As long as we talk about autonomous driving, the most important and basic factor is perception.
Simply put, if you want a car to achieve autonomous driving, you need to allow the car to "see clearly" or collect traffic environment data around the vehicle, so that it can make decisions at any time based on the "seen" environment, so that the self-driving car can Safe Driving. So, if you want to "see" the surrounding environment clearly, you cannot do without the hardware equipment such as millimeter wave radar , lidar and vehicle camera installed on the vehicle.
We all know that the observation ability of a camera is more susceptible to extreme weather, light and other factors than the human eye, so it is usually used with millimeter wave radar. However, this solution has low resolution and insufficient data acquisition, so it has been abandoned by most car companies, and many car companies have turned to the lidar camp.
Although lidar can make up for the "blind spot" problem of vision algorithms, for now, its high cost limits its large-scale mass production application. It was under this circumstance that 4D imaging radar was born, and it seems promising to become a compromise between millimeter wave radar and lidar.
So, what is 4D imaging radar? Can it really replace lidar? Which 4 "D"s are
D?
In fact, 4D imaging radar can be regarded as an advanced version of millimeter wave radar. Millimeter wave radar is a relatively mature technology. It can calculate the distance, orientation and speed of an object by emitting and receiving reflected electromagnetic waves. This is what we learned in high school physics. Puller effect .
Distance, orientation, and speed also constitute the meaning of 3D, so the millimeter wave radar we are familiar with is already 3D. But this 3D does not have the same meaning as what we commonly call "space 3D", so do not confuse this.
Traditional millimeter wave radar can already detect static obstacles and can accurately know the distance, orientation and speed information between the target and radar. However, because it does not have the ability to measure height, it is difficult to judge whether the stationary object ahead is on the ground or on the ground. up in the air. Car cameras can only capture 2D plane images. Even with the assistance of deep learning, they still cannot accurately measure the distance between surrounding objects and self-driving cars. Therefore, this is also part of the cause of early "autonomous driving" accidents caused by system misjudgments.
Therefore, while camera technology is developing rapidly, improving the judgment ability of millimeter wave radar has also been put on the agenda. After all, for millimeter wave radar, the most difficult thing is to judge stationary objects. How to distinguish small stationary objects such as manhole covers and speed bumps from tall "air obstacles" such as traffic signs and gantry has become the reason why millimeter wave radar has been upgraded and evolved again. Because of this, 4D imaging radar came into being.
D imaging radar can be regarded as an advanced version of millimeter wave radar. On the basis of ordinary millimeter wave radar, height data is added, which is the meaning of "3D+1D". With the data of an additional height detection value, millimeter wave radar not only improves the "vision" range, but also enhances the detection range and resolution capabilities.
By integrating the fourth dimension of height into traditional millimeter wave radar, the environment can be better understood and mapped. By making the measured traffic data more accurate, the contours, behaviors and categories of measured targets can be effectively analyzed to adapt to more complex roads. If more small objects, occluded objects, and stationary or lateral objects can be identified, it will be possible to accurately understand the circumstances under which the vehicle needs to brake, thereby providing more reliable information for decision-making and planning.
Take Huawei 4D imaging radar as an example. This radar uses 12 transmitting channels and 24 receiving channels. The overall number is 12*24, which is 288 channels. It is 24 times higher than the conventional millimeter wave 3 transmitting and 4 receiving antenna configuration. times.More channels mean more points can be formed, resulting in higher resolution. Through these points, the outline, category, and behavior of the target can be identified, which is the meaning of imaging.
In terms of performance effects, 4D imaging radar can be regarded as an upgraded version of 3D millimeter wave radar. On the other hand, from a cost perspective, the cost of 4D imaging radar is only 10%-20% of that of lidar. Therefore, after lidar, 4D imaging radar is very likely to become the next hot spot in the autonomous driving sensor market. Can
replace lidar?
In fact, as early as 2018, Texas Instruments proposed the concept of 4D imaging radar. It was not until 2021 that 4D imaging radar made its "official debut." For a time, technology giants such as Continental, Waymo, Aptiv , Huawei and other technology giants have launched new products. At the same time, many emerging players such as Aoku, Suzhou millimeter wave, and An Zhijie have bet on the layout. However, the mass-produced cars equipped with this technology have not yet been launched, and lidar has already begun to be delivered for use.
In terms of working principle, lidar and millimeter wave radar are basically similar, both use echo imaging to display detected objects. However, the electromagnetic wave emitted by lidar is a straight line, using light particle emission as the main method, while the electromagnetic wave emitted by millimeter wave radar is a cone-shaped beam. The band antenna mainly emits electromagnetic radiation .
From the perspective of detection accuracy, lidar has the advantages of high detection accuracy, wide detection range and strong stability. In terms of accuracy, the detection range of millimeter-wave radar is directly restricted by frequency band loss. It cannot detect pedestrians, non-motorized vehicles, etc., and it cannot accurately model all surrounding obstacles. From this point of view alone, the accuracy of 4D imaging radar is not as good as lidar.
In terms of anti-interference ability, since lidar detects by emitting light beams, it is greatly affected by the environment. Once the light beam is blocked, it cannot be used normally, so it cannot be used in bad weather such as rain, snow, haze, and sandstorms with low visibility. Open in . Millimeter-wave radar has a strong ability to penetrate fog, smoke, and dust, so it can be used in bad weather. In terms of anti-interference ability, 4D imaging radar can win again.
Therefore, lidar and millimeter wave radar actually have their own advantages and disadvantages, and currently neither can completely replace the other. The two even have a complementary relationship and learn from each other's strengths and weaknesses. However, in the face of the high cost of lidar, the cost-effectiveness of 4D imaging radar is highlighted.
The cars we can currently buy on the market are basically equipped with L2 level driving assistance, as well as some pre-embedded L3 level driving assistance systems. In fact, they do not require such high wire harness lidar. At this time, if 4D imaging radar can assist in the implementation of L3 functions, it can be widely used.
From the perspective of lidar, manufacturers of high-precision automotive lidar products are mainly concentrated abroad. Judging from the industrial layout of millimeter wave radar, it is currently mainly controlled by overseas giants, and it is still in its infancy in China. Lidar has high accuracy, but the cost is also high, dozens of times that of millimeter wave radar. Therefore, millimeter wave radar is also widely used due to its higher cost performance and other performance advantages.
Of course, 4D imaging radar is still facing many technical difficulties. In addition to issues such as the consistency and reliability of the product itself, in a large array with multiple channels, how to ensure phase calibration is also a test. This includes horizontal angles and pitch angles, making calibration more difficult.
Master’s observation
Judging from today’s technology, it is still quite difficult to achieve fully autonomous driving. The emergence of 4D imaging radar may reduce or replace the use of lidar to a certain extent, but it is not easy to achieve mass production and installation.Limited by the upstream material supply chain, 4D imaging radar is still in the early stages of development, and there is still a long way to go before true large-scale commercial application.
