The 512 line era is coming. Is the automotive lidar industry ready?

What are the bottlenecks restricting the large-scale commercial use of lidar?

If it were placed two or three years ago, the opinions given by practitioners might include "expensive", "difficult to meet vehicle regulations", and "unable to be mass-produced on a large scale".

With the gradual mass production of models equipped with lidar on the market, these problems seem to be solved.

However, this does not mean that the problem of "large-scale commercialization of lidar" has been solved. Compared with tens of millions of cameras and millimeter-wave radars shipped each year, the loading volume of vehicle-mounted lidar this year has just exceeded tens of thousands.

In the words of BeiXing CEO Li Yuan, the automotive lidar market is still in its early stages.

Precisely because it is in an excellent window period, Beiwake is actively promoting the evolution of the new generation of lidar, trying to achieve large-scale commercial use of lidar through "cross-generation mass production".

01. Let’s start with 512-line lidar

The "first appearance" of the 512-line lidar was at the China Automobile Supply Chain Conference held by the China Association of Automobile Manufacturers in June this year. Li Yuan disclosed some of the core parameters of the 512-line lidar:

• 
  

• Under 10% reflectivity, the effective detection distance is 200 meters;

• 
  

• The maximum field of view can reach 120*25.6°.

Five months later, another news came out - Beiwake's 512-line lidar received orders from car companies.

A more specific development is that Beiwake has obtained the designated target of 180,000 lidar units from a certain car company and will release the SOP in the second half of 2023.

It can be said that Beiwake has single-handedly brought vehicle-mounted lidar into the 512-line era. Such rapid progress at least shows that Beixing has been accumulating strength in the automotive field for a long time.

Picture from Beixing official website

The emergence of 512-line lidar is not the product of working behind closed doors.

According to Li Yuan, it is a solution deduced based on the needs of car companies, especially the detection of small target objects at long distances.

Founded in 2015, Beiwake currently operates in 90 countries and regions around the world and is already a leader in the fields of mass transportation (including rail transit, highways, civil aviation, and shipping) and industrial sensing.

As a veteran of the lidar industry, Li Yuan firmly believes: "As long as it can solve the pain points of the industry, it must be a good product."

This logic also applies to the automotive field. "The pain point of cars is actually safety. Only by solving the safety problem of cars can lidar enter an incremental closed loop."

Currently, there are two major technical routes for autonomous driving perception solutions:

• 
  

• One type is a "pure vision" solution that relies on cameras;

• 
  

• The other type is a "fusion sensing" solution that combines multiple sensors such as cameras, lidar, and millimeter-wave radar.

Regardless of "pure vision" or "perception fusion", cameras are used as the basic perception solution. However, the problem with purely visual perception solutions is that they cannot cover all long-tail scenarios, and the perception function will be limited in some special scenarios.

For example, it is difficult for cameras to fully function when driving on roads without lighting conditions; it is also difficult for cameras to work effectively in environments with insufficient visibility such as rain and fog.

At the same time, conditions such as occlusion caused by dirty lenses and direct exposure to strong light will affect the image quality captured by the camera and affect the vehicle's perception function.

The value of lidar in the automotive field is that lidar can obtain more dimensional information, such as the distance, angle, speed, reflection intensity and other information of the target, so that the vehicle can obtain more comprehensive and powerful perception capabilities, thereby achieving safer purposes. .

The overall performance of lidar (assisted driving) currently used in L2+ systems is low. Especially for special scene detection such as long-distance small targets, the detection effect is not ideal.

Therefore, the performance of lidar needs to be further improved to meet the needs of higher-level autonomous driving.

This improvement is reflected in the lidar’s wiring harness.

Traditional lidar can be divided into 1 line, 4 lines, 8 lines, 16 lines, 32 lines or 64 lines according to the laser beam. Multiple laser beams are emitted at different angles in the vertical direction, and the target area is scanned in the horizontal direction. 3D contour detection.

The more wire beams, the higher the resolution and the greater the density of the laser point cloud - the clearer the target object detection and identification, thereby helping the autonomous driving system to obtain more accurate detection results.

To achieve such an effect, the most direct way is to upgrade the wiring harness to enhance the sensing ability of detecting targets, especially small targets, so that the autonomous driving system has more time to make decisions.

The solution proposed by BeiXing is to upgrade the current mainstream 128-line lidar to 512-line lidar, which will fundamentally improve the performance level of vehicle-mounted lidar.

"For a child 200 meters away or a long-distance cone, basically only the 512-line lidar can outline the basic situation of the person or object. In this way, the self-driving vehicle can prepare at least 2-3 seconds in advance. This 'life and death' "Three seconds" is our value," Li Yuan said.

Industry insiders say that with the arrival of 512-line lidar, high-resolution lidar will become a key means for high-end autonomous driving to converge on the corner case.

02. How can 512-line lidar reduce costs and increase efficiency?

As leading car companies actively promote high-end autonomous driving, it is imperative to put lidar on cars globally.

Since this year, NIO, Ideal and Xpeng have delivered models equipped with lidar. New and old forces such as General Motors, Volvo, Changan, Chery, BYD, Jikrypton, Avita, Zhiji, Feifan, etc. have also successively launched plans to launch models with lidar.

These models equipped with lidar will have autonomous driving capabilities in high-speed, urban, parking and other scenarios. Therefore, the requirements put forward by car companies for lidar can be said to be "both necessary and necessary":

• 
  

• With long-distance sensing capabilities, vehicles in high-speed scenarios need to detect and identify objects at least 150 meters away;

• 
  

• It has a scanning effect of at least 100 lines, so that when an object 150 meters away is encountered, enough laser point clouds can be reflected back for identification;

• 
  

• Wide viewing angle: a horizontal field of view of more than 120° to meet target detection in scenes such as intersections;

• 
  

• It can meet vehicle-level standards and support pre-installed vehicle integration and large-scale mass production.

• 
  

• After meeting the performance requirements, lidar still has to climb the price mountain before large-scale mass production.

Simply put, putting lidar on a car is not only demanding, but also challenging.

From a technical perspective, lidar must not only ensure a substantial improvement in wiring harnesses, but also balance stringent requirements and restrictions including the shape, size, weight, stability and working environment of the product itself. In terms of cost, it must also control the price. Within the range acceptable to car companies.

For example, by stacking light emitters, you get more layers that emit laser beams outward. The more layers there are, the greater the accuracy, but it also means a larger sensor size.

For another example, if scanning is used to relieve the pressure on the device end, facing a high resolution of 0.05 degrees, the scanning angle accuracy needs to be controlled within one ten thousandth, which is also a huge technical challenge.

On the cost side, technical architecture optimization, supply chain upgrades, and large-scale mass production amortization costs are considered three ways to reduce costs for lidar.

Li Yuan believes that the real cost reduction solution for lidar is to decouple performance and cost through the optimization of lidar architecture.

All of the above actually test the lidar company's ability to balance lidar system design. Regarding this point, Li Yuan will focus on sharing it at the subsequent product launch conference.

It is understood that under the guidance of the idea of ​​"decoupling performance and cost", the price of BeiWake's 512-line lidar is close to that of the current mainstream 128-line lidar in the market.

03. 512-line lidar is not the end, but the beginning

From 2005 to 2007, the lidar used by teams participating in the U.S. DARPA Challenge was priced in the tens of thousands of dollars.

One of the core reasons for Tesla's refusal to use lidar is that during the 2014-2015 period when Autopilot development began, there was no hope of mass production of lidar.

Nowadays, through the efforts of lidar companies such as Sagitar Juchuang, Hesai, Tudatong, and Beixing, the price of lidar is advancing from tens of thousands to thousands of yuan.

This means that high-end autonomous driving functions such as city/high-speed NOA will accelerate the popularization of mass production.

In Li Yuan's view, the most important value of vehicle-mounted lidar for autonomous driving is "safety."

As models equipped with lidar are proven to reduce accident rates and personal casualties, and as users gradually form the understanding that "installing lidar equals greater safety," more and more car companies will choose to use lidar as a standard. matching, thus forming a positive cycle.

"Once this kind of life-saving thing is statistically clearly understood by everyone, lidar will definitely be popularized quickly." Li Yuan said.

At the same time, he also released an early notice: 512-line lidar is not the end, but the beginning of Beiwake’s launch of the first vehicle-mounted lidar. Next, BeiXing will also launch higher-performance lidar to meet the safety needs of car companies.