Even Musk can’t stop it! On the eve of mass production of L3 autonomous driving, LiDAR becomes standard for smart cars

The mass production of LiDAR has reached the most critical stage.

After more than five years of technological accumulation and market verification, this emerging automotive sensor will see a market explosion in the next two years.

Five years ago, Velodyne was the only company in the global automotive lidar field.

Three years ago, numerous LiDAR start-ups sprang up like mushrooms after a spring rain, presenting a flourishing scene. After a round of survival of the fittest, a number of high-quality companies and products remained.

From 2019 to date, with the entry of industry giants such as Bosch, DJI, and Huawei, competition in the LiDAR industry has become more intense and more dynamic.

In 2017, the world's first mass-produced L3 autonomous driving car, the new Audi A8, was launched, and the world's first pre-installed mass-produced LiDAR, Valeo's first-generation Scala, was also launched.

After that, the curtain slowly opened on mass production of front-end equipment in the entire LiDAR industry.

As 2020 is coming to an end, more than a dozen car manufacturers around the world have confirmed that they will equip their mass-produced models with LiDAR, including:

• 
  

• International top car manufacturers Mercedes-Benz, Audi, BMW, Volvo, Honda;

• 
  

• Among domestic brands, Great Wall, Changan, and BAIC New Energy;

• 
  

• Xpeng Motors and Lucid Motors are new car-making forces;

• 
  

• As well as leading commercial vehicle manufacturers such as FAW Jiefang, Shaanxi Automobile Heavy Truck, etc.

In addition, among the new domestic car-making forces that are popular in the capital market this year, NIO, Ideal, and Xpeng are also conducting on-vehicle testing of laser radars, and it is only a matter of time before they are installed on mass-produced cars.

According to incomplete statistics from Auto Heart, the total amount of confirmed LiDAR pre-installed mass production orders worldwide has exceeded RMB 20 billion, and the mass production scale that these OEMs will be able to leverage in the future will reach millions or even tens of millions of vehicles.

Such a large-scale market has created broad development space for lidar companies.

Autonomous driving is destined to change the global transportation system, and lidar will be one of the "nuclear weapons" in this change.

1. No matter what Musk thinks, the trend of laser radar front-end mass production is irreversible

There is still debate in the industry about whether lidar is necessary for autonomous driving.

Tesla CEO Elon Musk is the number one bear on LiDAR.

Musk is full of confidence in his own camera and vision technology, claiming that high-level autonomous driving can be achieved without relying on lidar. He also threatened that even if lidar is free, Tesla will not use it.

Then at the Guangzhou Auto Show in November, Xpeng Motors announced that it would add lidar sensors to its next-generation autonomous driving system and planned to mass-produce models equipped with lidar in 2021.

When the news reached Musk on the other side of the ocean, he indeed mocked him. But He Xiaopeng, the helmsman of Xpeng Motors, did not back down and directly responded: he would beat Tesla in the field of autonomous driving.

The "cross-ocean confrontation" between the two giants is just a microcosm of the industry's long-term debate on whether to use lidar.

If we look at it purely from a technical perspective, cameras and visual technology do have shortcomings, such as parallax illusions in cameras and unstable performance when ambient light changes.

In many application scenarios, the defects of visual technology itself are difficult to make up for. The many accidents that have occurred in Tesla Autopilot in the past have exposed the limitations of visual perception time and time again.

An engineer in charge of hardware at a well-known domestic autonomous driving company told us:

“A pure vision solution can solve 90% of the problems, and a pure lidar solution can also solve 90% of the problems. However, each of the two 90% has its own areas of expertise, and together they can solve more than 97% of the problems.”

If supported by sensors such as millimeter-wave radar, many perception problems can be better solved.

So in essence, adding lidar to the perception solution is to increase the reliability of the system.

You know, even Mobileye, the pioneer in the field of visual ADAS, cannot reject lidar.

Recently, Mobileye has reached a cooperation with Luminar to equip its autonomous driving road test fleet with lidar.

On the one hand, cameras and visual perception technology themselves have flaws.

On the other hand, LiDAR is gradually becoming ready for large-scale pre-installation mass production.

After several years of exploration by many manufacturers, some lidar products are already capable of withstanding the stringent automotive-grade standards.

In terms of cost, the price of lidar is gradually being reduced to around US$1,000 per piece.

This reduces a lot of pressure on car manufacturers to get their cars on board.

More and more car manufacturers are vigorously promoting the installation of L2+ and L3 autonomous driving functions in vehicles.

Assisted driving functions such as NOP and NGP will become the focus of competition among many automakers in the future. In particular, if these functions are to be enabled on ordinary urban roads, there must be sufficient perception redundancy, which will also force automakers to promote the use of lidar in vehicles.

At present, a considerable number of car companies on the market have made it clear that they will install 1 to 5 laser radars on their vehicles.

According to Xiaopeng Motors’ autonomous driving engineer:

“Its autonomous driving system will still be based on visual perception and deep learning, and lidar provides an extra layer of redundancy.

In future car models, LiDAR may not necessarily be used to achieve 360° coverage. Mainly for more critical locations, 1-2 LiDARs will most likely be used.

With the support of LiDAR, the application scenarios of its newly launched NGP high-speed autonomous navigation driving system will be expanded, and its safety and stability will be further improved. "

He Xiaopeng also said in an interview with the media, "In the first few years, we will place safety at a very high level. We would rather have hardware and software redundancy than fail to ensure safety."

Of course, who will pay for the additional cost of lidar sensors ultimately depends on the decision made by the car manufacturers.

Taking all the above factors into consideration:

• 
  

• Camera and vision technologies have defects, and LiDAR is needed to provide an extra layer of redundancy. Therefore, the mass production of LiDAR is necessary.

• 
  

• In terms of technological maturity and price-friendliness, LiDAR has already met the requirements for mass production.

• 
  

• Due to the need to launch higher-level autonomous driving functions, car manufacturers now have clear procurement and mass production plans.

Taking several factors into consideration, mass production of front-end LiDAR has become an inevitable trend.

No matter what Musk thinks, the only thing he can decide is whether to use lidar on Tesla cars, but he cannot stop the direction of the entire market trend.

2. The wave of LiDAR pre-installation

1. 2017-2018: Scala is the only language that stands out

Since its launch in 2017 with the new Audi A8, Valeo's first-generation Scala LiDAR has shipped more than 100,000 units worldwide by September 2019.

However, after Valeo Scala, there is still no second lidar product on the market that has been mass-produced in consumer-grade vehicles.

This situation has arisen because:

• 
  

• Early laser radar products on the market basically adopted mechanical rotation solutions. There were many moving parts inside the laser radar, which made it difficult to meet automotive-grade requirements.

• 
  

• The shape and size of the mechanical rotating laser radar are also extremely unfriendly to pre-installed mass production;

• 
  

• In addition, a laser radar cost tens of thousands of dollars at that time.

Due to several reasons, many car manufacturers with demand are reluctant to do so.

At the same time, emerging lidar companies that adopt solid-state and semi-solid-state technology routes have emerged on the market, and traditional lidar manufacturers are also developing lidar products that better meet automotive standards and have lower costs.

However, between 2017 and 2018, these technologies and products were not mature enough and had not yet reached the level of mass production.

But even despite numerous difficulties, there are still car companies willing to make advance arrangements.

In April 2018, BMW announced that it would become the world's second automaker after Audi to install LiDAR sensors on mass-produced cars. And the product to be installed on the car was a solid-state LiDAR product.

At that time, BMW announced that it would use InnovizOne, the first-generation product of Israeli lidar startup Innoviz, in its future self-driving cars.

They also brought in parts supplier Magna and plan to mass-produce this L3 autonomous driving car in 2021.

However, foreign media have recently reported that BMW will postpone the mass production of L3 autonomous driving, which means that InnovizOne will not be installed in cars in 2021 as originally planned.

There are also reports that the cooperation between BMW and Innoviz will shift to the second-generation product InnovizTwo.

Compared with its predecessor, InnovizTwo reduces costs by more than 70% and also improves performance.

InnovizTwo redesigned the MEMS mechanism used in the beam steering mechanism and detector to achieve lower cost.

In terms of performance, the official estimated detection range of InnovizTwo for targets with a reflectivity of 10% will exceed 200 meters.

Innoviz plans to begin shipping InnovizTwo samples to customers in the third quarter of 2021, with plans to have it available for mass deployment in the fall of 2023.

2023 may be the time when BMW plans to mass-produce L3.

2. 2019: Established LiDAR manufacturers gain favor