4D Imaging Radar
When we talk about when 3D capture, always first think of optical sensors . When we talk about (time) to discuss the visual data in the fourth dimension, we tend to consider the scene data scheduling. These are our concern over the years lidar (LiDAR) and photographic measurement , and users for large projects slow-moving, on the time scale of these techniques to prejudice caused by static objects.
On automatic pilot car demand contributed to the incredible variety of sensor development, because we can not just get an application in sensor technology, no changes will be put into other applications. Like 4D imaging radar such new sensor, which uses echolocation (such as dolphins, bats, or some method may determine the position of the object) and time of flight ( the ToF) measurement principle to capture 3D spatial information. In addition, they are also used in fast-moving car or fast flying UAV , the time scale to achieve imaging.
Radar vs. laser radar
According to Mai Musi Consulting reported that recently Sensors Online article presents a rather persuasive argument: The 4D transducer for achieving Level 4 and is essential for Level 5-speed automatic driving. In this article, Arbe Robo Ti , CEO and co-founder of cs Kobi Marenko explains "Without the help of 4D imaging radar, optical sensors can not achieve Level 4 and Level 5-speed automatic driving causes". Before further discussion of this issue, briefly explain the autopilot rating: - Level 0 means no automation, just like a car manual: The driver needs its own control everything, while the car itself can not make any judgment and control. - Level 1 ~ 3 an increase of varying degrees of automation, current Tesla autopilot level between Level 2 to Level 3 between which the guide autopilot system may be implemented, acceleration, braking, driving control can sometimes itself (although this should not exist). - Level 4 means that the car can be run without the driver's control of the situation, but only under certain conditions. For example, a university campus running autonomous vehicles. - Level 5 means that the car can be fully realized autopilot, you rest or take a nap in the car, your car will be able to send you home safe. From the above we can see the importance of the automation level car radar. Compared with the camera and the laser radar, 4D imaging radar can work under any conditions, may be provided"In the fog, heavy rain, darkness and include air pollution highest detection reliability under adverse weather and other environmental conditions ." 4D imaging radar sensing area may reach 300 meters, and to capture the object may be displayed relative to the vehicle is close to or away from the Doppler shift, which can meet the higher requirements of automotive automation level.
Autonomous vehicles lidar survey plan
4D imaging radar: "bottom pressure " technology? ! It is noteworthy that, Marenko do not think 4D imaging radar capable of handling independent tasks alone. He believes 4D imaging radar only includes an optical sensor including automatic driving car part of the sensor system. This is a self-driving car on the old concept of 3D capture - that "each technology are a separate tool in the toolbox ", the current concept still applies. Marenko that: "furthest 4D imaging radar detection range for all sensors, which makes it possible to identify the first and dangerous,. 4D imaging radar can detect the camera and the laser radar sensor is guided to the relevant area, which will greatly enhance automatically driving safety. "Marenko last most convincing argument is the cost. The entire cost of the car automatically driving the sensor to be reduced to less than $ 1000, in order to achieve commercialization. In addition Marenko There are more extreme views: 4D imaging radar that allows autonomous vehicles completely out of the need for laser radar.
Autopilot implementation requires sensing layer, as well as the decision-making system to perform the layers cooperate with each other, by collecting and processing the environment information and the vehicle perception layer information, is transmitted to the decision-making, the decision-making based on information acquired by the determination decision, then back the execution level to perform the task.
The first step therefore is to achieve autopilot perception. Currently L2 of, L3 level laser radar autopilot mainly used, millimeter-wave radar and a camera as a sensor to receive information.
Camera ranging lack of capacity and vulnerable to extreme weather factors, such as light, which is often used with millimeter-wave radar use. However, this scheme is low resolution, and rich enough to obtain data. Although the laser radar can make up for these problems, but the high cost limits the large-scale application. In this case, many companies thinking about how to provide greater awareness for the autopilot. Which Waymo, Arbe, Echodyne is a pioneer in this field, they developed a new 4D imaging radar to make up for the lack of a millimeter-wave radar and laser radar.
Source: Arbe
4D imaging radar, highlighting the advantages
Typically, the radar is able to detect objects for distributing the electromagnetic waves. Such an object is called scattering body into the radiator and the distribution point radiator. Radar data collected by receiving information sent by the radiator, and passes the data to the decision-making.
However, around the observation target is often more than a heat sink, such as clutter is distributed in the body of the heat distribution around the target. Radar need to extract useful information clutter and other information.
In the currently available programs, the millimeter-wave radar that favorable conditions than the camera and laser radar, can adapt to extreme weather. But the disadvantages are also obvious, and that is clutter processing power is weak, and unable to locate the target distance information, a lower resolution.
And identifying defect resolution millimeter-wave radar capability, which will lead to too much information false positives. That is, if the radar can not raise the handling capacity of the clutter, it will cause significant underreporting problems, leading to safety issues.
In fact, the more clutter information provided, the better to make the right decisions. But the premise is that perception needs to have the ability to ascertain all targets within a certain area. Thus ultra-high levels of vertical and horizontal resolution is important.
In order to improve resolution and more accurate information is received, the company has developed a number of autopilot 4D imaging radar.
Compared with the camera and the laser radar, 4D imaging radar can work under any conditions, even under harsh weather and environmental conditions, fog, heavy rain, and air pollution black can also provide the highest detection reliability.
Radar market typically have 12 channels ( 3 rounds and 4 receiving), and 4D imaging radars use 2300 channels (transmit 48 * 48 receiving ). Channel array may provide a 1 ° azimuth resolution and 2 ° elevation resolution, the detection distance of up to 300 meters in range accuracy 10-30 cm, under a wide range of vision and can simultaneously track the remote target hundreds . Capture and display the object relative to the vehicle is close to or away from the Doppler shift, which can meet the higher requirements of automotive automation level.
In addition, 4D imaging radar on cost advantage over laser radar. At present, the sensor suite to achieve mass production, the cost should be less than less than $ 1000, but the current cost of some vehicle components and systems used in the testing phase and even the price of 100 times. It is understood that while the use 4D imaging radar used cost equivalent to the cost of using only a single element in the laser radar, and therefore it can help manufacturers achieve cost reduction targets.
Hella, etc. have incoming British LingFei
2020 on the CES, the world's millimeter-wave radar giant hella cool proud to announce a strategic investment, and the establishment of strategic cooperative partnership. And a month before that, the world's radar chip giant Infineon also announced strategic cooperation cool and proud, proud utilize cool 4D imaging radar technology for significantly improved its L2 / 3 ADAS tailored 77G single chip solution angular resolution performance.
It is reported that, in general, based on the hardware level, proud cool can rely on software algorithms to implement virtual aperture radar, antenna simulation many times, point cloud imaging and greatly improved angular resolution.
Source: Cool proud radar
Cool proud 4D imaging techniques may be implemented in the rest of the still detection probe, moving at low speed, the high speed of movement, the periphery of the vehicle, the pedestrian (with crossing vehicles, pedestrians) other high-definition holographic. Meanwhile, four FOV120 angle of the radar, may form a 360-degree view point cloud, detecting a radius of up to 200 meters.
4D imaging point cloud can greatly improve radar performance. Common angle radar, radar to radar before collaborative awareness and road vehicles, under the blessing of this technology, can appear a whole new look.
2019 Nian December, the Israeli company Vayyar complete D round of $ 109 million financing. Vayyar focus on the development of radar technology, the purpose of this round of financing to develop its 4D imaging radar technology.
The Vayyar 4D imaging radar technology mainly by the imaging system on chip ( the SOC), integrated on a single chip 72 and transmitter 72 receivers, covering the 3 GHz ~ 81 GHz band radar and imaging. With high performance integrated large memory the DSP, Vayyar sensor without any external CPU to perform complex imaging algorithms.
By using a broadband radio waves, Vayyar sensors can penetrate different types of materials, and can be run under any weather or light conditions, making it ideally suited for automotive and industrial markets.
Another house of Israel R & D 4D imaging radar technology is Arbe. In the same year in December, Arbe received investment from Beijing Automotive Industry Investment Group Co., Ltd. Yuen-ching capital and other institutions.
Arbe developed its own 4D imaging radar RFIC chip. The chip is based on the industry's first 22nm RF CMOS process products. RFIC chip with a self-developed algorithm and the original antenna design can provide finer than the original image is 100 times the accuracy of the image, it is possible to distinguish between different target size. Which resolution of 1 ° and the azimuth angle of 2 ° elevation, the width of the field of view reaches 100 ° in azimuth and 30 ° elevation angle, can be up to 300 meters per second range 30 (near real time ) to track hundreds of targets simultaneously .
Arbe CEO Kobi Marenko that the future of 4D imaging radar that allows autonomous vehicles completely out of the need for laser radar, which also will be upgraded to the core components redundant autopilot. Waymo, the Gates Foundation have been put into this track, accelerated autopilot landing.
Summary: 4D imaging radar having a high resolution can be acquired in any extreme environment to valid information, a millimeter-wave radar to make up for the defect. Meanwhile, at the cost lower than laser radar, but having the possibility of mass production. At present, whether giant companies, or start-ups, are trying to incoming 4D imaging radar in this area, and gained the support of capital. OEMs have started working with them, to promote the landing 4D imaging radar. With the advance of technology and technology cooperation, 4D imaging radar may gradually replace out laser radar, autopilot become a core part of.