How 4D mmWave Radar Helps Smarter Cars

How 4D mmWave Radar Helps Smarter Cars

At present, as the world’s major car companies have entered the field of electric vehicles, it reflects the determination of various countries in the world to replace fuel vehicles with electric vehicles.

This major change directly promotes the comprehensive combination of vehicle electrification and intelligence. The autonomous driving system has become the most important battlefield for major car companies, and the advent of 4D radar has become a major breakthrough in the field of radar.

most obvious difference between 4D imaging millimeter-wave radar and traditional millimeter-wave radar is that it increases the perception of target height information, followed by a substantial improvement in perception fineness. Traditional millimeter-wave radars were mainly used to detect the distance, relative velocity and azimuth of objects due to limited elevation channels and processor performance limitations, which is the meaning of “3D” in the past. However, in terms of altimetry, the traditional radar capability is very weak, and it can even be said that it has no altimetry capability at all. The result of this is that when there is a static obstacle on the road, although the radar can detect the reflection point of the obstacle, it is often difficult to make an accurate judgment because it cannot identify the height and size of the obstacle. Others include the wrong perception of static targets such as road manhole covers, speed bumps, street signs, and overpasses, which leads to the general low confidence of millimeter-wave radar in fusion systems.

In order to reduce unnecessary braking and improve the user experience, some radar companies, autonomous driving companies or car companies will choose to directly filter out such static targets perceived by radar in the actual application process. Maybe in most scenarios, this strategy is enough to deal with, but for some uncommon Corner Cases, it seems a little powerless. In particular, many L2-level autonomous driving systems use vision as the main sensor and millimeter-wave radar as auxiliary perception. If obstacles are found visually, it is okay. If they are not found in time, they have to rely on millimeter-wave radar, which is prone to problems.

Compared with traditional millimeter-wave radar, 4D radar can not only provide height information of the target and achieve higher angular resolution due to its more channels and more advanced processors, but also output dense point cloud information to outline the surrounding area. The outline of the object, that is, the point cloud imaging.

In terms of angular resolution, the current 4D millimeter-wave radar can generally achieve a horizontal angular resolution within 1° and a pitch angle resolution within 2°, which is particularly important for improving driving safety. Taking ACC as an example, according to relevant sources, if you want to accurately distinguish two vehicles 300 meters away, the horizontal angular resolution must be less than 1°. And if you want to identify a traffic light that hangs about 6.5 meters 150 meters in front of the car, the angular resolution needs to reach 2° to meet the demand.

In terms of point cloud output, the density of traditional radar point cloud is very sparse, and it is basically impossible to classify obstacles, while 4D radar has the point cloud output capability similar to lidar, which can help analyze the contour and category of the target, and improve the perception reliability of the entire system. At the same time, it can also be redundant with the camera perception, as a classification supplement when the image sensor fails. Thanks to the significant improvement in altimetry capability, angular resolution, point cloud density, etc., 4D millimeter-wave radar is being considered as an important support for the evolution of autonomous driving from L2 to L3 or even higher-level L4/L5.

Share this post