The 4D imaging radar track is lively again.
Since February 2023, the world's first mass production model of Sensitech's 2-chip cascade 4D imaging radar STA77-6 - Ideal L7 has been officially released and launched, setting an important milestone in the mass production of domestic 4D imaging radar products in passenger cars, and the industry generally believes that 2023 will usher in the first year of large-scale mass production of 4D imaging radar.
Although domestic companies including Foretek, Chengtai Technology, Chuhang Technology, Muniu Technology and other leading companies are developing 4D imaging radars, it is a pity that the volume of 4D imaging radars in 2023 has not reached the optimistic expectations of the market.
InLi Xuyang, co-founder and CEO of Sain LingdongIt seems that this is in line with the normal law of market development, after all, last year's domestic 4D imaging radar products generally have problems such as immature technology, compared with other sensors, the competitive advantage is not obvious, and the cost performance is not high.
However, judging from the bidding for 4D imaging radar in the past period,Q1 and Q2 in 2024 are expected to achieve rapid volumeAmong them, the 4D imaging radar product led by Sain is expected to be mass-produced in the first half of this year in the new model of a leading new force, and the whole system is standard.
As one of the few new players in China with the mass production strength of 4D imaging radar, Sain's actions are indeed fast enough.
Founded in November 2021, it has completed the product development of two S-series high-performance radars SIR-4K and SFR-2K in less than a year, and won the designation of the full-platform mass production project of the leading new energy vehicle company in February last year. In October, the industry's first fully automated production line for 4D imaging radar was officially completed and delivered in small batches, with an annual production capacity of 800,000 units expected to reach 800,000 units after it was put into operation.
In fact, domestic 4D imaging radar manufacturers represented by Sain Lingdong are expected to achieve large-scale volume in 2024, and the biggest confidence is stillValue for money
According to Gaogong Intelligent Vehicle, the current delivery of double-cascade 4D imaging radar has dropped to 500-1000 yuan, the lowest has even fallen below 500 yuan, and the four-cascade scheme is generally more than 1000 yuan. A number of industry insiders said that it is expected that in 2024, it will drop by 30%-40%, and the cost of the dual cascade scheme is rapidly approaching that of traditional millimeter-wave radar.
In addition,Market demandIt is also one of the promoters. As the focus of intelligent competition among various car companies points to high-end intelligent driving, the requirements of end-to-end capabilities in high-speed and urban scenarios pose higher challenges to the performance of various sensors, and there must be a battle between several different sensors.
Li Xuyang said that in L3 and above autonomous driving, multi-sensor redundancy is required, and vision, lidar and 4D imaging radar can become "brothers"; However, L2 (including L2) sensor configurations are subject to cost pressures, and lidar and 4D imaging radar are in competition.
It is true that with the decentralization of cost pressure, even high-end intelligent driving does not accept the high cost of perception hardware, especially lidar, high-end. In contrast, under the premise of maintaining excellent perception performance, providing a cost-controllable perception solution has become the core advantage of 4D imaging radar.
However, the traditional millimeter-wave radar is limited by the idea of "cost reduction", and the cost saving from the development source has led to a performance breakthrough that has failed to keep up with the pace of industry development, which has been questioned by the market. 4D imaging radar is expected to be accompanied by large-scale large-scale measurement to solve the real-world perception challenge, that is, to detect targets as early and stably as possible.
After all, the 4D imaging radar takes into account the characteristics of low latency, high dynamic target detection accuracy and all-weather characteristics of millimeter-wave radar, and its output point cloud has accurate height information and target contour restoration ability, which greatly improves the confidence of target perception in all scenes.
In particular, its core components have been verified by the market for a long time, the functional safety maturity is high, and the active safety system based on 4D imaging radar is self-closed-loop, with low verification complexity, and will surely become the last firewall for the safety function of the intelligent driving system.
In addition, 4D imaging radar can also compensate for the ability of optical sensors to solve the pain points of typical scenes. For example, the camera cannot see obstacles on the road in foggy scenes; Unable to correctly identify overturned cars, special-shaped cars, etc.; When the color of the object is very similar to the background or the contrast is very low, the optical sensor will have the disadvantage of "not being able to see clearly" and "not being able to read".
This also means that 4D imaging radar is expected to use the best of traditional millimeter-wave radar to achieve the perception performance of lidar and make up for the perception pain points of traditional cameras and millimeter-wave radar. However, from the perspective of the development status of the industry, 4D imaging radar still needs to solve many problems.
The first thing to break through is:Design ideas of traditional millimeter-wave radar systemsAlthough the number of virtual channels is increased through chip cascade, the traditional TDMA wave generation method cannot solve the problem of sending all CHIRPs within the radar perception cycle. SecondlySide-lobe inhibition is insufficientIt is easy to appear false targets (ghosts), and the dynamic range of radar signal processing is insufficient, so it cannot detect small and weak targets.
In addition, in the process of commercializationChoose how to optimize efficiency with multiple goniometric algorithms, and how to choose processors to balance cost and computing powerand other links, which will affect the final product performance indicators, and often have unsatisfactory results such as limited number of point clouds, low quality of point cloud angle measurement, and many false targets.
Based on this, many domestic manufacturers are also impacting the technical highland of 4D imaging radar and exploring the commercialization of this product in the automotive field. Led by Issain4D imaging radar SFR-2KFor example, through a completely independent innovation technology breakthrough, it has broken the industry's traditional perception of millimeter-wave radar performance.
Sain leads the 4D imaging radar SFR-2K
One isWaveform designCombined with the self-developed industry-leading speed deambiguation algorithm and intelligent anti-interference algorithm, the DDMA waveform design is mass-produced at the product level.
The second isGoniometric algorithms, unique to Sain LeadJoint spatial and Doppler transformation techniquesThe angle measurement algorithm can be intelligently optimized according to the scene, the high dynamic range greatly improves the detection ability of weak and small targets, and the industry-leading sidelobe suppression level controls the appearance of false targets (ghosts).
The third is adoptionIndustry-leading dedicated processors for imaging radarWith strict vehicle-grade verification and development standards, we create the ultimate cost-effective and truly landable.
In the real application scenario, SFR-2K (48 channels) outputs 2048 point clouds in a single frame, with a maximum detection distance of 370 meters and a horizontal angular resolution of 1°, which provides a prerequisite for accurate perception at the subsequent target level.
For example,Beyond the limits of visionEspecially in moderate rain weather scenarios, when the driver's vision is obstructed, the SFR-2K point cloud can detect pedestrians crossing T-junctions in advance and stably in a "ghost probe" manner.
The second is inReal road scenes at nightThe falling tires seriously affect driving safety, and the SFR-2K point cloud can be seen to appear stably about 120 meters to make up for the lack of visual ability.
It can be seen that from the perspective of product performance, reshaping the performance limit of 4D imaging radar, Sain Lingdong is ready to go.
According to reports, in addition to following the large-scale mass production of new models of the head new forces, based on a comprehensive consideration of market demand, in 2024, Sain Lingdong will make a review of the releasedSIR-4K (192 channels, four-cascade scheme).The product will be comprehensively improved, and the product performance will achieve a comprehensive breakthrough at the "ceiling" level during mass production.
The employer's high recognition of the company's innovative technology and 4D imaging radar market prospect has also promoted the company's positive development.
Following the completion of hundreds of millions of yuan in Series A financing at the beginning of last year, Sain led the completion of 100 million yuan in Series A+ financing in December, which was jointly led by Gimpo Investment and Huaqiang Capital, and the additional investment in Series A by old shareholders NIO Capital and Xiaomi Industrial Investment, the funds will be used to increase R&D investment and continue to promote the process of product industrialization.
Overall,Excellent technology and products, large orders in hand, and high-frequency financingIt is the troika that leads Sain to lead into a positive cycle of business.
However, there are still a few domestic manufacturers with full potential for large-scale 4D imaging radar, such as Sain Lingdong, and it remains to be seen whether 2024 can usher in the first year of real 4D imaging radar.