Millimeter wave radar sword refers to autonomous driving, mainland, Huawei launched 4D imaging will be the “optimal solution”?

Musk tweeted last month that the Beta 9.0 version of Tesla’s FSD will no longer rely on radar. If it does arrive, it will mean that Tesla has regrouped and returned to a purely visual autopilot solution.

Millimeter wave radar sword refers to autonomous driving, mainland, Huawei launched 4D imaging will be the "optimal solution"?

Tesla has now updated the Model 3 promotional page on its North American website, with only the vision and ultrasonic sensor sections remaining about FSD, after the millimeter wave radar information has been removed. In its place: “250 meters of powerful vision processing capability.

Unlike Tesla’s ambiguous attitude towards millimeter wave radar, some players at each end of the industry chain treat millimeter wave radar as one of the important components in autonomous driving and assisted driving, except that it has not been discussed with much heat by outsiders in the past. When we talk about the perception route debate, the discussion is all about cameras vs. lidar, but in fact both programs incorporate millimeter wave radar.

But after Continental launched the world’s first 4D imaging millimeter wave radar in the middle of last year, it seems to have become a small windfall that is crying out for attention. On the eve of this year’s Shanghai Auto Show, Huawei also held a new HI product launch, releasing a new generation of high-resolution 4D imaging radar for use as a core sensor in autonomous driving solutions.

Why is the traditional millimeter wave radar in the vehicle side of the market talk less? What about the performance of 4D imaging radar that has emerged in the last two years?

Millimeter wave radar, lost vertical and blurred horizontal dimension

Millimeter wave radar is still relatively new to people, but in the industry, the vehicle millimeter wave radar in Europe and the United States has more than twenty years of successful commercialization experience and market accumulation. Some data show that the top five global giants, such as Bosch, Continental and Denso, occupy more than 75% of the market share.

The development so far, millimeter wave radar in the traditional giants here is relatively mature automotive sensing sensor, the cost is relatively low, and its perception and camera fusion is also the early realization of L2 level assisted driving program of choice.

Traditional Tier 1 not only occupies the “cake”, but also holds the “knife” to cut the “cake”. Therefore, it is difficult for the new players to tear a hole in it and affect the long-term bundled interests between them and the downstream car companies. Naturally, emerging players focusing on millimeter wave radar will not choose to “fight hard” with these Tier 1s, and will prefer to explore new technological innovations to achieve a bend in the road.

The market pattern is stabilizing, which is one of the reasons why traditional millimeter wave radar does not seem to have a strong presence in the outside world. Another reason is that when autonomous driving is heading towards L3 and L4, the technical inadequacy of traditional millimeter wave radar is gradually magnified. In the view of “intelligent relativity”, there are two aspects as follows.

  1. “Invisible Y-axis”

Traditional millimeter wave radar is lacking in the longitudinal altimetry capability, which can be interpreted as lacking the ability to “understand” the vertical plane.

Because of this lack of “understanding” ability, millimeter wave radar “can not see” the height of bridges, road signs, for example, in its “eyes”, these stationary objects are considered to be on the ground This plane. Based on this premise, if all their reflected signals are not filtered out, millimeter wave radar will undoubtedly give false warnings of obstacles ahead, resulting in “ghost braking”.

However, when there are stationary vehicles under bridges and road signs, this can lead to traffic accidents.

Tesla has been involved in several crashes into trucks, and this is a typical example. In one of them, the Tesla’s camera sensing failed to recognize the stopped truck in front of it. The millimeter wave radar, a backup sensor, was supposed to identify the obstacle ahead and issue a warning, but the millimeter wave radar did not work either.

Because the information of the stationary truck is mixed with those information, the former will also be filtered out by the recognition algorithm of the radar together. Millimeter wave radar recognizes a stationary object, but therefore “ignores” its presence. Thus, the millimeter wave radar is invisible, the camera is disabled, the self-driving car becomes blind and deaf, and finally crashes into the stationary truck.

  1. “Blurred X-axis”

Another limitation of conventional millimeter wave radar is the low lateral resolution, which can be interpreted as a weak “understanding” of the horizontal plane.

Lateral resolution refers to the angle formed by the left and right two scanned laser points, the smaller the angle degree, the higher the lateral resolution. If compared with LIDAR, the lateral resolution of millimeter wave radar does not have the advantage.

For example, Tesla past problems: the front of a car parked next to the road, there may be half the body in the lane, this time Tesla will be because of the millimeter wave radar lateral resolution is not enough to identify the vehicle and more likely to hit it.

On such a problem, Tesla has had two accidents in Florida, is because the millimeter wave radar can not measure the lateral speed, resulting in the recognition of the vehicle in front is not moving, and ultimately lead to the vehicle too late to brake.

So all together, the traditional millimeter wave radar in “understanding” the vertical and horizontal plane are inadequate, but also decided that the traditional millimeter wave radar is difficult to adapt to the sensor of higher-order autonomous driving perception system. However, the recently active 4D imaging radar seems to let us see some new possibilities.

From quantitative change to qualitative change, two routes of 4D imaging radar

The reason for the emergence of 4D imaging radar is somehow the inverse of the bright performance of LiDAR and camera. The demand for perception of autonomous driving increased, the camera was upgraded from 2 megapixels to 8 megapixels, and the semi-solid-state LIDAR started to accelerate on board. And the initial millimeter wave radar was upgraded in parallel to today’s 4D imaging millimeter wave radar.

From the demos or ppt released by the manufacturers in the past two years, 4D imaging radar does technically solve some major shortcomings of traditional millimeter wave radar and amplify the advantages of millimeter wave radar. This year’s Shanghai Auto Show exhibitors such as Continental and Huawei are among the representative manufacturers.

In this process, the general implementation is nothing more than increasing the number of transceiver channels in hardware and expanding the antenna aperture while meeting the horizontal and vertical requirements for resolution. This is also the mainstream solution of 4D imaging radar in this Shanghai Auto Show.

In fact, from the technical point of view alone, we see more technical innovation to bring “quantitative change” rather than “qualitative change”, if the manufacturer only focus on highlighting parameters at the moment inevitably have “excessive marketing” of suspected of “excessive marketing”. The real use of it to promote the relevant products to achieve the landing, is the embodiment of the “qualitative change”. In the view of “intelligent relativity”, there are two feasible directions as follows.

1, focus on segmentation needs, find a new idea of commercial scene realization

4D imaging radar has the characteristics of all-weather operation without fear of rainstorm, bright light and other harsh environments, while on the other hand, 4D imaging radar is a kind of “ascension” of millimeter wave radar, which also continues the advantages of the past, that is, the cost.

These two features actually give it the opportunity to do driverless technology implementation and commercialization in closed parks and other commercial scenarios. One of the “unmanned delivery cart” is a suitable option, because at this stage of mass production there are still problems with its sensor cost performance and all-weather operation, so that 4D imaging radar to improve.

Unmanned delivery vehicles travel at low speeds, and at this stage mostly use LIDAR with fewer beams. LIDAR is developing rapidly, is the cost of the decline, just not enough pro-people. LIDAR-based sensor solutions make the cost of a single vehicle too high, is the main obstacle to the commercialization of unmanned delivery landing, which has also long been the industry consensus.

Therefore, people are very sensitive to the cost of sensors for unmanned delivery vehicles and will be happy to see a driverless solution that guarantees adequate performance and is also more cost-effective. By combining multiple 4D imaging radars, it may be possible to do this.

For example, GaoGong Intelligent Vehicle has reported that Great Wall Motors, Yihang Intelligence and Oculii have jointly built an unmanned logistics cart that provides a customized sensing solution for low-speed logistics park scenarios based on 4D imaging radar. The unmanned cart can image, identify and track point clouds of 360° low-speed or stationary pedestrians, obstacles and small objects in the surrounding area around the clock, and its point cloud effect is very close to LIDAR.

4D imaging radar can help shorten the mass production cycle of the unmanned car and reduce its commercialization difficulty after further lowering the cost. Therefore, 4D is only 3D “up-dimensional”, but “cost efficiency” is better, and the “added value” of closed scenes is higher. It is foreseeable that it will replace low-cost LIDAR in more specific scenarios for balance considerations.

If we continue to zoom in, manufacturers choose to commercialize in closed commercial scenes, for themselves is also a “from near to far” development ideas. After the maturity of the application of special scenarios with less risk, the gradual transition to the more difficult intelligent car autonomous driving. Although you may not be able to catch the “early set”, but the steps forward may be solid.

2、Be a friend of time and build the best landing posture for the whole set of autonomous driving program

4D imaging radar compared to the car’s other two pairs of “eyes”, see the distance farther.

For example, ZF will provide the radar to SAIC in 2022, the farthest illumination distance can reach 350m; Huawei released a high-resolution 4D imaging radar, detection distance can be 300m, the traditional usually 200m. so it is worth affirming that 4D imaging radar to the automatic driving system to leave more processing time, which is difficult to surpass the advantages of cameras, LIDAR.

This also gives a way of thinking that 4D imaging radar, camera and LIDAR can complement each other when landing a whole set of autonomous driving solutions. If the signals of the three do effective fusion and redundancy, will promote the whole program gradually approach the goal of the ideal sensor.

How to define the goal of the ideal sensor? In the opinion of “Intelligent Relativity”, “omnipotent” is a core keyword to define it. From this core extends several basic points, such as “technical omnipotence” and “scene omnipotence”.

For 4D imaging radar, combined with the guidelines of automotive functional safety, “automotive-grade applications are not too safe”, just in terms of detection distance, not afraid of bad weather, these characteristics, it has been verified that it is the autonomous driving to achieve “technology omnipotence” when It is an indispensable sensor for autonomous driving.

4D imaging radar with these technical characteristics into the integration, you can “unlock” more functional scenarios, to promote the whole set of autonomous driving solutions “scene omnipotent”.

Take Huawei as an example, in the HI new product launch, Huawei has mentioned 4D imaging radar “three major capabilities” and “six values”. The former contains most of the technical content in the industry is not new, not no one I have. However, the latter involves new functional scenarios, we as users are worth to look forward to.

For example, there are high-speed cruise, safe obstacle avoidance, urban cruise, non-visual distance before the car detection and so on. Huawei points out that for urban cruising, 4D imaging radar’s large field of view without blurring capability can match some urban scenes (mixed traffic, large and small targets in parallel, obscured scenes); for high-speed cruising, if two vehicles 220m away are exactly the same speed and distance, located in adjacent lanes, 4D imaging radar can be distinguished by the angle.

So for now, 4D imaging radar will be a perception sensor that cannot be ignored in the field of autonomous driving. Huawei, Continental, ZF, Arbe and other OEMs and suppliers are laying out among them. Behind the arms race of major manufacturers, the technology of 4D imaging radar will definitely become more and more mature, and with it, it may be a step closer to higher-order autonomous driving.


High-order autonomous driving has put forward new requirements for in-vehicle millimeter wave radar, giving rise to 4D imaging radar, but we cannot ignore the reality that ADAS localization is not yet mature, and the market is still highly monopolized. It is not a healthy signal for millimeter wave radar manufacturers if everyone is just overly pursuing the “new” challenge of overcoming higher-order autonomous driving and missing the immediate ADAS front-end market development window.

The inherent advantage of ADAS localization is that manufacturers are closer to end customers and China’s localized road scenarios and have a more accurate grasp of demand. Smart cars and autonomous driving are in full bloom, with higher requirements for hardware and software customization than in the past, and overseas Tier 1 is precisely unable to customize flexibly for the Chinese market in the short term, which to a certain extent creates opportunities for domestic manufacturers to break through.

Then, applying to ADAS actually means close product landing and seizing the market, applying to higher-order autonomous driving means a long persistence to run to the starry sea, manufacturers can choose one or the other simultaneously. Many choices, 4D imaging radar future will be how to present in front of everyone, still need to wait patiently.

Posted by:CoinYuppie,Reprinted with attribution to:
Coinyuppie is an open information publishing platform, all information provided is not related to the views and positions of coinyuppie, and does not constitute any investment and financial advice. Users are expected to carefully screen and prevent risks.

Like (0)
Donate Buy me a coffee Buy me a coffee
Previous 2021-06-05 03:50
Next 2021-06-05 03:54

Related articles