Millimeter-wave radar: the real “center of attention” in the three-piece set of autonomous driving?

If you were to take a self-driving car on the road without a safety officer , and it was driving at high speed, what would you be most concerned about? Worried that the car would get lost? Drive too slowly? Or feel lonely on the road?

Probably not. What you care most about is whether the car is safe. Compared with traditional cars that rely on the driver's vision to ensure driving safety, self-driving cars rely on a large number of sensors to exercise the car's "vision" function, and the higher the level of self-driving, the more redundant sensor systems will be used.

Safety and reliability that exceeds that of human drivers is the fundamental prerequisite for autonomous driving.

At present, to ensure safety, the autonomous driving sensor system is mainly composed of three sets of cameras, laser radars , and millimeter-wave radars . These three sets can be found in almost all existing unmanned driving sensor solutions.

However, in vehicles that are still in the assisted driving stage, the three-piece set of millimeter-wave radar, lidar, and camera is not set in stone. Tesla, which insists on taking the camera route, has been repeatedly dissed by its head Musk, but the industry has fought back, believing that only lidar is the C position of the autonomous driving sensor system.

In contrast, millimeter-wave radar is the least controversial of the three sensors. Millimeter-wave radar is used in all current solutions that support autonomous driving and advanced driver assistance systems (ADAS).

Millimeter-wave radar has advantages such as long transmission distance, stable performance, and controllable cost, but it also has defects such as weak angular resolution and low recognition accuracy. In the context of multi-sensor fusion in existing autonomous driving, it is important to understand the advantages and disadvantages of millimeter-wave radar, as well as the latest technological evolution trends and its value in the autonomous driving industry, to determine whether we should focus on the development of millimeter-wave radar.

Balancing the two benefits: the application advantages of millimeter wave

Millimeter-wave radar has been used in the automotive field for many years. It was initially added to automotive sensors mainly to achieve blind spot monitoring and fixed-distance cruise control. With the development of technology, these two features have gradually spread from high-end models to almost all models. With the demand for high-precision perception of the driving environment by autonomous driving and ADAS, millimeter-wave radar has played an important role in it with its many advantages.

The so-called millimeter wave refers to an electromagnetic wave with a wavelength between 1 and 10 mm. The wavelength of the millimeter wave is between the centimeter wave and the light wave. Therefore, the millimeter wave has the advantages of both microwave guidance and photoelectric guidance.

Millimeter-wave radar refers to radar that works in the millimeter-wave band. Millimeter-wave radar transmits millimeter waves through the antenna, receives the target's reflected signal, and quickly and accurately obtains the relative distance, speed, angle, direction of movement, etc. between the car body and other objects after calculation, and then returns it to the vehicle's central processing unit (ECU) for intelligent processing and decision-making.

The working principle of LiDAR is to use laser as a signal source, and the laser beam emitted by the laser is used to detect the target's distance, direction, height, speed, posture and other characteristic quantities. Since the laser beam continuously scans the target, all data points on the target can be obtained, and after 3D imaging processing, an accurate three-dimensional stereo image can be obtained.

What are the obvious advantages and disadvantages of the two performance characteristics?

First, in terms of detection accuracy and resolution, LiDAR is significantly better than millimeter-wave radar. For example, when both millimeter-wave radar and LiDAR detect an "obstacle" on the road, the former may only "see a vague shape, while the latter can clearly distinguish" whether the obstacle is a shoulder or a slope. Once the vehicle determines that it is a slope, it can make a decision to move forward safely.

其次，在抗环境干扰上，毫米波雷达则显著优于激光雷达。由于激光雷达使用的是光波段的电磁波，透射与绕射性能补强，在遇到雨雪、雾天、雾霾、灰尘等环境，其探测性能将大幅下降。而相比较于光学传感器，处于毫米波波段的电磁波则不会受到雨、雾、灰尘等常见的环境因素影响，因此，毫米波具有全天候（除大雨天气外）、全天时的强抗干扰的探测性能。

In terms of anti-signal source interference, unlike millimeter-wave radar, which is easily affected by electromagnetic waves in nature, there are very few sources in nature that can interfere with laser radar. Therefore, laser radar has a stronger ability to resist signal source interference.

In addition, in terms of detection distance, since millimeter waves have weak attenuation in the atmosphere, they can detect and sense farther distances. Medium and long-range millimeter-wave radars can reach a detection distance of 250 meters, while laser radars can only reach a maximum of 200 meters. Therefore, during high-speed driving, millimeter-wave radars can judge the status of obstacles ahead earlier than laser radars, and play a role in safety warnings or emergency braking judgments.

In addition, in terms of manufacturing process and cost, millimeter-wave radar is significantly better than laser radar. Due to the short wavelength of millimeter waves and small antenna diameter, millimeter-wave radar has the characteristics of small size, light weight, easy integration, and is easy to install on the car; while laser radar has complex internal structure, high requirements for manufacturing process, large product size, high installation difficulty or poor aesthetics. At the same time, in terms of cost, the price of millimeter-wave radar can be controlled at around 1,000 yuan, while laser radar with better performance still costs tens of thousands of yuan.

At present, for LiDAR, in addition to miniaturization and aesthetics, reducing costs has become the top priority for manufacturers pursuing mass production. For millimeter-wave radar, improving its measurement resolution and accuracy has become a top priority.

Ultra-long distance + high resolution, new advantages of millimeter-wave radar upgrade

Compared with lidar and cameras, millimeter-wave radar is undoubtedly widely used in autonomous driving vehicles due to its long measurement distance, stable operation around the clock and low cost, but its shortcomings in detection accuracy also need to be compensated by continuous technological iteration.

According to the millimeter wave frequency band classification, there are currently three main frequency bands for automotive millimeter wave radars, 24GHz, 77GHz and 79GHz. The former is mainly responsible for short-range detection, and the latter two frequency bands are mainly responsible for medium and long-range detection.

The first trend is that major manufacturers around the world are now focusing on the application of 77GHz millimeter wave radar and making technological breakthroughs in the 79GHz frequency band. Compared with 24GHz radar, 77GHz millimeter wave radar is smaller in size; in addition, it can simultaneously meet high transmission power and large working bandwidth, enabling it to achieve long-distance detection and high distance resolution at the same time.

The leading advantage of 77GHz millimeter-wave radar also means that the technical threshold for realization is very high. The design and manufacturing of antennas, RF circuits, chips, etc. are more difficult. Currently, only a few companies in the United States, Japan and other countries have mastered it, and domestic manufacturers are in the stage of catching up. 24GHz millimeter-wave integrated circuits have been mass-produced and tried, and the localization of 77GHz millimeter-wave radar chips is still in progress.

The 79GHz band has a bandwidth more than three times higher than the 77GHz band and has a higher resolution. However, it has not yet been put into large-scale mass production, and domestic and foreign companies are still at the same starting line.

The second trend is that CMOS (complementary metal oxide semiconductor) technology is becoming the mainstream in the system integration process of millimeter wave radar. In addition to reducing costs, CMOS can mainly integrate additional digital modules such as MCU and DSP, so that the control of radar chips and even digital signal processing can be completed locally without the need for dedicated processors, reducing system complexity and cost.

Another important trend is the improvement of the spatial resolution of millimeter-wave radar. In blind spot monitoring, high-resolution millimeter-wave radar needs to achieve the transformation from only judging whether there are objects within a safe distance to forming environmental modeling and judging the shape of specific objects (people or cars, etc.) corresponding to each point in the radar point cloud. The way to achieve this feature is to increase the number of integrated transceivers in the millimeter-wave radar chip.

Improving system integration and increasing the number of transceivers represent two technical application directions for autonomous driving. The former is mainly aimed at assisted driving. Since it is more sensitive to cost and radar module complexity, assisted driving cars are more concerned about the reduction of module complexity brought by CMOS system integration. L4-L5 autonomous driving is more concerned about the spatial resolution of millimeter-wave radar to obtain higher-precision point clouds, so it is more concerned about the number of transceivers.

From the above trends, we can see that millimeter-wave radar is trying to make up for its shortcomings in terms of detection distance, high-precision resolution, and spatial resolution to improve detection accuracy, thus challenging LiDAR. At the same time, LiDAR is also trying to reduce costs as much as possible to consolidate its market share. Therefore, the two will continue to be used in combination for a long time in the future, forming a multi-sensor fusion application trend with cameras, ultrasonic sensors, etc.