Can it kill lidar?

Why has 4D millimeter wave radar become the industry’s “new favorite”?

“Bosch has completely given up on developing autonomous driving lidar.” I believe everyone already knows the news just revealed by the German Handelsblatt.

Bosch gave up on lidar because of the complexity and long development time of its MEMS (Micro-Electro-Mechanical-System) technology, which it invested heavily in research and development. A spokesperson for the company confirmed that “due to the complexity of the technology and time to market, Bosch recently decided not to invest any additional resources in the hardware development of lidar sensors.”

However, many domestic media reports on this matter did not explain that as one of the four giants of global automotive millimeter wave radar "ABCD", Bosch still has technology that "will never give up", that is, working at 77Ghz 4D millimeter wave radar in the frequency band. This is where Bosch is heavily protected.

博世也坦承，激光雷达对L3级自动驾驶非常重要。因为，虽说行业內4D毫米波雷达正在风生水起，但也并不能“平替”激光雷达。不过，4D毫米波雷达究竟有什么重要的潜力，值得博世这样的巨头宁可放弃激光雷达也不放弃呢？

"New favorite" was born from "slap in the face"

It is worth mentioning that before this, another auto parts giant ZF had also withdrawn from the lidar competition. ZF once invested about US$100 million in Ibeo (the "originator" of lidar), but after Ibeo went bankrupt, ZF also gave up lidar.

As for vehicle companies, Tesla gave up lidar very early.

However, in February this year, Tesla’s vehicle change application submitted to European regulators clarified that the upcoming fourth-generation Autopilot hardware HW4.0 added a high-resolution 4D millimeter-wave radar. Previously, Tesla has successively removed millimeter wave radar and ultrasonic radar, trying to All in visual route.

As soon as this "slap in the face" incident came out, the industry was abuzz. It turns out that you, a thick-browed and big-eyed person, are also doing this. Then, 4D millimeter wave radar suddenly became the industry’s “new favorite”. According to incomplete statistics, at least nearly 20 local companies in China are currently developing 4D millimeter wave radar products.

why? The reason is that compared with traditional millimeter-wave radar, 4D radar can more accurately identify static objects due to the added height information, which is the so-called "imaging" capability. In addition, the price is about 1,000 yuan. Compared with lidar, which can easily cost 6,000, 7,000 yuan or even tens of thousands of yuan, the cost advantage is undoubtedly revealed at a time when the automotive industry is "rolling" to the extreme.

Moreover, compared with lidar, 4D millimeter wave radar inherits the all-weather anti-interference advantage of traditional millimeter wave radar and is not affected by light, smoke, dust, haze, and can work normally at night, rain, snow and other environments. More adaptable.

For example, CICC believes that 4D imaging radar can improve the performance of millimeter wave radar in all aspects, and is expected to make millimeter wave radar one of the core sensors in the ADAS system, which is an important direction for the future development of millimeter wave radar.

In fact, 4D millimeter wave radar is not a new technology. At the end of 2018, in order to compete with the two major manufacturers Infineon and NXP, Texas Instruments TI proposed the concept of 4D imaging millimeter wave radar and launched a complete set of 4-chip cascaded 4D millimeter wave radar based on the AWR2243 FMCW (frequency modulated continuous wave) single-chip transceiver. The design solution integrates the most difficult antenna for radar developers.

In March 2020, Waymo, a subsidiary of Google, released a fifth-generation autonomous driving system perception solution, upgrading millimeter-wave radar to 4D imaging radar, making 4D millimeter-wave radar technology used in vehicles for the first time.

In this year's CES speech, Amnon Shashua, CEO of Mobileye, a subsidiary of Intel, emphasized the application scenarios of 4D imaging millimeter wave radar in automobiles. He said: "By 2025, except for the front of the car, we only want (4D) millimeter wave radar, not lidar."

"Spring River Plumbing Duck Prophet", in fact, the "volume" of 4D millimeter wave radar in the domestic market has already started. Moreover, since the second half of 2022, 4D millimeter wave radar has accelerated the pace of "onboarding", and the main models include Feifan R7, Deep Blue SL03, Lili L7, etc. Some models require both lidar and 4D millimeter-wave radar.

Of course, as early as October 2021, more than half a year ago, Bosch exhibited the fifth-generation millimeter wave radar Supreme Edition (as it was called at the time) for the first time in Shanghai, with a maximum detection range of 302 meters and a horizontal field of view of 120 degrees. Vertical field of view 24 degrees.

At the same time, Bosch’s old rivals Continental, ZF and Aptiv are also not idle. For example, Continental has the world's first 4D millimeter wave radar ARS540 put into mass production, and ZF's 4D radar has been designated by SAIC. At that time, many companies in the domestic market, including Senstech, Freetech, CubTEK, and Sichuan Microwave, had also entered the 4D track.

By 2023, competition will become more intense, and more companies will invest in the research and development and mass production of 4D millimeter wave radar. Just on August 15, Li Yifan, CEO and founder of Hesai Technology, participated in the first round of financing of Aotu Technology in his personal capacity.

It is reported that Altos V1, Altos V1, Altos V1, Altos’ first product to enter the mass production stage, is “currently the only mature, non-FPGA 4-chip cascaded (12TX, 16RX) 4D imaging radar product in the world.” Moreover, in terms of comprehensive performance, it is While the performance of similar products is weaker than that of industry giants ZF, Bosch, Continental, etc., the price of Altos V1 is only 1/2 of the former or even lower.

Bosch can give up lidar, but it will never give up 4D millimeter wave radar. In August last year, Bosch reached an agreement with GapWaves, an innovative radio frequency antenna technology company from Sweden, to jointly develop and produce millimeter-wave radar antennas that meet the high-resolution requirements of vehicles.

Therefore, the giants nicknamed "ABCD" (i.e. Autoliv, Bosch, Continetal, Dephi), as well as the top domestic start-up companies, have invested heavily in this. It is obvious what the future of 4D millimeter wave radar will be like. thing.

Not achieved overnight

Tesla’s “slap in the face” fact proves that even a purely visual solution still requires 4D millimeter wave radar as system redundancy. In addition, "multi-sensor fusion" is a widely recognized intelligent driving solution in the industry, and 4D imaging radar will obviously occupy a place in it.

However, an important fact is that 4D millimeter wave radar cannot "replace" lidar. This is due to its technical heterogeneity with lidar. Of course, this does not affect the bullish market for 4D millimeter wave radar.

We won’t go into details about the benefits of 4D millimeter-wave radar. We only need to clearly know that the development and application of 4D millimeter-wave imaging radar is not achieved overnight. There are still many technical difficulties that need to be continuously optimized and improved. We can also see from our table that there are still only a few companies that can mass-produce 4D millimeter wave radars, and most are still in the research and development stage.

Specifically, first of all, 4D radar needs multiple indicators to meet the conditions and requirements of the original equipment manufacturer at the same time. It needs to improve the distance resolution, angle resolution, and speed resolution at the same time to achieve a better imaging effect. The power of a single indicator is crucial to the final Imaging is of little significance, which is not an easy task for 4D millimeter wave radar manufacturers to achieve "excellence in every aspect".

Secondly, if you want to give full play to the technical advantages of 4D millimeter wave radar, you need to fuse it with the camera. However, 4D millimeter wave radar has many channels and a large amount of data, so it cannot be compared with vision before fusion. The power requirements are relatively high, and it also involves algorithmic problems. The computing power on the sensor side is not enough because the memory of the millimeter wave radar chip is limited.

Therefore, if the point cloud density of the data is relatively high, pre-fusion needs to be done on the domain controller. But there are also problems. On the one hand, the high data rate and data compression of 4D millimeter wave radar will bring challenges to the centralized architecture. On the other hand, the bandwidth and rate of signal transmission between the antenna and the processor will also affect the detection accuracy.

Therefore, to resolve these contradictions, 4D millimeter wave radar manufacturers need to have a deep enough understanding of the central domain controller, or be deeply bound to a domain controller manufacturer or chip manufacturer. For example, after Ambarella acquired Oculii, it launched a new generation of 4D imaging radar. But we know that these core technologies, at least currently, are rarely controlled by domestic companies.

In addition, front fusion requires joint calibration of 4D millimeter wave radar and camera, but joint calibration is difficult. In addition, 4D millimeter wave radar has distance information, but cameras do not. So how to deal with the joint calibration of the two is also a big problem.

Moreover, the issue of the number of antenna channels and chip cascading also has a strong technical content. For example, some domestic companies mostly refer to the traditional design experience of ABCD. If the 12T16R (12 transmit channels, 16 receive channels) antenna array with four chip cascades is designed from scratch, the simulation volume and server requirements alone will be too much. constitute a threshold.