tech

A few days ago, the response to the high-speed collision accident of M7 was made. One point that can be seen from the accident is that there are multiple factors currently affecting the intelligent driving assistance and AEB active safety, including the performance of sensors and the rigid conditions behind triggering the active safety mechanism.

Intelligent driving has always been a focus of competition among car companies. From the perspective of intelligent driving, cameras, millimeter wave radars, and LiDAR are the three mainstream methods.

Traditional millimeter wave radars cannot perceive stationary objects, and the heavy responsibility of perception and recognition is handed over to cameras, but cameras still have great limitations, which are greatly affected by the environment, and rain, snow, and darkness may cause the camera to fail. LiDAR can directly scan each pixel of the object, estimate the distance from the object by calculating the time it takes for the laser to reflect back from the object, and more accurately restore the scene, and darkness or too dazzling sunlight will not affect its efficiency.

Advertisement

At present, apart from Tesla and Jie Yue, which insist on the pure vision route, other car companies' new cars would be embarrassed to talk more about their intelligent driving experience if they do not carry LiDAR.

In the past two years, when domestic smart cars were still in the material stacking period, the person in charge of Great Wall's salon robot dragon once said: "Below 4 (LiDAR), please don't speak."

However, by 2024, car companies are no longer "keen" on LiDAR.

01

Shuffling anxiety, car companies reduce configuration

"On the current table, if you only have the card of electrification in your hand and are not prepared for intelligence and ecology, the car company will still face reshuffling." Some time ago, Jia Ke, the chairman of the China Automobile Blue Book Forum, pointed out the new anxiety faced by car companies to the point.Tesla has reduced the prices of the Model 3 and Model Y, and Xiaomi has entered the public eye with its own SUV 7. At the same time, companies such as Wenjie, Li Auto, and BYD have launched high-end models such as the Wenjie M9, Li Auto MEGA, and Yangwang U8.

On one side, there is a new round of price wars triggered by Tesla, and on the other side, domestic car companies are making a breakthrough in the high-end market, which is actually a further intensification of competition in the automotive market.

Looking back at the smart electric vehicle market, the simplest way to achieve a price war is to reduce prices and reduce configurations. As the most expensive core component of intelligent driving on the car, under the pressure of the price war, the hardware reduction and "sacrifice" of LiDAR has been put on the agenda.

As mentioned earlier, at present, the "perception" part of intelligent driving is mainly based on three routes: cameras, millimeter-wave radars, and LiDARs. Most of them need to be integrated through the "Kalman filter", and the integration is fundamentally to improve the accuracy of "perception". Therefore, from a technical point of view, the more perception components, the better.

However, intelligent driving will eventually return to the user, and in the trade-off between cost, scale, and accuracy, smart cars will ultimately consider their "product" attributes.

Under the premise of prioritizing intelligent experience and safety, car companies have three ways to reduce costs on intelligent perception hardware:

The first is to directly reduce the configuration of LiDAR, and car companies can only provide users with high-speed NOA assisted driving.

The second is to change the technical route, replace the perception hardware with visual cameras, and take the all-visual route, such as Tesla and Ji Yue. However, it is not easy to copy Tesla's homework, and the threshold and barriers are extremely high. From the price of the intelligent driving of these two companies, the cost of the pure visual route is actually more expensive at this stage.

The third is to pursue cost-effectiveness. Reduce the number of LiDARs, replace some of the LiDARs with 4D millimeter-wave radars, and reduce the cost of the number of LiDARs on the car.

024D Millimeter Wave Radar Gains Attention

Whether a technology is practical or not often depends on whether we have successfully developed it. This statement is more evident in the industry of 4D millimeter wave radars.

Before the advent of 4D millimeter wave radars, Tesla abandoned the use of millimeter wave radars.

Two years later, Musk, who was once firmly against millimeter wave radars, also softened his stance on Twitter, saying: "High-definition millimeter wave radar is a feasible choice." Subsequently, it was revealed that Tesla had reserved a gigabit network interface for millimeter wave radars in HW4.0.

In addition, Bosch also began to favor 4D millimeter wave radars.

In July last year, a spokesperson for Bosch confirmed: "Given the complexity of the technology and the time required for market promotion, we will no longer invest more funds in the hardware development of lidar." In short, the cost of lidar is too high, and car manufacturers are unwilling to use it.

On the contrary, Bosch plans to transfer technology and investment to other sensors, namely 4D millimeter wave radars. It is reported that as early as last August, Bosch reached an agreement with the Swedish innovative radio frequency antenna technology company (GapWaves) to jointly develop and produce millimeter wave radar antennas that meet the high-resolution requirements of vehicles.

Recently, Huawei's Kunpeng announced the first domestic high-precision 4D millimeter wave radar, which once again ignited the popularity of 4D millimeter wave radars.

So, what is the charm of 4D millimeter wave radars? The frequent consideration of 4D millimeter wave radars is nothing more than two reasons: performance improvement and low cost.

First, let's look at the performance. The main functions of the current mainstream millimeter wave radars are angle measurement, distance measurement, and speed measurement, hence they are also called 3D millimeter wave radars. Previously, Tesla's Auto Pilot was supported by 3D millimeter wave radars for computation. However, the inherent drawback of 3D millimeter wave radars is that they cannot measure the height of objects, making it impossible for them to identify whether stationary objects in front will affect the vehicle's passage. Compared to traditional radars, 4D millimeter wave radars have an additional "1D," which is the ability to detect the height of objects.The 4D millimeter wave radar adds measurement information of the pitch angle, and the angle resolution can reach the sub-degree (<1°) level, which can clearly present the outline of the target obstacle through a large number of measurement points. In addition, the 4D millimeter wave radar not only provides three-dimensional data such as speed, distance, and azimuth, but also "evolves" a point cloud imaging effect similar to that of LiDAR, which can also compensate for the shortcomings of pure visual algorithms that cannot cover the entire scene. In terms of clarity, some indicators are close to the 16-line LiDAR. In rainy or foggy weather, its detection range can still reach about 300 meters.

Huawei's latest high-precision 4D millimeter wave radar can support parking mode, with a vertical field of view of 60°, which is three times the vertical field of view of 18° of traditional radar; the distance accuracy is 5 cm, which is four times the accuracy of 20 cm of traditional radar.

In terms of price, from the perspective of cost, the price of LiDAR is mostly around 1,000 US dollars, while the cost of 4D millimeter wave radar is about 1/10-1/5 of that of LiDAR. At present, the overall price of 4D millimeter wave radar is about a thousand yuan, and the price is rapidly approaching that of traditional millimeter wave radar. For example, the imaging radar of Arbe company is priced at 690-1036 yuan, and the unit price of products from Tier 1 such as ZF and Continental is 1036-1381 yuan.

From the perspective of vehicle application, car companies have voted with their feet. It is understood that the new generation of NIO models has determined to install 4D millimeter wave radar. The NIO NT3.0 platform models will come standard with a 4D millimeter wave radar from SAIC Motor. The first model of Le Dao, L60, also removed LiDAR and added 4D millimeter wave radar, with visual capabilities basically the same as NIO, and it can be used in urban NOP+ as soon as it is launched, which may be the first time among new forces.

Domestic manufacturers have set off a wave of 4D millimeter wave radar, and car companies have found the best cost-effective ratio of L2+ autonomous driving perception hardware at the same time. According to Arbe's forecast, 4D millimeter wave radar will begin to be equipped in L1-L4 autonomous driving models in the European and North American markets in 2024. In terms of market size, according to Yole's forecast, the compound annual growth rate of 4D millimeter wave radar from 2021 to 2027 will reach 48%, and the compound annual growth rate of 4D imaging radar will reach 109%.4D Millimeter Wave Radar: Who's Joining the Game?

From the perspective of the capital market, the popularity of 4D millimeter wave radar is continuously rising.

Last year, NIO and Xiaomi invested in Sensetime, with the financing scale exceeding 100 million yuan, mainly used for the mass production research and development and industrialization of 4D imaging radar. At the same time in August, Li Yifan, CEO of Hesai Technology, known as the "first stock of LiDAR in China," also invested in the new 4D radar company Autotalks.

Not only Sensetime, Xiaomi has also invested in multiple rounds of financing for Geometric Partners and Zongmu Technology. Among them, Baidu also invested in Geometric Partners, and each round of financing amounted to at least 100 million yuan.

According to incomplete statistics from the semiconductor industry, since 2023, at least 10 domestic 4D millimeter wave radar companies have obtained financing, and the disclosed total financing amount has far exceeded one billion yuan, showing the heat of this market segment is not ordinary.

Let's take a look at the current "battlefield" of 4D millimeter wave radar.

In the 4D millimeter wave imaging radar track, the main players include three types: traditional Tier 1, technology giants, and emerging companies. Among them, traditional Tier 1 companies such as Bosch, Continental, Aptiv, and ZF have laid out earlier and generally use multi-chip cascading technology to increase reception and transmission channels. Technology giants such as Mobileye and Huawei have also launched 4D millimeter wave imaging radar.

In addition, there are emerging companies such as Oculii, Sensoro, Radar Technology, Muniu Technology, Chengtai Technology, Zongmu Technology, Geometric Partners, Chuhang Technology, Xingyidao, Sensetime, and Huayu Auto.

Oculii's current 4D millimeter wave radar solution has been applied to the unmanned logistics vehicle jointly built with Great Wall Motors, and at the same time, it has also fixed multiple OEM mass production projects.The global debut mass production model of Sensitec's 2-chip cascaded 4D imaging radar STA77-6 is on the Ideal L7. In addition, Sensitec's 4-chip cascaded 4D imaging radar STA77-8 has been mass-produced and launched on a leading domestic vehicle model.

Mu Ni Technology currently equips its 4D imaging radar products on Bobcat mechanical vehicles and has reached a strategic cooperation with the latter for mass supply. In addition, last year, a new 4D imaging radar I79 was released.

Sain Lington, which has received financing from Xiaomi, previously launched the 4D imaging radar SFR-2K product. The SFR-2K (48 channels) can output 2048 point clouds in a single frame, achieving a horizontal angle resolution of 1°. From the latest progress, Sain Lington has currently obtained a full platform mass production project from a leading new energy vehicle company, and is expected to be mass-produced and launched in the second half of this year.

04

Conclusion

From the overall market perspective, 4D imaging radar has not yet achieved large-scale vehicle installation and is still in the small-scale trial stage.

To promote the early installation of 4D imaging radar, the entire industry still has a lot of work to be initiated and completed: clearly define and set the minimum performance requirements of the sensor suite: resolution, false detection rate, range, delay, and sensitivity, including 4D imaging radar; clustering, object boundary, free space mapping algorithms, and tracking algorithm integration; reasonable installation positions and application methods; unified performance testing and evaluation standards, etc.

Under the current mainstream of L2/L2+ intelligent driving, 3D millimeter wave radar can basically meet the needs of most scenarios. 4D imaging radar is currently in a "nice to have" situation. If it is better, the question is: is it necessary for companies to spend more money to replace sensors with higher clarity?

However, the emergence of 4D millimeter wave radar has filled the automotive market with hope. The industry is looking forward to 4D millimeter wave radar overcoming difficulties and reducing the BOM cost to below 50 US dollars, which will not only be the starting point for its installation, but once entered, it may face a vast automotive market in the range of 100,000-150,000.

Driven by technology and capital, 4D millimeter wave radar is starting to heat up.