How to build automotive-grade LiDAR?

In the global autonomous driving landscape, China and the United States have always been at the forefront. In the various sub-sectors such as Robo-Taxi, Robo-Truck, and Robo-Delivery, you can find corresponding players in China and the United States, and there are even more Chinese players. For example, in the Robo-Taxi sector, the United States has Waymo, Cruise, Uber ATG, etc., and China has Bai du Apollo, Pony, AutoX, WeRide, DiDi, DeepRoute , etc.

In addition to more players in the scene, there are more players in the autonomous driving industry chain. For example, in the L4 level autonomous driving essential sensor - LiDAR , Chinese players are catching up.

Velodyne is said to be the originator of lidar , but China's Hesai has won more autonomous driving orders. The patent mutual licensing cooperation reached by the two parties will help Hesai win orders from American autonomous driving companies.

According to Che Zhijun, the leading autonomous driving companies in the United States, except Waymo (which has always developed all its autonomous driving software and hardware in-house), include Cruise, a General Motors-owned autonomous driving company with a total financing of up to US$8.25 billion (currently the world's largest), Nuro, which set a record of US$940 million in a single round of financing for a startup company, and Zoox, which was acquired by Amazon for a record US$1.2 billion. All of these leading autonomous driving companies in various fields have adopted China's Hesai's lidar.

The US super unicorn autonomous driving companies with a cumulative financing amount of nearly US$10 billion have all adopted Hesai's lidar. In fact, many of China's leading RoboTaxi autonomous driving companies, such as Baidu Apollo, Pony, WeRide, AutoX, etc., also use Hesai's lidar.

Che Zhijun found that among the top 15 RoboTaxi players in terms of test mileage announced by the California Department of Transportation in 2019 , more than 10 of them used Hesai LiDAR as their main radar. Does this mean that Chinese LiDAR represented by Hesai has the opportunity to be at the forefront of the global industry in LiDAR, a new sensor component industry in the autonomous driving industry, before the mass production of autonomous driving vehicles?

New requirements for LiDAR in L4 autonomous driving

Since 2009, Google X Lab has started to study driverless cars, which has opened a new round of autonomous driving research. Especially from 2014 to 2016, a large number of autonomous driving companies have emerged in China and the United States, working in various sub-scenarios. Two major technical routes have emerged: one is computer vision-based, and the other is lidar-based.

After more than 10 years of development and verification, LiDAR has been proven to be an essential sensor for achieving L4 autonomous driving and above. Companies around the world engaged in L4 autonomous driving, including startups and large companies, have adopted LiDAR, and most of them purchase LiDAR rather than develop it themselves.

At present, the software and hardware platforms of various autonomous driving companies have undergone several iterations, and with the evolution of technology, higher requirements have been placed on lidar, the most important sensor for autonomous driving.

For example, reliability, stability, recognition accuracy, point cloud quality, and even automotive grade standards are being considered to prepare for the upcoming mass production. Currently, L4 autonomous driving is in the intense testing phase before mass production. Any autonomous driving player hopes that their technology is more advanced than their competitors, which requires better software architecture, better hardware platform, and better lidar.

This also means that newly established LiDAR companies have lost the opportunity to grow together with their customers, and more mature players will have a clear advantage. Some autonomous driving companies may choose to develop their own LiDAR, such as Waymo, but there is only one Waymo in the world, and because Google is willing to provide continuous cash support, it was able to raise a record-breaking $3 billion in the first round of financing.

But input and output are not necessarily proportional. For example, in early 2020, Cruise carried out a certain scale of layoffs and laid off the LiDAR technical team, which means that it gave up the route of self-developed LiDAR because there are more cost-effective and mature options on the market. Currently, Cruise uses Hesai's LiDAR.

As autonomous driving technology is in the intense testing phase before mass production, the required LiDAR has higher requirements in terms of reliability, stability, performance, etc., and even requires the demonstration of automotive-grade products. Just yesterday, September 1, Hesai released its new flagship product Pandar 128, a main LiDAR designed with automotive-grade requirements in mind.

How to Build Automotive-Grade LiDAR

At present, only Audi A8 has used Ibeo's low-line-speed lidar, which is also the only automotive-grade lidar to date, helping Audi A8 achieve L3 level autonomous driving. However, this function has not been activated in China, the United States and other places because it is not allowed by regulations.

According to the plans of some car companies, in the next three years, laser radars that can achieve autonomous driving will be gradually made standard. For example, Volvo announced that it will mass-produce new models with laser radars as standard in 2022, and hopes to realize the autonomous driving function Highway  Pilot on highway scenarios. According to Volvo's definition of Highway Pilot - creating an unmanned driving system for specific highways , this can be considered as L4 level autonomous driving.

As car companies are planning L4 autonomous driving, technology companies and startups must keep up with the pace of car companies in terms of their demand for automotive-grade LiDAR. The automotive-grade requirements are much higher than those for consumer electronics, requiring component products to adapt to automotive-grade requirements from design to manufacturing to verification to production and large-scale application, and a lot of verification work is required.

On the one hand, there is a demand for higher performance, and on the other hand, there is a demand for automotive-grade standards. The combination of the two has put forward higher requirements for LiDAR manufacturers. From 64 lines to 128 lines, the technical difficulty is an increase in technical level, including power consumption, heat dissipation requirements, anti-interference difficulty, etc., and it must also meet automotive-grade requirements. The challenge of Pandar 128 is even greater.

As a company that has been supplying main lidars to leading autonomous driving companies in the market for a long time, Hesai has seen the needs of its customers and the development trends of the industry reflected behind these needs. It is also facing huge market competition pressure. Pandar 128 is designed to meet market demand.

In terms of performance, Pandar 128 has improved the line speed and point frequency compared to the previous Pandar 40/64 products, and balanced the vertical and horizontal resolution, making the point cloud more refined; in terms of size, previous lidars were like a large weight on the top of the vehicle, which made it difficult for many users to accept. The future lidars must be embedded in the car, and Pandar 128 is working hard in this direction, with a smaller size and lighter weight.

Autonomous driving practitioners once worried that laser radars would interfere with each other. To solve this problem, each laser beam emitted by Pandar 128 has its own key, and each laser radar can only receive and identify the light it emits (using reflected light to identify the shape of objects), thus avoiding interference between laser radars.

As mentioned earlier, the autonomous driving industry's demand for LiDAR is aimed at automotive-grade. In order to meet user needs, Pandar 128 has been designed to meet automotive-grade requirements from the beginning. During the period, DV conducted more than 50 tests, including electrical, mechanical , climate, sealing, material and battery compatibility tests, and other automotive-grade tests. The designed service life under typical working conditions is more than 30,000 hours.

In addition to meeting higher performance, lidar, as a core component of autonomous driving, also needs to pass the ISO 26262 functional safety standard and the ISO21434 network security standard. Hesai's automotive certification will not take too long, thus meeting the requirements of automakers for mass production.

Only after obtaining the automotive-grade certification can we be considered to have obtained the ticket to mass production. As an indispensable component for autonomous driving, LiDAR needs to go through many steps and pitfalls from the current testing requirements before the large-scale application of L4 autonomous driving to the future mass production application. Any LiDAR company, whether it uses mechanical, MEMS or pure solid-state, needs to go through a similar process. Successful experience will avoid some pitfalls.

With the advent of autonomous driving, the automotive industry will have a huge demand for lidar in the future. Some institutions predict that by 2035, China will have 8.6 million autonomous vehicles, of which 3.4 million will be fully driverless and 5.2 million will have assisted driving functions. If each car requires a high-resolution image-level lidar, the Chinese market alone will reach a market size of hundreds of billions of US dollars, which is large enough to accommodate many players.

For Chinese lidar companies, this is a rare new opportunity to participate in the transformation of the automotive industry. As the process of automotive intelligence, networking and automation deepens, more and more new technologies will be installed in vehicles and bring huge business opportunities. This is the opportunity given to us by the times.