autonomous driving technology development, lidar should see some healthy adoption on the horizon, even as Ford disclosed in October that it would shut down Argo, its multi-billion joint venture with Volkswagen, and instead focus using lidars for Level 2/3 advanced driver assistance systems (ADAS) for now with help from existing suppliers such as Intel.
Kozlov said there was a bit of bubble of high expectations for autonomy that’s burst – robot taxis are not going to take over the streets anytime soon, at least not until the next decade or even longer. But some autonomous functionality does improve safety, which means LightCounting is forecasting that sales of lidars for L2/3 ADAS applications will result in rapidly growing lidar sales in 2023-2027. Level 3 is defined as autonomous driving “under certain conditions,” but the research firm anticipates that most of these deployments will actually be Level 2.5, which always requires driver’s attention, but does improve safety significantly.
Safety is the top priority and car manufacturers are willing to make all the investments necessary to improve it, Kozlov said. “It's a bit ironic that expectations for autonomous driving are being pushed out, but the market for lidars is actually growing now.” Electrical vehicle (EV) adoption in China is one of the most active parts of the lidar market, he added.
Even for only Level 2/3 use cases, the car market is significant for lidar, but whether the profit margins are huge remains to be seen, Kozlov said, and there are different types. Frequency modulated lidars not only track position of vehicles, but speed and direction, too. “It can anticipate where the vehicle is moving, where it's going be in next half a second, or where a bicycle is moving.”
Lidar can’t go it alone
But lidar is only a piece of the automotive safety puzzle. The big companies with scale don’t necessarily make only lidar, Kozlov noted. Intel makes a lot of chips for the ADAS functionality, for example, and redundancy is critical for safety functionality. “If lidar stops working, then the car can still rely on radar. If radar stops working, they can still rely on cameras.” He said having this redundancy is a big deal for manufacturers to be able sell these technologies.
The industry is coming to the realization that lidar by itself is not an option, said Chuck Gershman, CEO and cofounder of Owl Autonomous Imaging (Owl AI). “We're going to have to supplement it.” Tesla’s move back to radar after stating it wanted a car driven with purely computer vision demonstrates that relying solely only camera technology is insufficient, he told Fierce Electronics in an interview.
Owl AI’s 3D monocular thermal camera is designed to enhance detection and location because thermal cameras inherently expand imaging viability over a much wider operating range:
Its ADAS development platform for 3D thermal computer vision is now shipping and is comprised of a thermal camera and a Nvidia Jetson AGX Orin. Owl AI CMO Wade Appelman said a thermal camera outperforms an RGB camera in both chaotic urban environments and highway environments when the vehicle is running high beams – both examples of where 3D thermal cameras can fill in sensor fusion gaps.
RGB cameras see color and fine detail but are blind in degraded visual environments and limited by headlights – most ADAS and RM cameras operate on visible band reflected light. However, with a dependence on an external light source, these cameras do not work well in low-light or no-light conditions. Thermal cameras operate in wavebands that require no illumination source whatsoever by detecting energy emitted directly from objects in the environment, and can detect objects 300m to 400m away, even in the dark.
Lidar has been somewhat oversold, Gershman said, which has led to vendor consolidation. He sees thermal cameras as filling some of the gaps of other available automotive sensing options – the goal is to build a perception suite so that data is robust under all circumstances, and in automotive, it’s unacceptable to have a gap in certain conditions, he said. “Thermal works well in degraded visual environments.”
Gershman said lidar suffers from two fundamental problems – overhyped capability and high cost. “Lidar does work at night very well, but it doesn't necessarily give you all the information as detailed as a camera system.”
Radar also has its shortcomings, Appelman added. It’s good at detecting things but not classifying them, whether it’s another car, a person, or an animal.
Lidar, meanwhile, is great at long range on a highway, but knowing what it is matters because the vehicle's going to react differently, Appelman said. Factoring in velocity also matters because it could be a pedestrian running across the street and out of the way before the car is close enough for it to matter. “You need to make decisions based on information from multiple sensors,” he said.
Less autonomy drives lidar vendor consolidation
Gathering all this information accurately is critical, but so is the cost. Camera systems and radar come in under $100 while lidars tend to be in the thousands, which is driving market consolidation. The recent merger of Ouster and Velodyne is a notable example, while lidar sensor maker Quanergy has filed for bankruptcy. Some of the assets of insolvent Ibeo Automotive were recently acquired by MicroVision, a developer of MEMS-based solid-state automotive lidar and ADAS solutions.
“The industry is inching forward on the autonomous vehicle front and not exactly leaping forward as many of us hoped,” said CEO MicroVision Sumit Sharma. “Lidar and autonomous vehicles are no doubt tied together.” Manufacturers on the path to Level 5 autonomy will come to rely on lidar, but right now the automotive market is making the transition from Level 2 to Level 3 autonomy in the coming year. With increased adoption of EVs, he said, car companies are searching for new means of differentiation. “Safety is emerging as the next source of differentiated capabilities, and lidar will play a key role in allowing OEMs to build new safety, proprietary features.”
While top-level L5 autonomy does represent huge potential, MicroVision see a multi-billion-dollar market for lidar in helping automakers deliver new safety features in L2, L3 and beyond. But as ready as the company is for L5, Sharma said the debate over which sensing technology is best must be put to rest to truly shift to the kind of innovation needed to elevate from Level 2 to Level 3 autonomy. “Clearing this hurdle will require a solution that incorporates cameras, radar, and lidar--not just one technology,” he said. “Safety is growing as a major differentiator for automotive OEMs and they have the ability now more than ever before to extend safety features more broadly and deeply through their fleets.”
Sharma said every OEM is on its own path towards autonomy with each at different stages and heading towards different end goals. “Our strategy is to be the Tier 1 partner for lidar hardware and perception software with the most OEM-friendly solution and business model.” For MicroVision, that means being flexible to meet OEM’s specific program requirements, he said, right down to the form factor of its sensor, he said. “Our ASIC allows us to build a one-in-box sensor that can be deployed in more places, like roofline mounting.”
Many OEMs are looking to accelerate their ADAS and autonomy programs, Sharma added. MicroVision is building its sensor using materials known to the OEM’s supply chain and incorporating perception software from Ibeo that has already been validated by them, he said, making it easier to get to market faster.
Sharma said there are many opportunities for lidar technology beyond the automotive industry, including commercial vehicle, industrial, robotics and even smart infrastructure production.
LightCounting’s Kozlov noted that lidar in smartphones is much easier to sell in volume as the cost per unit isn’t much more than a dollar. “Suppliers can scale down the cost quickly if volumes are significant.” He said lidar’s ability to measure distance over hundreds of meters makes it well suited for drones, which is a growing market, along with other niche opportunities for smaller players.
For automotive, lidar will remain part of the solution because it has some technological advantages, but it's not going to be the only solution, Kozlov said. “It’s going to be a mix of technologies. We'll see how the market is going play out, but we certainly feel that lidar is making good progress.”
RELATED: Lidar sensing industry sees merger, bankruptcy, stock price crunch