What Autonomous Vehicle Sensors do Driverless Cars Use?
Author： Neuvition, IncRelease time：2021-03-12 02:09:26
Without human operating, driverless vehicles perceive the environment through autonomous vehicle sensors, which are like the eyes of a vehicle. Collecting basic information about the surrounding environment through the environmental perception of sensors is the basis of autonomous driving. There are many types of autonomous vehicle sensors, such as cameras, millimeter-wave radars, and LiDARs.
Camera can collect image information, and it is the closest to human vision. Through the collected images, through computer algorithm analysis, camera can identify a wealth of environmental information, such as pedestrians, bicycles, motor vehicles, road trajectories, curbs, road signs, signal lights, etc. Also, as an autonomous vehicle sensor, it can achieve vehicle distance measurement, path tracking through algorithm processing, to realize the forward collision warning (FCW) and lane departure warning (LDW).
Compared with other autonomous vehicle sensors, camera has been applied in advanced driver assistance systems (ADAS) in a large scale. The camera technology is mature and relatively inexpensive and is the first sensor to be widely used. The car collects image information around the vehicle by installing a camera and then analyzes the computer algorithm to realize the functions of object recognition and early warning.
Although the camera has been widely used, its shortcomings limit its application in the field of advanced driver assistance systems (ADAS). Its main shortcomings include: it lacks 3D information; it is greatly affected by the environment, for example, its recognition rate is greatly reduced under backlighting or complex conditions of light and shadow, and in the case of low visibility in bad weather conditions (night, rain, snow, heavy fog, etc.), so that it is difficult to achieve all-weather work.
Compared with cameras, millimeter-wave radars are more adaptable to the environment. Besides, as an autonomous vehicle sensor, it also has advantages including high resolution, good directivity, anti-interference, and good detection performance. Because the millimeter wave has a small attenuation to the atmosphere and good penetration of smoke, dust, etc., it is less affected by the weather.
Although millimeter-wave radar has better environmental adaptability as an autonomous vehicle sensor, its inherent characteristics limit its application in the field of advanced driver assistance systems (ADAS). Its main shortcomings include: the high cost of millimeter-wave radar components, relatively high requirements for processing accuracy, and small detection angle; the attenuation of high humidity environments such as rain, fog, and wet snow, and poor penetration of trees; the reflected waves are weak and difficult to identify for pedestrians and other non-metallic objects.
The key technology of advanced driver assistance systems (ADAS) – LiDAR is a system that transmits detection signals to the measured targets, and then measures the arrival time, intensity and other parameters of the reflected or scattered signals to determine the target’s distance, azimuth, movement status, and surface optical characteristics. As an autonomous vehicle sensor, the advantages of LiDAR include: 1) Very high range resolution, angular resolution, and speed resolution; 2) Strong anti-interference ability; 3) The amount of information acquired is rich, and the distance, angle, speed, and reflection intensity of the target and other information can be directly obtained, to generate multi-dimensional images of the target; 4) can work all day.
Compared with millimeter wave radar, LiDAR can detect the human body. Compared with a camera, LiDAR has a longer detection range. By scanning the surrounding environment, the point cloud data of the surrounding space is obtained, and the three-dimensional space map around the vehicle is drawn in real time to establish a decision basis for the next step of vehicle manipulation.
The maturity of LiDAR technology is not enough (not only MEMS LiDAR, but mechanical LiDAR also has the same problem), and unable to meet the requirements of vehicle regulations as an autonomous vehicle sensor. On the other hand, because the supply chain of LiDAR is not mature, the price of components is too high, and the cost of LiDAR is difficult to decrease. There is still a gap between the price of current LiDAR and the price expected by OEM manufacturers.
Types of LiDAR
There are two main types of LiDAR: mechanical LiDAR and solid-state LiDAR.
Mechanical LiDAR can directly drive the laser beam to realize the perception of the surrounding environment through mechanical rotating parts. Its advantages include greater measurement accuracy, a 360°field of view, and moving laser sources can emit higher power than solid-state laser sources, i.e. the detection distance is farther. However, mechanical LiDAR is not highly adaptable due to its complex mechanism and large volume, and it is difficult to achieve mass production due to high manufacturing costs.
The solid-state LiDAR mainly relies on electronic components to control the laser emission angle and does not require mechanical rotating components, so its structure is simple and the cost is lower. However, its disadvantages include: the scanning angle is limited, and more ambient light noise will be introduced because the solid-state LiDAR receives signals from an entire surface, and the measurement accuracy is lower than that of the mechanical LiDAR.
In fact, for cost considerations, most auto companies still adopt the “millimeter-wave radar + camera” fusion solution in the current passenger car market. To be precise, for the L2 aided driving, the millimeter-wave radar is used to collect distance and speed information, and the camera is to collect the image information of roads; then each transmits its information to the chip, and the chip makes the decision. This is the fusion scheme.
Compared with the “millimeter-wave radar + camera” fusion solution, LiDAR is superior in the ability to accurately model the surrounding environment in 3D. LiDAR can achieve high-precision recognition of targets, detection, and tracking of dynamic obstacles. It enables the self-driving car to have a clearer recognition and easier understanding of obstacles in front of them. And the high-precision perception of the surrounding environment will be improved by leaps and bounds, which means the improvement of autonomous driving capabilities is bound to be a leap forward.
At present, Xiaopeng, Volvo, Honda, Changan, Jihu, NIO, BMW, and other car automakers have clearly announced that they will apply LiDAR as autonomous vehicle sensors on future mass-produced car models. Let’s look forward to an automated world supported by LiDAR!