Whether a technology is good or not depends on whether we make it or not.
Nowhere is this more evident than in the 4D millimeter-wave radar industry.
Before the advent of 4D millimeter-wave radar, Tesla abandoned millimeter-wave radar. Only two years later, the 4D millimeter-wave radar came out, but it was brought on fire by TeslaTesla's next-generation self-driving hardware system HW40, reserved for millimeter-wave radars100 Gigabit Ethernet interface
At the same time, domestic car companies are also keeping an eye on the 4D millimeter-wave radar outlet.
According to Auto Heart, NIO's new generation models have been confirmed to install 4D millimeter-wave radar. NIO NT3The 0 platform model will be equipped with a 4D millimeter-wave radar from Sain Lingdong as standard.
Domestic manufacturers set off a 4D millimeter-wave radar craze, not to follow the trend, but at the same time car companies found the best cost performance of L2+ autonomous driving perception hardware.
Very simply, 4D millimeter-wave radar is less than one-sixth of LiDAR, and in the increasingly fierce market environment, everyone wants to try it.
Can 4D millimeter-wave radar really replace lidar and reach the top of perception upstarts?
The heart of the car threw the problem to Dr. Li Xuyang, the founder of Sainling, who believes that in L3 and above autonomous driving, multi-sensor redundancy is required, and vision, lidar and 4D millimeter-wave radar are brothers.
L2 sensor configurations, including L2, are subject to cost pressures, and LiDAR and 4D mmWave radar are at odds.
The perception upstart 4D millimeter-wave radar is inevitably stirring up a replacement storm.
Why is 4D mmWave radar the new upstart?
Looking back at the smart electric vehicle market, the easiest way to achieve the best battle is to reduce prices and allocation.
But the reality is that even if you want to fight the best war, you must reduce the price under the premise of giving priority to intelligent experience and safety.
As a result, car companies have three ways to reduce costs in intelligent sensing hardware:
The first is the most brutal,Reduced LiDARCar companies can only provide users with high-speed NOA assisted driving, which directly saves hardware overhead, but also cuts the user's intelligent driving experience in half.
The secondChange the technical route, the perception hardware is mainly a visual camera, attacking the all-visual route. Although it saves money on the perception hardware side, from the perspective of Tesla and Jiyue, the cost of the pure visual route is more expensive at this stage.
The third is:Seek a balance between ** and technology, replace all or part of the lidar with 4D millimeter-wave radar, so as to reduce the number of lidars on the vehicle and reduce costs. At present, this method has been adopted by NIO and XPeng.
It is not impossible for the new forces to turn to the combination of millimeter-wave radar + lidar as standard in the whole system.
There are two reasons why 4D millimeter-wave radar has been pushed to the front of the stage, on the one hand, the technology meets the perception requirements of developing urban NOA, and on the other hand, it is cost-effective enough.
Compared to traditional millimeter-wave radar, 4D millimeter-wave radar can not only detect the distance, relative velocity, and azimuth of an object (angular measurements in a spherical coordinate system), but also use time to determine highly correlated information.
Led by Issain4D millimeter-wave radar SFR-2KFor example, a single frame outputs 2048 point clouds to achieve a horizontal angular resolution of 1, and the number of point clouds is a traditional millimeter-wave radar8 times
In addition, the four-chip SIR-4K has 192 channels, and the 4096 point cloud output can compete with the 64-line LiDAR, and the horizontal angular resolution is also increased to 05º。
Wang Weiping, technical engineer of Sain Lingdong, gave an example to the heart of the car
In the traditional millimeter-wave radar, only a point cloud will be displayed, but in the 4D millimeter-wave radar, not only the vehicle height information can be known, but the point cloud accuracy can also be knownUp to 10 times
The increase in altitude information and point cloud accuracy is to improve the confidence of the judgment of the intelligent driving system. On the city road of Zhangjiang Science Park in Shanghai's Pudong New Area, the heart of the car tested the discrimination effect of SFR-2K.
In the 4D millimeter-wave radar point cloud map, the altitude information is distinguished by different color depths, which can help the system determine whether the object is on the ground or in the air, so as to avoid false braking and improve the intelligent driving experience.
During the test drive, it was clear that the faster electric vehicle was driving at a distance of 50 meters, but the 4D radar still provided about 30 point clouds.
Especially after adding altitude information, 4D millimeter-wave radar also makes up for the shortcomings of traditional millimeter-wave radar to identify static objects.
Traditional millimeter-wave radar is inaccurate for static obstacles。According to the Doppler effect, millimeter wave can return to the target point of static objects, but the traditional millimeter wave radar can easily filter out the static foreground information directly due to its limited resolution and inability to recognize altitude information, resulting in the shortcomings of difficult to accurately identify static objects.
With the addition of altitude information and more accurate point cloud information, 4D millimeter-wave radar is more capable of identifying static objects.
Wang Weiping introduced that the traditional millimeter-wave radar is equipped to do AEB testing, and the radar can confirm whether the front is an obstacle when it is meters.
Now, the distance can be stretched up to 100 meters.
, is the second killer feature of 4D millimeter-wave radar.
According to industry market research, the overall 4D millimeter-wave radar is about 1,000 yuan, and it is rapidly approaching the traditional millimeter-wave radar.
Dr. Li Xuyang told the heart of the car that in the next one or two years, the four-chip 4D millimeter-wave radar led by Sain will be achievedLess than 1000 RMB, a single chip to doMore than 500 RMB
In contrast, 4D millimeter-wave radar can basically achieve the same effect as 64-line lidar, but it can bring at least half of the cost reduction space for car companies.
Driven by both cost reduction and performance, it is reasonable to speculate that the combination of a lidar and a 4D millimeter-wave radar will likely replace the dual lidar configuration in urban NOA, and in the 200,000-level smart car market, as the cost of 4D millimeter-wave radar continues to decrease, it will replace the traditional 3D millimeter-wave radar.
Impacting mass production, the dual maturity of manufacturers' capabilities and car companies' needs
Mass production is the first step in the 4D millimeter-wave radar industry.
In 2019, the Israeli company Arbe was the first to launch the 4D millimeter-wave radar Phoenix.
Subsequently, Tier1 and vertical startups such as Waymo, Continental, ZF, Aptiv, and Bosch tried to follow up on the technology.
But overall, 4D millimeter-wave radars are still inSmall-scale mass production was introducedstage, in other words, at this stage, there are not many manufacturers with the ability to mass-produce 4D millimeter-wave radar.
From a technical point of view, the cascade method mostly used in 4D millimeter-wave radar still needs to be solved to increase the probability of detection success, reduce echo noise, and heat dissipation in the case of increasing point cloud density.
In addition, the EMC (electronic compatibility), test standards, and policies of 4D millimeter-wave radar are still in the mature stage.
In the words of Dr. Li Xuyang, it is necessary to have from the demo to the mass production of the productHardware reliabilityStability of software algorithmsAt the same time, it must be able to meet the requirements of car companies.
These variables make the competition for 4D mmWave radar more intense, but the scoring criteria have also become more obvious: whoever goes into mass production first is already half a position ahead.
The first year of mass production of 4D millimeter-wave radar, is it far from us?
Dr. Li's answer is that it's not far away, just in 2024.
He believes that to enter mass production, it is the dual maturity of the capabilities of millimeter-wave radar manufacturers and the needs of car companies.
Starting from the production line, the investment of a 4D millimeter-wave radar production line is often tens of millions, and not all manufacturers have the capacity to build the production line, but only by firmly holding the production and manufacturing in their own hands can the efficiency and quality of large-scale mass production be guaranteed.
Therefore, having a production line is already one of the signs of the maturity of domestic millimeter-wave radar manufacturers.
At present, there are 2 domestic production lines for 4D millimeter-wave radar, and Sain Lingdong is one of them.
It is understood that after the automated production line was implemented in November last year, Sain Lingdong has achieved small batch delivery with an annual production capacity800,000Taiwan years.
Some people in the industry expect that by the middle and end of this year, the number of manufacturers with 4D millimeter wave radar production lines may reach one, and the competition will continue to be fierce.
Second, it depends on the maturity of the market. In other words, the amount of demand for 4D millimeter-wave radar by car companies will directly determine the outbreak time.
In Li Xuyang's view, the sign of maturity is that car companies are willing to equip all models with 4D millimeter-wave radar as standard, rather than only carrying proportionsThe high-end version.
In the past year, there has been no shortage of domestic car companies equipped with 4D millimeter-wave radar. For example, SAIC Feifan R7, Changan Deep Blue SL03 and other models are also equipped with 4D millimeter-wave radar, but they are all in the high-end version**.
NIO NT3The 0 version of the model is equipped with Sain Lingdong 4D millimeter-wave radar as standard, which is undoubtedly a good opportunity for the 4D millimeter-wave radar market to go to the masses.
There are two benefits of 4D millimeter-wave radar moving towards mass production, one isUse scale to drive the cost flywheel
An industry insider once said to the heart of the car that for sensor hardware, even if it is a cost of 1 yuan, the solution is not afraid to tear down and start again.
Behind the saving of one yuan is the scale of tens of millions of products, but because of this, the large-scale increase of 4D millimeter-wave radar standard high-end driving can also usher in the era of thousands of yuan as soon as possible.
Li Xuyang thinksAutomotive-grade radarShipping one million pieces a year is a threshold。OnlyPlatformizationwithScaleto turn on the price-performance ratio of 4D millimeter-wave radar.
In addition, mass production is also an important part of the performance upgrade of 4D millimeter-wave radar.
Li Xuyang said that the 4D millimeter-wave radar has entered the mass production stage, and with the support of data, through the later OTA optimization, the performance of the 4D millimeter-wave radar can be20% 30% increase
Last year, the performance parameters of the four-chip SFR-4K were still 4096, and after a year, under the background of continuous optimization and iteration of data and AI, it was upgraded to about 6000 points and a horizontal resolution of 02 degrees.
Currently, the recall and accuracy of SFR-4K are thereMore than 98%.
At present, the bell of mass production of 4D millimeter wave radar is about to ring, in June this year, SIR-2K will start mass production, and in the second half of the year, the mass production of new models will be launched, with the 4D millimeter wave radar of Sain Lingdong and other manufacturers on the car one after another, the second half of the year will usher in the mass production window of domestic 4D millimeter wave radar manufacturers.
Algorithm + hardware, domestic players are making great progress
If you look back at the 4D millimeter-wave radar market, you will find that it has been less than 4 years since the technology was born and brought to the public eye.
At present, there are two types of players in the industry, one is represented by traditional Tier1s such as Continental, ZF, and Bosch, and the other is vertical track start-up players such as ARBE, Sainlingdong, Aoku, and Chengtai.
The trend is already very clear, as early as two years ago at the CES exhibition, NXP, TI, and Mobileye launched imaging radar solutions, such as Mobileye put a high-performance 4D millimeter-wave radar in the front of the car, and a low-end 4D millimeter-wave radar in the corner.
However, under the interweaving of the two waves of new and old forces of international Tier1 and new domestic players, domestic players have more advantages in the mass production of 4D millimeter-wave radar.
In fact, compared with traditional millimeter-wave radar, the most difficult part of 4D millimeter-wave radar is that it has to redo the signal processing and point cloud processing framework, and do target classification through algorithms.
Therefore, 4D millimeter-wave radar is not just about hardware performance.
In addition to comparing the number and quality of point clouds, it is also necessary to be able to provide 4D millimeter-wave radar algorithms and supporting services for car companies. This is also why more and more new power car companies have chosen domestic first-class companies.
For example, the SFR-2K led by Sain uses the speed deambiguation algorithm and the intelligent anti-interference algorithm to realize the DDMA waveform design at the product level, and also develops the scene intelligent optimization angle measurement algorithm to improve the detection ability of weak and small targets.
These are the added value of technology brought by domestic players to car companies.
Domestic players are gaining momentum, but the competition is just as fierce. One is the same kind of new players who started in the same year, and the other is the international tier1 who has heard the news.
Li Xuyang believes that among the same batch of domestic players, the competition point is cost-effective. Sain Lingdong hopes to use 3-4 years to achieve a higher gross profit margin and higher cost performance than similar players.
In the comparison with the international Tier1, Li Xuyang bluntly said that Bosch's visual capabilities and AI capabilities are very strong, and it will be a strong competitor to all emerging manufacturers in the 4D millimeter-wave radar market.
Overall, with the mass production of 4D millimeter-wave radar on the car, the cost performance is perceived by the market, and the weight of 4D millimeter-wave radar in car companies will be more obvious.
Perhaps, in 2-3 years, 4D millimeter-wave radars will no longer be placed on the inside of the front and rear bumpers, but will be generous like today's lidarsExternalIt has become the logo of high-end intelligent driving on the consumer side.