Lidar’s winter is quiet
Latest update time：2022-11-15
Reads：
The cold wind of autonomous driving blows from Robotaxi to lidar.
A week ago, Velodyne, the world's first listed lidar company, announced its merger with another lidar startup, Ouster, which shocked the industry.
The founder of Ouster comes from Quanergy, another well-known lidar company. It is a start-up trying to subvert the industry with "solid-state" lidar. Velodyne was once the well-deserved overlord of the industry, and its "big flowerpot" (mechanical lidar) is almost a totem of the industry.
But in the cold winter, the landlords had no food left. The two companies said that investors' enthusiasm for autonomous driving is waning, and they must work together to reduce costs and increase efficiency. After the merger, the two parties will have 355 million in cash and a market value of approximately US$400 million. Prior to this, the stock prices of the two companies had fallen by more than 80%. It is said that the actual situation of the merger is that Ouster acquired Velodyne at a low price.
In contrast, although domestic laser companies are still in the stage of losing money and making money, they are living a comfortable life because of the support of financing. Companies represented by Sagitar, Hesai and Tudatong have already tied up With major customers such as Xpeng, Ideal, and NIO, various technical routes are also blooming, and it is still unclear who will win.
So, how did Velodyne, which once dominated the industry, end up being merged and acquired? Can other
companies avoid Velodyne's pitfalls and successfully take over? What is the biggest concern in the lidar industry?
01
Overlord falls
Velodyne is a legend in the lidar industry.
In 2005, David Hall, the American audio tycoon and founder of Velodyne, learned about the driverless driving challenge launched by the U.S. Department of Defense and decided to join the effort to develop new sensors for driverless cars. He developed a 64-wire machine right away. Type LiDAR HDL-64, its performance far exceeds that of the single-line LiDAR that was originally widely used, and the effect is immediate.
Just two years later, in the new Autonomous Vehicle Challenge, a total of 6 vehicles completed the competition, 5 of which were equipped with Velodyne’s HDL-64, including the team that won the championship[1].
Velodyne 64-line lidar HDL-64
This competition became
the beginning of the industrialization of autonomous driving in the United States and even around the world. Participants later became the backbone of the autonomous driving projects of companies such as Google, Uber, and Ford. Velodyne also came close to success and began its 10-year dominance of automotive lidar.
Li Yimin, CTO of Tudatong, once said that if it were not for Hall, the development of the global autonomous driving industry would be delayed by 10 to 25 years.
Before 2017, Velodyne basically had a monopoly on high-performance automotive lidar. Even though its HDL-64 was quoted at US$80,000, it was mainly because all mechanical components required manual adjustment, which was time-consuming and labor-intensive. However, even so, it was still expensive but not marketable. , the price on the black market once reached around 2 million yuan. In 2016, Baidu invested US$75 million in Velodyne just to obtain priority delivery rights.
During that period, how many new test vehicles were added for autonomous driving each year depended on how many radars Velodyne produced.
However, during the autonomous driving entrepreneurial boom that began in 2016, the whole world became aware of the prospects of lidar, and dozens of start-ups poured in within a few years. Under saturated competition, Velodyne's dynasty declined rapidly. In the competition, the two Chinese companies that cause Velodyne the most headaches are Hesai and Sagitar Juchuang.
In 2017, Hesai Technology launched a 40-line lidar, and in the same year Sagitar Juchuang 16-line lidar was put into mass production. The two companies competed with Velodyne in the mid-to-high end and mid-to-low end respectively, competing for customers with the cost-effectiveness that Chinese companies are good at.
The following year, Velodyne's 16-line products responded with a 50% price reduction. However, Hesai and Sagitar continue to expand their shares. In 2019, Velodyne initiated a patent lawsuit, but Hesai and Sagitar continued to make rapid progress after paying heavy patent fees.
In 2020, when Velodyne reluctantly listed on the US stock market as a SPAC as the first lidar stock, the best-selling lidar in the world was actually Hesai.
Many of Hesai's customers, such as Baidu and WeRide, switched sides from Velodyne not only because of price, but also because of faster product updates and after-sales service.
Compared with Hesai, Velodyne only has a sales department in China and its service support capabilities are weak. WeRide COO Zhang Li once complained that Velodyne's lidar needs to be sent overseas for repair, and the whole process takes 1-2 months[2]
. Domestic companies can provide 24/7 nanny-style services, and even guarantee returns and exchanges within 7 days [6].
And changes in technology have made Velodyne's situation worse.
After 2020, the accelerated popularity of smart driving in passenger cars has created a demand for automotive-grade lidar, which is a much larger blue ocean market. However, traditional mechanical lidar has hundreds or thousands of internal parts and a large number of moving parts, and its reliability cannot meet vehicle regulations. Solid-state/semi-solid lidar with fewer moving parts is about to emerge.
In fact, Velodyne has already seen the trend clearly. It released a solid-state lidar prototype product in 2017 and released a "mass production version" in 2020. However, Velodyne has no experience in the production and manufacturing of solid-state lidar. The company's management fought among themselves after the listing, and founder Hall was also revealed to have neglected management [3]. To this day, Velodyne's solid-state lidar has not been put into production vehicles. .
Among automakers’ fixed orders for solid-state lidar, Velodyne’s share is only 2%
Velodyne's failure to transform solid-state lidar directly led to Hall's expulsion from the board of directors last year and a management change. At the same time, Velodyne's few customers in the automotive industry are still very "Buddhist". Their timetable for putting lidar on cars is mostly in 2023 or even later, which is far less aggressive than domestic car companies.
After going public, Velodyne fell into continuous losses. Under multiple negative circumstances, Velodyne is in decline and can only stay together to keep warm.
Velodyne's experience is actually just a microcosm. Since last year, the share prices of lidar companies listed on the U.S. stock market have fallen by more than 80%. Not only was Velodyne forced to merge with Ouster, but another company, Quanergy, plummeted 99% and was delisted from Nasdaq.
While performance on the east coast of the Pacific plummeted, mass production on the west coast of the Pacific steadily advanced.
Fortunately, the seed customers of Chinese lidar manufacturers are naturally new car-making forces.
Under Wei Xiaoli's enthusiastic smart driving competition, since 2021, the solid-state/semi-solid-state lidar of Sagitar, Tudatong, and Hesai have successively appeared on the mass-produced models of Xpeng, Ideal, and NIO. In these three pairs of cooperation, new forces have made in-depth involvement in the product definition of lidar. Among them, Weilai personally designed the circuit board of Tudatong Falcon.
Velodyne actually lost to the Community of Destiny on the other side of the ocean, which was full of crisis and fighting power.
But the industry is far from reaching a point where the outcome is decided. If the commercialization of vehicle-mounted lidar is a 100-kilometer cross-country marathon with heavy loads, those players running at the front have just passed the first checkpoint.
02
The throne is empty
What happened to Velodyne is a cruel fact: Although it entered the industry nearly 10 years earlier than many companies, it had no moat.
The reasons are of course management, business and even political, but in the final analysis they are technical: in the past few years, the main market for vehicle-mounted lidar has migrated from robotaxi to passenger cars, and the form has changed from mechanical rotation to solid-state/semi-solid state. Velodyne, which started out as a mechanical lidar, failed to successfully complete this difficult technology and product reconstruction.
But the challengers to unseat Velodyne are also treading on thin ice. Because the entire industry's technical routes are diverse and rapidly developing and switching, a technology gamble that attempts to build barriers is likely to turn into a loser's game.
This uncertainty is specifically reflected in the fact that among the key technologies of lidar - from ranging mode to laser emission, scanning, and receiving modules, almost every item has not converged to an optimal solution, but has There are many paths with their own advantages and disadvantages for enterprises to choose (du) and choose (bo).
As a result, a hundred flowers bloom in the industry: there are dozens of companies in the industry, and almost every company has come up with a solution that is different from others through a combination of technologies.
A typical example is that although the first passenger car lidars AT128, M1 and Falcon, which are the "domestic three masters" of lidar, Hesai, Sagitar and Tudatong, are on the car, their
technical duplicate detection rate is very low.
Among them, the Sagitar M1 prefers to use more mature components. It has been iterated many times to improve component integration. It has low theoretical cost and is suitable for acting as a price butcher.
Hesai AT128 uses a new VCSEL array on the light source, pursuing the semiconductorization of components and minimizing moving parts, which is beneficial to product reliability.
The Tudatong Falcon emphasizes the use of force to achieve miracles, exchanging greater volume and power (and more expensive parts) for higher performance, allowing it to see farther and with higher resolution.
Before dozens or even millions of lidar units are delivered for verification, no one knows which solution will win, or whether the three will divide the market into three levels, or other companies will bring breakthrough technologies and sweep them into the past. The hot laser radar has attracted a large number of talents from optics, optical communications, and semiconductors. This is not an industry lacking new technologies.
For lidar companies, the more definite answer is to launch more smart electric vehicles as soon as possible at a low enough cost, ensure the reliability of this precision optical equipment in complex vehicle environments, and make their solutions a reality as much as possible industry standards.
Therefore, the leading lidar companies focus on two key words:
engineering and manufacturing.
Hesai Technology CEO Li Yifan said in an interview with "Nine Chapters Smart Driving" that Hesai's CTO in charge of engineering, Xiang Shaoqing, oversees thousands of people, while he and the chief scientist each manage a hundred people. In May last year, Hesai Investment began to build its own "Maxwell" lidar super factory.
Coincidentally, Sagitar Jutron also established a joint manufacturing company "Liteng Innovation" with Luxshare Precision last week - the two are trying to take the lead in bringing the endless technology competition back to the mass production competition of precision manufacturing.
Even so, lidar is still a money-burning business in the short term. Hesai Maxwell Super Factory has invested US$200 million and has a planned annual production capacity of one million units. On September 29 this year, Hesai announced that its monthly delivery volume of automotive-grade lidar had just exceeded 10,000 units - making it the fastest among leading players.
Two days later, Tesla’s annual AI Day was held, and Musk passed the cold air to every lidar company.
03
The biggest enemy is not your peers
A fact that is easily overlooked is that the biggest enemy of lidar companies is not their competitors in the same industry, but cameras. To be more precise, they are companies that develop purely visual autonomous driving. Tesla is the talker in this camp. .
In the past few years, Musk has repeatedly dissed lidar, believing that the latter is a "crutch" for autonomous driving and that anyone who relies on lidar will fail. But all along, the attitude of most practitioners towards lidar has been
"you spray yours, I'll use mine
. " This is because purely visual autonomous driving without lidar relies heavily on deep learning and once had major flaws in environmental perception:
On the one hand, the camera itself is not an all-weather sensor, and it is difficult to work normally in rain, snow, fog and at night; on the other hand, in the previous visual algorithm framework, the objects captured by the camera must be recognized before they can be considered to exist by the system. This results in purely visual autonomous driving being extremely unstable when dealing with untrained obstacles and stationary objects, often missing or misdetecting them.
Lidar can detect obstacles through accurate ranging without training, providing guarantee for autonomous driving.
Therefore, the previous mainstream view in the intelligent driving industry was that a multi-sensor fusion perception system should be built so that the advantages of cameras and lidar can complement each other. However, the hardware advantages of lidar are gradually being equalized by Tesla’s advantages through software algorithms.
At this year's AI Day, Tesla introduced the Occuppancy Network in detail. This algorithm can restore the three-dimensional world with high precision and real-time based on two-dimensional images. It can not only perceive the volume of objects, but also determine their movement. state. This is essentially no different than what lidar is capable of.
The picture above shows LiDAR sensing, and the picture below shows occupancy network sensing.
If cameras can become the replacement for lidar, the latter’s survival space will be in jeopardy.
Ideal is investing heavily in intelligent driving based on visual perception this year. After the network occupation was made public, Ideal took the lead in raising the ambitions of others -
CEO
Li Xiang said on Weibo that the essence of lidar is to occupy the network.
It is said that Wu Xinzhou, the person in charge of Xpeng’s intelligent driving, also privately told lidar manufacturers to prepare for transformation.
However, the industry is not all Tesla followers. By the middle of this year, Sagitar Jutron has won more than 40 lidar vehicle model approvals, and Hesai also claimed that there are millions of pre-installation approvals from OEMs. More car companies are waiting and watching: whether lidar can be used depends on whether it is cheap enough and whether its performance is stable enough.
Prior to this, the price of automotive-grade lidar has been reduced from tens of thousands of yuan to more than 3,000 yuan. However, compared with high-definition cameras that cost several hundred yuan each, the price of lidar still embarrasses most models. Lowering the price by another order of magnitude is the ardent hope of car companies for lidar, and it is also a prerequisite for large-scale vehicle installation.
A secret war between cameras and lidar has actually begun.
The strategic goal of lidar is to reduce costs. According to Li Yifan's outlook, the final price of lidar will be 2-3 times that of cameras [4]; while the strategic goal of cameras is to improve efficiency and make the visual algorithm more accurate and more confident. , as close to lidar as possible.
At present, the voice of coexistence of the two is still the mainstream, but in this race, lidar, as an emerging sensor, faces greater worries -
in history, the primary factor that determines the rise and fall of a new technology is often not its theory. The advanced nature of performance is the ability to use existing technologies and facilities
, which can be translated as
ecology
.
Compared with lidar, the camera ecosystem is complete and vast.
Image-based computer vision has always been an AI science, with the largest number of sensors (cameras), the largest amount of data, and the most intensive talent pool. This advantage has been directly inherited into the field of intelligent driving. At present, most intelligent driving functions are completed by cameras + visual algorithms, or at least mainly cameras. This brings a complete data closed loop and an extremely high evolution speed of visual algorithms.
In comparison, the ecological construction of lidar is still in its infancy, with less data and talents, and the algorithms are more immature. Even more, because the human eye is familiar with images rather than point clouds, lidar data annotation efficiency is lower than that of images, and the price is higher: annotating an image usually takes tens of seconds and costs a few cents, while The typical annotation cost of a lidar point cloud is a few minutes and starts at ten yuan [5].
The roots of these differences may be traced back to the formation of civilization or even the evolution of eyes in our ancient ancestors.
Tesla's former AI director Andrej said in a podcast recently that
the artificial world created by humans is built from the perspective of facilitating human eye perception, and vision sensors will naturally occupy a core position
. Tesla, which has figured this out, is breaking through the ceiling of visual intelligent driving every year. Just three days ago, Tesla began to push FSD V11 in North America.
This means that lidar has to fight an unequal war. Facing rapidly evolving opponents, if lidar wants to compete for a place in autonomous driving, it needs to run faster, cooperate more closely with downstream companies, and break through the impossible triangle of "cost, performance and stability" as soon as possible.
References:
[1] It Began With a Race…16 Years of Velodyne LiDAR, Velodyne
[2] Misjudgment, counterattack, overturn, lidar has been on the road for ten years, latePost
[3] LIDAR maker Velodyne boots its founder after an investigation into 'inappropriate' behavior, The Verge
[4] A leading lidar player’s thoughts on the scale from 1 to 10 - Jiuzhang Zhijia’s conversation with Hesai CEO Li Yifan, Jiuzhang Zhijia
[5] After burning hundreds of billions of dollars in ten years, driverless cars still have no way to go. LatePost
[6] 18 years of lidar battle: "Twilight of the Gods" in the West, "New King Emerging" in the East, HiEV Garlic Granule Vehicle Research Institute
京公网安备 11010802033920号