Changan Deep Blue 4D radar, the trend behind the competition between Zongmu Technology and Sensortech

The battle for 4D millimeter-wave radar has quietly begun.

Recently, ZongMu Technology officially announced that the second-generation 4D millimeter-wave radar (SDR2, four-corner radar) has been mass-produced on the DeepBlue G318 model. In the future, it will also be installed on dozens of models on many platforms of Changan Automobile, and it is expected to reach millions throughout its life cycle.

As the first company in China to equip mass-produced cars with 4D millimeter-wave radar, Deep Blue's two earlier models were partially equipped with the STA77-8 (4-chip cascade 4D imaging radar) and the upgraded version of the STA79-2 short-range radar (officially known as high-density point cloud millimeter-wave radar) from Sensortech.

However, SensTech also has a higher performance STA79-8 corner radar, which is labeled as a high-resolution corner radar (that is, a 4D imaging corner radar), and has been mass-produced on the Hongqi EHS7. The "failure" in the new Deep Blue model further highlights that the market competition for 4D radars has intensified.

As we all know, in the past few years, as one of the leaders of domestic millimeter-wave radars, Sensortech has quickly gained a foothold by leveraging the resources and financial endorsement of Hikvision (the controlling shareholder), its cost-effectiveness advantages, and the opportunity of localizing the supply chain of automobile companies.

However, the weakness of the single radar business has already laid hidden dangers, especially since most Tier 1 intelligent driving companies are involved in the research and development and layout of perception hardware.

Data shows that Senstech, which officially merged with Hikivision at the end of February last year, had revenue of 590 million yuan and net profit of 15.5597 million yuan (net profit margin of less than 3%) from March to December 2023; in comparison, the net profit margin in 2022 was only about 3.3%.

With the (reorganization) merger of Hikvision Automotive Electronics and Sensortech, whether the former, with its business model mainly focusing on panoramic view, parking and after-sales installation, can grab market share in the mid-to-high-end intelligent driving market will also have a certain impact on the subsequent growth space of Sensortech's radar business.

However, with the start of a new round of intelligentization of auto companies, the supplier switching of vehicle models has entered a new active cycle. Behind this, in addition to the competition among new Chinese suppliers, traditional foreign suppliers are also actively grabbing orders.

In fact, judging from the parameters of some 4D radar products currently available to the public, product performance indicators are still in the exploratory stage, and the comparison between products from different companies also shows that each has its own strengths.

According to official specifications, Sensortech's STA79-8 can achieve a horizontal resolution of 2.5° and a pitch resolution of 4° in a wide field of view of 150°, with a detection range of up to 200 meters. In comparison, Zongmu Technology's SDR2 has a horizontal resolution of 4.5°, a pitch resolution of 3°, and a detection range of more than 150 meters.

As another strategic partner of DeepBlue, Huawei has participated in co-creating another new model, the DeepBlue L07, which will also be equipped with Huawei's full-stack intelligent driving solution. Interestingly, during the Shanghai Auto Show in April this year, Huawei also released a new generation of high-precision 4D millimeter-wave radar for the first time.

Public information shows that this high-precision 4D radar is based on a waveguide antenna design, has a detection range of 280 meters, supports parking mode, has a vertical field of view of up to 60°, a distance accuracy of 5 cm, and can identify high and small obstacles. According to the plan, Huawei's ADS 3.0 advanced intelligent driving will be the first to be equipped with this 4D radar.

In fact, 4D radar has entered the pre-installation track as early as two years ago. In the Chinese market, the first model is the RISING PILOT intelligent driving system of SAIC Feifan R7, which is equipped with PREMIUM 4D imaging radar from ZF.

Since then, models including Lotus, Ideal, and Deep Blue have been gradually installed. However, due to the iterative improvement of performance maturity, higher cost than traditional radars, and low integration with high-end systems (involving point cloud processing and multi-sensing technology running-in), its popularity in the original equipment market has dropped sharply.

At the same time, during this stage, with the leap-forward improvement in pure visual perception capabilities of automakers and some intelligent driving solution providers, coupled with cost reduction and efficiency improvement as well as the installation of lidar on vehicles, the market demand for 4D radar did not meet expectations.

Last year, Sensortech delivered 4D forward radar (two-chip cascaded 4D imaging radar STA77-6) for the first time on a large scale in the Ideal L7 model. However, according to the evaluation of the car company, "at present, it is still in the stage of exploring software application, and it needs to be fully verified before we know whether it can play the expected performance."

Similarly, Tesla also confirmed that at present, the company is only conducting relevant verification work on the Model S and X to evaluate the actual performance of the products. There is currently no plan to install 4D radar on the Model 3 and Y models.

As we all know, in addition to adding 1D height measurement parameters, the biggest advantage of 4D point cloud is that it is much cheaper than LiDAR, but can also achieve high-resolution obstacle recognition and classification. However, the technical threshold is also raised, "Hardware becomes easier, but the difficulty of software increases geometrically."

Previously, some 4D radar core chip module manufacturers said bluntly: We are providing software stacks as much as possible to help our end customers (automakers) adapt to this technology as soon as possible. For automakers and Tier 1, it also takes time to optimize software and hardware performance.

As for 4D radar, the opening of the OEM market has also forced traditional millimeter-wave radar giants such as Bosch and Aptiv to catch up. For example, at the CES exhibition earlier this year, Aptiv first demonstrated the seventh-generation 4D millimeter-wave corner radar developed by a Chinese team and equipped with the first integrated radar chip.

In addition, the company is promoting the Gen7 radar series, among which the FLR7 forward 4D radar has a horizontal angular resolution of 2°, a ranging range of 290m (target objects the size of a motorcycle), and supports 100M bit/s Ethernet data transmission rate and 1Gbit/s CAN-FD data transmission rate. It is expected to be mass-produced and delivered in the third quarter of 2025.

At the same time, manufacturers such as Aptiv are also actively promoting the "satellite architecture" sensor configuration solution, making full use of the computing power redundancy of the domain controller for advanced algorithms for target detection, classification and tracking, thereby reducing the computing power cost on the terminal side (previously, 4D radars all used expensive FPGAs).

The benefits of this architecture are obvious. The data from multiple surround 4D radars are collaboratively integrated and processed in a central domain controller (for example, the intelligent driving domain controller), thereby reducing the data processing requirements on the end side.

At the same time, with the help of the radar's BEV feature map, it can be fused with the image BEV feature to further enhance the confidence of target detection. However, at the same time, the requirements for software development are higher.

As we all know, with the technological upgrade from traditional 3D to 4D imaging, software algorithm capabilities have become the shortcoming of millimeter wave radar companies that used to rely more on hardware design, development and manufacturing. (Especially from 4D to imaging, the requirements for software are also a qualitative change)

Bosch also began to promote its sixth-generation millimeter-wave radar this year, which uses the latest RFCMOS technology SoC chip and 3D waveguide antenna technology, point cloud output and processing, and provides AI-based target classification and road edge classification.

Public parameters show that the detection distance of this radar has been greatly improved from the original (fifth generation) 210m to 302m, and the horizontal angular resolution has reached 1°; in terms of performance, for some ADAS functions, it can support vehicle speeds of up to 250km/h.

Arbe, a U.S.-listed 4D imaging radar chip module solution provider, also announced earlier this year that the 4D imaging radar jointly developed with its Chinese partner (Hirorain Technologies) will be put into large-scale mass production by the end of this year.

The imaging radar, called LRR610, provides ultra-high resolution in azimuth and elevation based on 48 transmit/48 receive signal channels. It also has a wide field of view (120° azimuth × 30° elevation), long-range detection of more than 350 meters, high resolution and dynamic range for precise target separation, and all-weather reliability.

In addition, many car companies and listed companies are also gradually getting involved in the millimeter wave radar track. Among them, BYD is good at vertical integration of the supply chain. In March last year, BYD's Fudi released a new generation of RF1 platform forward millimeter wave radar, which was developed based on TI's AWR2944. According to the plan, the company will also launch a 4D imaging radar.