The battle for the autonomous driving industry is about to break out, and LiDAR has become a "gold mine" behind it

Hot money continues to pour into the autonomous driving industry, and a major war is about to break out.

Whether it is General Motors-Honda-Cruise, Toyota-Uber, Volkswagen-Ford-Argo, Hyundai Kia-Amazon-Aurora, Waymo-Renault-Nissan Alliance or Baidu, all have the hope of becoming the king in this war.

LiDAR companies are expected to become the largest "arsenal" behind these giants.

1. Finding the “gold mine” of LiDAR industry applications

Since 2007, Velodyne has provided lidar for autonomous driving projects large and small around the world.

As of March this year, Velodyne's lidar sales have exceeded 30,000 units, with sales reaching US$500 million (approximately RMB 3.45 billion).

——This sales volume exceeds the total shipments of all competitors worldwide, and the application scope covered mainly includes autonomous driving, surveying and mapping, industrial ports, logistics, security and other fields.

However, this is not the biggest gold mine for lidar.

In June this year, Valeo disclosed that it had received orders with a total value of approximately 500 million euros (approximately 3.87 billion yuan) from four major global automakers.

The life cycle of these orders is expected to extend to 2024-2025 and could ultimately result in long-term business (i.e. additional orders) worth a total of EUR 1-1.5 billion.

Currently, Valeo's 4th-line product ScaLa (first generation) is a laser radar that meets automotive-grade application standards and has been mass-produced. In 2017, it began to be installed on the A8 model with L3 autonomous driving function.

Now, LiDAR is finally installed on two types of cars:

One category is unmanned vehicles for autonomous driving tests. This market has high requirements for the number of laser radar lines and is willing to offer a higher price, but the order size is small;

One type is mass-produced cars with assisted driving and autonomous driving functions launched by automobile manufacturers. Because they are consumer-oriented, they have a large number of orders. However, these cars have very high expectations for the volume, size and reliability of LiDAR, and often require LiDAR to obtain various certifications.

For the second market, in order to ensure that the system has more safety redundancy, automakers are more willing to pay for lidar-driven  ADAS  systems.

Most mainstream car companies have stated that between 2025 and 2030, the price of lidar they can afford will be less than US$1,000.

LiDAR companies around the world are moving towards this goal.

One trend is that while many LiDAR companies are providing solutions to autonomous driving companies, they are also beginning to deploy more ADAS functions in an attempt to become a compensation option in L2/L3 autonomous driving solutions.

2. The camera-based ADAS system still has limitations

Tesla CEO Elon Musk has always regarded lidar as "heresy".

He even bluntly stated that lidar is futile for autonomous driving; lidar is an expensive and unnecessary product.

Is it really so?

In an article titled "ADAS Talk: Why LiDAR is a Must for Autonomous Driving", Zhou Yanwu, research director at Zoss, explained that camera-based ADAS and autonomous driving systems have many limitations:

First, the large field of view (FOV) angle causes the vehicle to have a very obvious blind spot.

Generally, the larger the FOV, the better, but the problem of distortion must be taken into consideration. Generally, the FOV of a CMOS  sensor lens does not exceed 76 degrees. If it exceeds 76 degrees, it is a wide-angle lens, which has obvious distortion at close range; if it exceeds 120 degrees, it is a fisheye lens, which has serious distortion at the edge of the image.

To solve the FOV problem, Volvo and Tesla chose a three-eye system.

Volvo's three-eye system has FOVs of 140 degrees, 45 degrees, and 34 degrees. Tesla's FOVs are 150 degrees, 50 degrees, and 25 degrees. But this system still cannot solve the problem of close-range blind spots.

Secondly, when the vehicle is driving at low speed, the monocular camera system is basically ineffective for suddenly appearing stationary targets or slow-moving targets (usually pedestrians).

Mobileye clearly states that pedestrian recognition only works at speeds above 50 km/h. Why is such a high speed required?

This is because machine vision is mainly used for dynamic target recognition, especially in the automotive field, giving priority to identifying dynamic targets such as vehicles, pedestrians, bicycles, electric vehicles, etc.

After understanding the above background, we can understand the role of LiDAR in autonomous driving systems and why the automotive industry always emphasizes:

Achieving a 99% recognition rate is not enough, we need to reach 99.99999...% - and lidar is the strongest guarantee for several decimal places.

In this case, the primary value of lidar is to serve as a safety redundancy in addition to cameras and millimeter-wave radar.

In 2017, Audi launched the Audi A8, which is known as the world's first L3-level autonomous driving model. This model also attracted the industry's attention to the world's first laser radar prepared for mass-produced cars - Valeo ScaLa.

Currently, the mainstream ADAS solutions in the industry use vision chips provided by Mobileye.

Although Mobileye has a high accuracy in identifying lane lines and vehicle rear ends, the system is still unable to match and identify some vehicles with strange shapes and local characteristics, such as modified vehicles and tricycles.

Millimeter-wave radar has insufficient resolution and a certain chance of missing non-metallic objects, making it impossible for the vehicle to accurately determine the position of itself and surrounding obstacles.

The Valeo ScaLa lidar equipped in the Audi A8 solves the above problems.

This can also explain to a certain extent why Mobileye, which has always focused on vision, spent tens of millions of dollars to acquire Eonite Perception, a lidar-related company, in November last year.

This is a company that specializes in developing software for 3D  mapping and tracking using lidar  . Relying on engineers from Eonite Perception, Mobileye established its laser sensor division, LiDAR.AI.

As the department is named, this acquisition will strengthen Mobileye's technology in the field of lidar and make up for Mobileye's shortcomings in the field of vision.

In the past, Mobileye's ADAS system relied on cameras, but now autonomous driving companies including car companies and technology giants prefer to use lidar.

Because lidar can accurately map the area where the car is operating in any lighting conditions, combined with cameras and radar, self-driving cars can clearly understand real-time road conditions.

3. LiDAR companies begin to focus on the development of ADAS functions

"The market is changing, and we now need to devote some of our energy to ADAS terminals." Velodyne founder David Hall told the outside world in March this year.

He believes that Velodyne LiDAR has its own advantages in the L4/L5 market, and Velodyne's production capacity is currently sufficient to meet user demand.

The ADAS market is about to usher in a new boom, where Velodyne's lidar can reap huge profits, so it is not a wise move to stick to the Level 4/5 market.

At this year's CES, Velodyne released its short-range lidar product VelaDome.

This product can cover the close-range range of the entire side of the vehicle, and is effective in detecting pedestrians or bicycles at close range/blind spots of the vehicle.

Velodyne also hopes to enter the ADAS market with its own lidar combined with software, thereby coming up with an ADAS solution as competitive as Tesla AutoPilot and GM Super Cruise.

The LiDAR software system Vella is a product launched in this context. Vella is mainly used with Velodyne's solid-state LiDAR Velarray, which can be embedded behind the car windshield or in the bumper.

Velodyne said, "Compared to the camera + millimeter wave radar system, the ADAS performance it achieves will be revolutionary."

Entering the ADAS market, Velodyne also expanded its business territory through mergers and acquisitions, considering the issues of product iteration and comprehensive performance:

In July this year, Velodyne acquired the intellectual property assets of San Francisco-based high-precision map startup Mapper.ai, and the two parties will work together to develop safer ADAS systems. Mapper.ai's high-precision map and positioning technology will accelerate the development of Vella software.

This acquisition seems similar to Mobileye's acquisition of Eonite Perception.

Velodyne’s suite of solutions now enables customers to unlock more ADAS features, including pedestrian and bicycle avoidance, lane keeping assist (LKA), automatic emergency braking (AEB), adaptive cruise control (ACC), traffic jam assist (TJA), and more.

Luminar CEO Austin Russell also has his own judgment on current lidar.

In his opinion, the self-driving taxi and truck projects led by companies such as Waymo will not be popular for a while, so the ADAS market will be more attractive in the next few years.

In June this year, Luminar launched Iris, a lidar platform that integrates hardware and software.

In order to build this solution, Luminar mobilized 60 software engineers to cooperate in research and development. The Iris platform is divided into two versions:

LiDAR solutions for advanced driver assistance systems (ADAS) costing $500 or less;

The solution for high-level autonomous driving (L4/L5) costs no more than 1,000 yuan.

Luminar's automotive lidar and software will begin mass delivery in 2022. The new lidar will be only one-third the size of current products and can be seamlessly integrated into the front grille, roof or headlights of mass-produced vehicles.

In August 2018, Cepton joined hands with Koito, Japan's largest automotive lighting company, to provide the latter with customized small-scale lidar solutions to install lidar in car lights.

Cepton's cooperation with Koito also mainly starts from ADAS and will extend to more advanced autonomous driving; the cooperation with May Mobility (low-speed campus vehicles) is aimed at L4 autonomous driving.