Lidar melee

In 2019, two major events occurred in the field of autonomous driving lidar: first, due to the fierce competition in the lidar market, Israel's low-cost lidar startup Oryx Vision announced its closure; second, in order to solve the cost and performance problems faced by lidar, Huawei confirmed that it would develop its own lidar.

Oryx Vision and Huawei's completely different attitudes towards LiDAR largely reflect the awkward situation of this technology: LiDAR is indispensable for autonomous driving, especially high-level autonomous driving, and has broad development prospects, but due to the high technical difficulty and slow commercial realization, not every player can make it to the end. In particular, as market competition continues to intensify, it is not ruled out that more LiDAR companies with poor financing capabilities and unclear prospects for technology commercialization will exit the market.

How does LiDAR help realize autonomous driving?

The industry has been debating for years whether LiDAR is needed for autonomous driving, but now the answer is becoming clear: LiDAR must be used for autonomous driving above Level 3. Why do we say that?

A very important reason is that LiDAR has high detection accuracy, wide detection range, and strong stability. It can perform real-time 3D modeling of the surrounding environment, provide decision-makers with road data such as the position, distance, speed, direction of movement, and flow of motor vehicles, non-motor vehicles, and pedestrians, and help self-driving cars accurately identify the attributes of targets on the road. L3 and above autonomous driving requires the system to share or even fully control the vehicle with human drivers, so it must have environmental perception capabilities comparable to or even beyond human drivers. At present, relying solely on millimeter-wave radars and cameras cannot meet this requirement.

Image source: Hesai Technology

Although the camera has a relatively good performance in identifying environmental information such as roads, vehicles, pedestrians, and traffic lights, the recognition accuracy is easily affected in harsh environments such as strong light, rain, snow, and fog. The fatal accident that happened to Tesla before is a good example. Although the millimeter-wave radar has strong anti-interference ability and can be used normally even in bad weather such as rain, snow, haze, sand, and dust storms, the detection distance of the millimeter-wave radar will be directly restricted by the frequency band loss, and it is impossible to accurately model all surrounding obstacles. It can be said that each has its own advantages and disadvantages. Even if these two sensors are integrated, they cannot meet the precise detection needs of high-level autonomous driving vehicles.

Of course, this does not mean that LiDAR has no flaws - LiDAR cannot be used normally in bad weather such as rain, snow, smog, and sandstorms. From this point of view, adding LiDAR cannot completely solve the perception problem of autonomous driving. It can only be said that it can complement other sensors to enhance the perception ability of autonomous vehicles and provide vehicles with a kind of perception redundancy.

The importance of LiDAR to autonomous driving goes far beyond this. LiDAR can extract road features in real time and match them with the relative spatial coordinate information of landmarks in high-precision maps, which can also help vehicles perform high-precision positioning, which is particularly critical for autonomous vehicles. Because only when autonomous vehicles know exactly where they are can they obtain a safer driving route.

If the LiDAR is installed at the roadside, it can form a roadside perception system together with cameras, millimeter-wave radars, etc., to help self-driving cars collect road environment information from another perspective, and can also achieve perception redundancy, improve the operational safety of self-driving cars, especially for dangerous scenes such as ghosts. It is a good solution. At present, LiDAR startups such as Innovusion and LeiShen Intelligence are already exploring this aspect. LeiShen Intelligence's 32-line LiDAR and vision-fused deep learning algorithm target recognition, classification and tracking solution are reportedly used in Yutong Bus's "Smart Island 5G Smart Bus" project.

Precisely based on the importance of autonomous driving, according to relevant forecast data, as L3+ autonomous driving cars gradually come into the market in the next few years, the LiDAR market will also usher in rapid development. By 2026, the market size of automotive LiDAR is expected to exceed 10 billion yuan. If applications in specific scenarios are included, its final market value will even far exceed this estimate.

Who will win among the competition among multiple forces?

If we count from the time when LiDAR was first used in the DARPA Challenge in 2004, the application of this technology in the automotive field has a history of 15 years. After more than a decade of development, the automotive LiDAR market has evolved from the initial situation where Velodyne was the only one to a situation where multiple forces coexist.

For a long time, Velodyne has dominated the automotive lidar market. Although the company's lidar has always been known for its high price and long delivery cycle, many companies have chosen Velodyne's multi-line lidar as the main sensor when conducting autonomous driving tests. For example, when Google was developing its autonomous driving project in the early days, it purchased Velodyne's 64-line lidar. It is reported that this "family bucket" costs $75,000, which is obviously not a price that the mass-produced car market can afford.

Because of this, in 2015, Velodyne launched a relatively economical new laser radar product, Puck (formerly known as VLP-16). Compared with HDL-64, Puck is indeed much cheaper, but it still costs $8,000 per unit. In 2017, Velodyne released another multi-line laser radar, Ultra Puck (formerly known as VLP-32C), which is priced between 16 and 64 lines, about $40,000 per unit. It can be said that Velodyne is indeed actively seeking change, but for mass-produced cars, such a price is still not "beautiful".

What's more, HDL-64, Ultra Puck and Puck are all traditional mechanical rotating LiDARs, which are not only difficult to mass produce, but also not easy to install. They are not the best sensor choices for mass-produced autonomous vehicles. Therefore, based on the existing multi-line LiDAR products, Velodyne is still seeking change, and finally launched its first solid-state LiDAR Velarray at the 2018 CES. It is reported that the resolution of Velarray is equivalent to 300-350 lines, which is suitable for both day and night, and can be easily embedded behind the windshield, in the bumper and other corners of the vehicle, fully realizing the perfect integration of LiDAR and vehicle. At present, the product has received orders from well-known foreign car companies.

Image source: Valeo

While Velodyne was busy seeking change, a large number of new LiDAR players began to emerge in the market in the past few years. For example, Valeo, in cooperation with Ibeo, launched the automotive industry's first and currently only mass-produced automotive-grade LiDAR, SCALA®, in November 2017. The LiDAR was installed on the new Audi A8 as soon as it was launched, and then on Audi A6, Audi A7, Audi Q7, Audi Q8 and other models to help vehicles collect road information and improve driving safety. So far, the delivery volume of SCALA® has exceeded 100,000 units.

In addition, auto parts companies such as Bosch, Aptiv, ZF, and Continental are also deploying LiDAR. Bosch mainly enters the field by investing in LiDAR startups. In February 2017, Bosch's investment company invested in a LiDAR company called TetraVue. It is reported that the company's ultra-high-resolution 3D LiDAR data and images will help automatic and highly automated driving vehicles to effectively identify obstacles. In September 2018, the investment company completed a follow-up investment in ABAX Sensing to support the latter's development of all-solid-state chip LiDAR for ADAS and autonomous driving vehicles. According to the relevant person in charge of Bosch, Bosch is also conducting LiDAR research and development, which is expected to be put on the market in 2022 .

The same is true for Aptiv. In 2015, Aptiv invested in a leading company in 3D LiDAR sensors.Quanergy, the two parties plan to jointly develop a series of low-cost, high-performance solid-state LiDAR products. Then at CES 2016, Quanergy released its first solid-state LiDAR S3 for self-driving cars. In August 2017, Aptiv acquired a minority stake in Israel's Innoviz Technologies. In September of the following year, Aptiv signed a commercial cooperation agreement with LeddarTech, a Canadian company that develops solid-state LiDAR technology, announcing the joint development of a low-cost angular LiDAR solution. At the same time, Aptiv also made a small investment in LeddarTech.

ZF's layout of LiDAR is mainly realized through Ibeo. In 2016, ZF announced the acquisition of 40% of Ibeo's shares, in order to develop more advanced driver assistance systems and achieve highly automated driving based on the latter's research in the field of LiDAR. In May 2019, ZF, Ibeo and ams reached a cooperation to jointly develop solid-state LiDAR technology for vehicles to ensure that the technology can be put into use quickly and safely before 2021.

Continental's strategy is similar to ZF's. In 2016, Continental announced the acquisition of LiDAR companyAdvanced Scientific After nearly three years of research and development, Continental has developed a solid-state lidar HFL110, and is pushing its samples to the commercial vehicle market to help achieve high-level autonomous driving, according to foreign media reports.

As for technology companies, Huawei, Intel, and Waymo are the representatives. Huawei is still in the early stages of its layout. On October 22, 2019, at the World Intelligent Connected Vehicle Conference Summit Forum, Huawei's rotating chairman Xu Zhijun clearly stated that Huawei will use its optoelectronic technology to develop LiDAR to solve the cost and performance problems faced by LiDAR.