LiDAR acts as an eye with a 360° view, and many autonomous-vehicle developers have been using it to build a three-dimensional map of the environment
LiDAR acts as an eye with a 360° view, and many autonomous-vehicle developers have been using it to build a three-dimensional map of the environment around the vehicle. The road to mass adoption, however, is paved with new challenges and added pressure.
Automotive LiDAR traces its origins back to the U.S. Defense Advanced Research Projects Agency (DARPA) Grand Challenge, an AV race to foster development of fully autonomous ground vehicles. LiDAR was introduced in the second edition of the race, in 2005. Two years later, five of the six vehicles finishing the race had roof-mounted LiDAR units. Since then, LiDAR innovation has moved fast, and automotive applications are expected to be the main drivers in the next five years, according to market research firm Yole Développement(Lyon, France).
“The LiDAR market for ADAS is set to achieve an annual growth of 114%, from US$19 million in 2019 to US$1.7 billion in 2025,” said Alexis Debray, a technology and market analyst at Yole. Expectations are high, but the LiDAR market currently faces headwinds and calls for bold moves.
Prices drop, but volumes are low
Historically, LiDAR systems have been too expensive to mass-produce for consumer vehicles. The trend is now reversing: Different LiDAR manufacturers have defined aggressive strategies, and the price drop over the past three years has been massive.
Last year, Luminar announced LiDAR-based solutions for under US$1,000. Velodyne, which came up with the first real-time 3D LiDAR in 2005, unveiled plans to reach an average unit price of US$600 by 2024, down from US$17,900 in 2017. And Chinese LiDAR manufacturers, whose unit prices are usually one-fifth those of other companies, are already fielding units priced below $1,000 and are gaining market share.
But a price drop does not necessarily imply a volume increase. So far, volumes have not grown significantly, and mass adoption has not yet occurred. “LiDAR must answer a need,” said Debray. “In the industrial market, including manufacturing and logistics, there is a clear trend toward automation, and LiDAR is playing a key role. In automotive, US$600 remains expensive for a car sensor in comparison with ADAS cameras, for which the average selling price is US$80. Therefore, we are now hearing about US$100 LiDAR for short-range automotive applications.”
Although Velodyne’s plan comes with some risks, said Debray, “things needed to change, and price reduction is a necessity for the automotive and the industrial markets.”
Synergies arise, but pressure mounts
Based on its tracking of several partnerships between LiDAR manufacturers and carmakers, Yole predicts that 3.2% of personal cars will integrate LiDAR units by 2025.
Audi and Valeo opened the way in 2017 with the integration of Valeo’s SCALA LiDAR into the Audi A8. Valeo’s LiDAR is now implemented in other car models such as the Audi A6, A7, Q7, and Q8. Valeo is also collaborating with Mercedes, and the recent release of the S-Class uses Valeo’s second-generation LiDAR, SCALA 2.
Subsequent partnerships have included Innoviz with BMW, Luminar with Volvo, Velodyne with Hyundai, and Ibeo with Great Wall Motors. “This is the beginning, and more partnerships are expected in the future,” said Pierrick Boulay, a technology and market analyst at Yole.
Consolidation is underway in the industry. Besides partnerships between manufacturers, Tier 1 suppliers, and OEMs, some acquisitions of LiDAR companies by Tier 1 suppliers have occurred. Continental acquired ASCar, a division of Advanced Scientific Concepts, in 2016, and self-driving car startup Aurora Innovation acquired Blackmore in 2019. Acquisitions of component companies also took place. In 2018, ON Semiconductor bought SensL Technologies, while Luminar bought Black Forest Engineering.
But Debray offered a caveat, noting that “there are too many uncertainties on the winning technologies and the size of the future market to trigger many acquisitions.”
The global automotive industry indeed faces mounting pressure. The introduction of connected, autonomous, shared, and electric (CASE) technologies has brought a new set of options for automotive companies. Each of them needs to determine the right direction to prioritize to align with customers’ preferences and requirements. “Two main trends are currently driving the automotive industry: electrification and autonomy,” said Boulay. Since 2009, European Union legislation has set mandatory emission targets for new cars. CO2 emissions, on average, must drop by 15% by 2025 and by 37.5% by 2030. These reductions would translate into a target value of 95 g/km in 2021 and 81 g/km in 2025. “Electrification requires time and a lot of R&D, and this could delay or reduce autonomous development projects for several OEMs,” said Boulay.
Caution is also recommended because LiDAR is not the only sensor technology that is being refined and improved. Radar is moving to ultra-high resolution, and thermal cameras could make their entry, Boulay noted. “LiDAR is still expensive, more than 10 times the cost of a camera or a radar, and if OEMs can find cheaper solutions with similar performance to a LiDAR, they will adopt it.”
Some automotive OEMs are not using LiDAR at all and are instead betting on a combination of sensor and artificial intelligence-based computing to achieve high automation levels. The most high-profile of these is Tesla’s whose CEO, Elon Musk, notably called LiDAR “a fool’s errand” at the company’s first Autonomy Day event in April 2019. “Anyone relying on LiDAR is doomed,” said Musk. “Expensive sensors that are unnecessary.”
Tesla’s approach is worth keeping an eye on, acknowledged Boulay. “They are using a pseudo-LiDAR approach based on the cameras placed around the vehicle with the goal to produce an accurate 3D image of the surroundings of the car.” AI computing is used to analyze lane markings and to predict the behavior of other vehicles. “It analyzes not only images, but also the individual pixels within the image to predict the depth of each pixel,” added Boulay. Toyota is investigating similar approaches.
No doubt Musk is an influential thinker whose words can have an impact on the industry. The Tesla Model S was the first car to offer autonomous features. Boulay called Tesla’s Autopilot “a great step toward autonomous vehicles” but added that “since its release, many accidents have occurred resulting from a misuse of this feature, and drivers have relied too much on this ADAS system.”
The 2018 death of pedestrian Elaine Herzberg, who was struck by an Uber self-driving vehicle, “really opened the eyes of OEMs, Tier 1s, and Tier 2s,” said Boulay; they realized that achieving AVs would be much more complex than anticipated and that it would take “decades to reach true autonomy.”
Debray noted that Musk is not opposed to LiDAR in all applications: SpaceX, another Musk company, is relying on the DragonEye LiDAR for its Dragon and Crew Dragon spacecraft. His question to Musk is: “Why can’t earth people enjoy the same safety as space people?”
Market shifts, but use cases remain narrow
The LiDAR market is dominated by two traditional companies in topography, Trimble, and Hexagon. With specific technologies and business models, these well-installed players face limited competition from newcomers. Debray believes they may even acquire LiDAR companies when the situation becomes clearer.
Germany-based Sick is well-established in LiDAR systems for factory automation. “They are probably under more pressure from newcomers, but they still have products well-suited for many industrial applications,” said Debray. “New LiDAR companies have to challenge this position with lower prices or better performance.” Asked about movers and shakers in the industry, Debray cited Ouster, a San Francisco LiDAR startup that launched out of stealth in December 2017 and has recently brought its total funding to US$140 million. Ouster’s differentiation lies in the development of industrial LiDAR based on technologies from the customer market, i.e., vertical-cavity surface-emitting laser and single-photon avalanche diodes. Also a newcomer, Valeo is starting to report LiDAR revenue.
Velodyne and Luminar have announced their intention to go public by the end of this year. “This will force these two companies to generate revenue at some point and will therefore strongly influence the LiDAR landscape,” said Debray.
Chinese LiDAR manufacturers such as Hesai, Robosense, and Surestar are also enjoying continuous growth and expanding their product portfolio. They not only benefit from lower LiDAR unit prices, but also from the numerous public projects for AV.
In parallel, many AV developers have expressed strong interest in LiDAR technologies. Mobileye, an Intel company, is crafting its own LiDAR tech, and Waymo is currently the only company making its LiDAR commercially available.
Overall, the LiDAR landscape is shifting. Unit prices are decreasing and will continue to decrease in the coming years. This should foster the adoption of LiDAR in new automotive applications.
But diversification must be encouraged, and new LiDAR models with more dedicated specifications are needed, the Yole analyst said.
“The development of LiDAR is clearly linked to the use cases that OEMs want to implement,” said Boulay. “There is not one LiDAR for multiple applications. It is the application or use case that will give the specifications and the position of the LiDAR in the vehicle.” For highway driving, he illustrated, a long-range LiDAR with a narrow field of view and a central position in the vehicle will be needed while for parking or city applications, a short-range LiDAR with a large field of view and placed on the corners of the vehicle will be required.
This article was first published on EE Times Europe