Alternative lidar? The "Heat" and "Pain" of 4D Imaging Millimeter Wave Radar

4D imaging millimeter-wave radar has recently heated up again.

At this year's CES, 4D imaging millimeter wave radar is gaining momentum, and a number of chip companies such as NXP, TI, and Mobileye have successively launched or updated their own imaging radar solutions, and millimeter wave radar system manufacturers Arbe, Zadar Labs, Smartmicro, etc. have also brought their own imaging radar products.

One of the most concerned by the industry is Mobileye CEO Amnon Shashua's emphasis on 4D imaging millimeter-wave radar in his CES speech, "(By 2025) in addition to the front, we only want millimeter-wave radar, not lidar." ”

Pierre Cambou, chief analyst at Imaging at Yole Développement, said the speech sparked speculation that Mobileye doesn't now think lidar is "more important than radar."

Danny Kim, partner and director of consulting services at VSI Labs, wrote in a post-CES report: "Unlike past CES activities, it feels like lidar companies have not brought so many breakthrough inventions to the industry." "On the other hand, 4D millimeter-wave radar is becoming increasingly popular," adding that 4D millimeter-wave radar is becoming a "new star" in automotive sensors.

In fact, under the entry of Huawei before, 4D imaging millimeter wave radar has long been subject to a wave of hot discussion in the market, and it is far from a new technology, but why is it being focused on at CES this time? What factors have changed? Can 4D imaging millimeter wave radar really PK with lidar? Or is it just a transitional technology solution?

Imaging radar is in the limelight

Yole summarized Mobileye's talk into four points, two of which were:

4D imaging millimeter-wave radar is an important enabler for consumer-grade autonomous vehicles;

Lidar is no longer the key.

Mobileye targets three segments: assisted driving, the geo-fenced L4 Robotaxi market, and the consumer-grade L4 Robotaxi market.

Shashua believes that if Mobileye's perception solution is to reach the level of consumer-level autonomous vehicles in 2025, one is how to significantly reduce costs, and the other is how to expand the ODD (operational design domain) to the level of L5.

Therefore, in Mobileye's plan, in addition to developing a consumer-grade autonomous vehicle program that can drive on camera alone, by 2025 they may also launch a consumer-grade autonomous vehicle solution that can travel separately on radar/lidar, which is equipped with a radar-LiDAR subsystem, when the vehicle only needs a forward lidar and a 360° all-inclusive millimeter-wave radar.

"Except for the front, we only want millimeter-wave radar, not lidar." Shashua said.

The problem is that while the cost of millimeter-wave radar and lidar is not an order of magnitude, at the current level of resolution of millimeter-wave radar, ordinary millimeter-wave radar cannot distinguish between pedestrians and vehicles that are very close to each other in congested traffic, and therefore cannot be used as a separate subsystem.

Mobileye turned his attention to 4D imaging millimeter-wave radar.

In terms of performance effect, 4D imaging millimeter wave radar is an upgraded version of 3D millimeter wave radar, on the other hand, from the cost point of view, the cost of 4D imaging millimeter wave radar is only 10%-20% of lidar.

Compared with the traditional 3D millimeter wave radar, the vehicle-mounted 4D millimeter wave radar can not only solve the distance, speed, and horizontal angle information of the target, but also solve the pitch angle information of the target, which can provide environmental information around the car, and can avoid the false alarm phenomenon caused by manhole covers, road shoulders, and speed bumps.

In addition, 4D millimeter wave radar can also provide more realistic path planning and passable space detection functions thanks to the ability to provide altitude information about the target and capture the spatial coordinates and velocity information of the target around the car.

Some insiders said to the new intelligent driving, the traditional millimeter wave radar is also a bit cloudy but the number is small, and there is no pitch information, 4D millimeter wave radar increases pitch information and more point cloud data, "point cloud can outline the outline of the object, that is, imaging."

However, whether 4D millimeter-wave radar needs to be imaged depends on the strategy of specific scene needs.

According to the above-mentioned industry insiders, the current L3 above the scene will have a greater demand for imaging radar, from a technical point of view, 4D imaging millimeter wave radar is an inevitable trend, the ability to do 4D millimeter wave radar manufacturers are basically doing 4D imaging millimeter wave radar.

"4D millimeter wave radar is the future development direction, and the focus of imaging is that there is enough point cloud data, whether to use 4D millimeter wave radar as an imaging output on the car depends on the overall comprehensive consideration of the main engine factory for sensor fusion, computing power and other factors."

New Intelligent Driving has been in "Just now, another smart car is available!" The main selling point is "4D millimeter wave radar", which carefully introduces the working principle of 4D millimeter wave radar.

Angular resolution as the pointing accuracy of radar, its numerical level and wavelength and aperture size are related, that is, the longer the wavelength, the lower the angular resolution, the larger the aperture, the higher the resolution.

Chu Xuanyan, founder and CEO of Chuhang Technology, a manufacturer of vehicle-mounted millimeter-wave radar systems, introduced to the new intelligent driving that the aperture size is the key to improving the resolution of the radar angle, and the number of antennas and the spacing between the antennas will affect the aperture size.

Over the past few decades, the world of vehicle-mounted millimeter-wave radar has often improved angular resolution by increasing the number of antennas.

At present, the common working mechanism of vehicle 4D millimeter wave radar is the frequency modulated continuous wave radar (FMCW) in continuous wave radar, which can continuously emit electromagnetic waves to the outside in a lower power consumption and larger bandwidth, so as to achieve the distance and speed information of the measurement target.

Depending on the number of input and output antenna arrays, FMCW radar can be divided into single input multiple output (SIMO) radar and multiple input multiple output (MIMO) radar.

For automotive millimeter wave radar systems, SIMO radar has long been widely used in 3D millimeter wave radar, and the MIMO radar concept was first proposed by Bliss and Forsythe in 2003, which is one of the key technical theories for the development of vehicle 4D millimeter wave radar.

Different from the traditional approach, in order to solve the problem of low angular resolution and low point cloud density of traditional millimeter wave radar, four 4D millimeter wave radar solutions have emerged:

First, based on the traditional CMOS radar chip, emphasizing "software-defined radar", the main manufacturers are Aoku, Mobileye, etc.;

The second is to integrate multiple transmission and multiple receptive antennas into a chip to directly provide imaging radar chips, such as Arbe, Vayyar, etc.;

The most traditional is to cascade standard radar chips with multiple chips to increase the number of antennas, such as Continental, Bosch, ZF and other companies;

The fourth is the development of new radar architectures through metamaterials, representing manufacturers such as Metawave.

Shashua believes that software-defined image millimeter-wave radar will be the key to improving the resolution of millimeter-wave radar.

The so-called software-defined radar, that is, through software post-processing, the system can be configured to receive/transmit radar signals and process, thereby greatly improving the performance of radar.

In fact, as early as two years ago, Mobileye has begun to build this image radar, and at CES 2021, Shashua has also introduced it, but it has taken far less time than this year's CES.

Is imaging radar just a transitional scenario?

From the data released by Mobileye, the performance of its imaging millimeter-wave radar is indeed remarkable in many similar products.

According to the data, Mobileye's software-defined radar will have more than 2,000 virtual channels, 48 signal transmitters and receivers, a horizontal angle resolution of 0.5°, a vertical resolution of 2°, and an effective detection distance of 150 meters.

"Mobileye's introduction of imaging radar is similar to that of The Arbe, because the current mainstream methods (including bosch, continental, and ZF mass production solutions) are all traditional cascading methods, and it remains to be seen whether the new solutions have both cost and performance advantages for traditional methods." But another insider said so about the new smart drive.

Here it is necessary to first understand the core devices that make up the vehicle-mounted millimeter-wave radar.

The core devices of vehicle-mounted millimeter-wave radar mainly include single-chip microwave integrated circuits and radar digital signal processing chips.

Monolithic microwave integrated circuits can implement low-noise amplifiers, mixers, frequency converters, power amplifiers and other functions, and the main players include ST, Texas Instruments, NXP, Gatland and so on.

Radar digital signal processing chip, is used for the millimeter wave radar medium frequency signal digital processing, divided into general digital processing chip and radar special processing chip, the current chip manufacturers that provide special radar processors are mainly Texas Instruments, NXP, Infineon, Gatland and so on.

In recent years, with the increase in market demand for vehicle-mounted millimeter wave radar, a number of emerging millimeter wave radar system manufacturers have emerged in China, such as Sensestar, Huayu, Lingbo Microstep, Hayabe Technology, Chuhang Technology and Geometric Partners.

In 2018, Texas Instruments proposed the concept of 4D imaging millimeter wave radar and launched a complete design scheme of four-piece cascaded 4D millimeter wave radar based on the AWR2243 FMCW single-chip transceiver.

This "turnkey" project has greatly reduced the threshold for enterprises to develop imaging millimeter wave radar products, and some millimeter wave radar system manufacturers have also begun to develop their own 4D millimeter wave radar products.

At present, the manufacturers who have launched 4D millimeter wave radar products mainly include Continental, Auco Radar, Arbe and Smartmicro.

The above-mentioned industry insiders told New Intelligent Driving that the development of 4D imaging millimeter-wave radar has gradually become an inevitable choice for millimeter-wave radar industry giants and startups.

The performance battle between 4D imaging millimeter-wave radars has become increasingly fierce.

At present, the on-board 4D imaging millimeter wave radar is more than 48 channels (6 hairs and 8 receipts), and some radar suppliers are also moving towards more than 192 channels, such as Sentec's 4D imaging millimeter wave radar STA77-8, Continental's 4D imaging millimeter wave radar ARS540, Aura, etc., huawei's 4D imaging millimeter wave radar has 288 channels (12 shots and 24 receipts).

However, there are also products comparable to the number of Mobileye imaging radar channels, which is a 4D imaging millimeter wave radar Phoenix provided by Israeli startup Arbe, which uses 48 rounds and 48 receipts, and more than 2000 virtual channels, providing 1° horizontal × 1.5° vertical angle resolution.

Huawei's intelligent car solution BU Marketing and the president of the sales service department have also previously revealed, "Huawei's 4D imaging millimeter wave radar antenna has achieved 128 rounds and 128 receipts, which is very advanced, and Huawei's millimeter wave radar is developed on the basis of communication technology."

Amba Semiconductor acquired Aoku Radar last year, and its vice president of marketing in China, Jianjun Xi, told New SmartDrive that the performance of its 4D imaging millimeter wave radar products is now similar to that of 32-wire-beam lidar.

"We will continue to improve the angular resolution to achieve 4-chip cascade, similar to the effect of 128-beam lidar resolution."

According to Xi Jianjun, the current cost of 4D imaging millimeter wave radar is similar to the cost of traditional millimeter wave radar, which is far lower than the cost of lidar.

"It takes at least 5 years for the cost of lidar to go from a few thousand yuan to a few hundred yuan, so 4D imaging millimeter-wave radar can replace low-beam lidar to some extent."

In addition, when 4D millimeter wave radar is combined with a multi-eye camera, it is theoretically considered that lidar can be completely eliminated.

So when the cost of lidar with higher ranging accuracy drops to an acceptable range, will 4D imaging millimeter-wave radar be obsolete? In other words, will 4D imaging millimeter-wave radar be a temporary "transitional solution"?

Chu Xuanyan's view of this is that 4D imaging millimeter wave radar and lidar have their own characteristics and can play their respective strengths, and there is no relationship between who replaces whom.

"For example, some L3-level functions do not need such a high-beam lidar, at this time, you only need to increase the resolution of millimeter-wave radar slightly, the cost will not increase too much, it is possible that many L3-level functions can land, which will be the main market for 4D imaging millimeter-wave radar."

Xi Jianjun believes that because ordinary millimeter-wave radar is not effective in detecting stationary objects and the angular resolution is low, the current imaging radar is still more of an alternative to traditional millimeter-wave radar.

Who will be the primary sensor?

But there is currently no 4D imaging millimeter wave radar product that can really land on a large scale.

Chu Xuanyan introduced that there are two difficult problems faced by this.

The first is that the demand for 4D millimeter wave radar is not clear.

Auto parts companies are mostly demand-oriented when developing products, but at present, car companies are not sure which function is in L3 autonomous vehicles, which requires 4D imaging millimeter wave radar, or only needs to output point clouds of 4D millimeter wave radar.

"In addition, as for autonomous driving functions such as AVP, HWP (high-speed automatic driving), TJP (medium and low speed automatic driving), there is no accurate definition of how high the resolution requirements of 4D imaging millimeter wave radar are, whether it is 1 ° or 2 °."

There is also the 4D imaging millimeter wave radar, although it can output more points, but the current millimeter wave system manufacturers have not figured out, "after exporting these points, what do I want to do."

Originally, millimeter-wave radar only outputs targets with distance and speed information, but when 4D imaging millimeter-wave radar also provides additional points with azimuth information, there is no clear solution in the industry for what kind of perception purpose enterprises want to use this feature.

The second major difficulty is that there is no test equipment specifically for 4D imaging millimeter wave radar in the industry, and the ecological chain of the industry is not mature.

Manufacturers can only use traditional millimeter-wave radar test equipment to verify the performance of their 4D imaging millimeter-wave radar products.

But the problem is that traditional test equipment such as the target simulator, whose resolution is not high, cannot verify whether the resolution of the 4D imaging millimeter wave radar reaches 1° or a few tenths of a degree.

"We can only rely on the road, with the truth value system, with the lidar to do the comparison." But in this case, if some of the basic performance of the product is wrong, such as antenna design, we simply can not detect it in the laboratory, can not do front-end testing, can not form a complete research and development closed loop. ”

However, the trend of 4D imaging millimeter wave radar has been irreversible, such as BAIC Group invested in Arbe, and its deputy general manager Chen Jiang has previously said that BAIC Group models are expected to be equipped with Arbe mass-produced 4D imaging millimeter wave radar products, such as SAIC Motor also announced last year that it has equipped 4D imaging millimeter wave radar in its R car, and then such as the Fisker Ocean model that Magna and Fisker will launch at the end of 2022, which will also be equipped with 4D imaging millimeter wave radar and so on.

Xi Jianjun also told Xinzhi Driving that its 4D millimeter-wave radar has been targeted by a number of OEMs.

"This year we see several OEMs ready to get on the 4D imaging millimeter wave radar, which is the same as the lidar of a few years ago, once there is a leading company on the large-scale 4D imaging millimeter wave radar, the companies behind will quickly follow." Xi Jianjun said.

As for the construction of the 4D imaging millimeter wave radar ecological chain, many enterprises, including Chuhang Technology, are working together to build industry standards, such as radar performance requirements and test methods.

"Everyone is looking in the direction of 4D imaging millimeter wave radar, although the development route is different, but I think in 1-2 years, the industry will form a simple standard, there will be attempts to land."

Chu Predicted that it will take 3-5 years for 4D imaging millimeter-wave radar to form a standardized, large-scale mass-produced product.

This may also be why Mobileye spent more ink and time at this year's CES to emphasize the importance of 4D imaging millimeter-wave radar.

Going back to the previous proposal by Mobileye, they will launch an AV solution that can travel on radar/lidar alone in 2025, that is, av vehicles will only need a forward lidar and a 360° all-encompassing millimeter-wave radar.

This means that 4D imaging millimeter-wave radar will become the main sensor for autonomous vehicles.

Is this approach feasible?

Chu Believes that it is feasible, but because millimeter-wave radar and lidar have weak perceptions of color and two-dimensional objects, the premise of this solution is that the vehicle can achieve vehicle-road coordination to accurately perceive the environment around the body.

However, cameras, lidar or 4D imaging millimeter-wave radar, who can become the main sensor of the future intelligent driving vehicle, may still depend on the specific scene, according to different functions and needs to configure different programs.

Chu Said yan said that for example, for the ADAS function, the mainstream standardized 4D millimeter wave radar products, 3 rounds and 4 receipts are enough, for the L3 level functions, 12 rounds and 16 receipts of 4D imaging millimeter wave radar products can meet the perception needs of the vehicle.

When the vehicle needs to be upgraded to the level of L4 or L5, such as mines and ports, these scenarios are not so sensitive to price, and manufacturers can choose lidar or 4D imaging millimeter wave radar according to their own needs to do a differentiated layout.

For example, an unmanned logistics car launched by Great Wall Motors did not use lidar, but was equipped with 5 (1 main radar + 4 angle radar) 4D imaging millimeter wave radar.

Different sensors have their own advantages and disadvantages, and the perception scheme of intelligent driving vehicles is far from fixed.

For 4D imaging millimeter-wave radar, in order to survive for a long time, the most important thing at present may be to solve the problem of whether it can be on a large scale.

Leifeng Network # Leifeng Network # Leifeng Network