Electric living
The "arms race" of lidar has quietly begun. The main hardware of this automatic driving advanced has been applied by more and more car companies, and its role is becoming more and more important, accompanied by a large amount of capital injection, so that lidar has become the fuse of the "war" of automatic driving.
This article will be the classification of vehicle-mounted lidar, the working principle for everyone to make a detailed description, while introducing some of the current lidar models, if you are interested in popular technology, or want to buy high-end automatic driving models, you can make a reference (full text dry goods recommended collection), you can also pay attention to the electric life public number to learn more.
· I ·
What is In-Vehicle LiDAR?
Definition: Vehicle-mounted lidar, also known as vehicle-mounted 3D laser scanner, is a mobile 3D laser scanning system and one of the most effective tools for urban modeling. Mainly composed of three parts: transmitting system, receiving system and information processing, the basic principle is to install the three-dimensional laser scanner + POS system on the car, which can build a 3D model in a longer and farther range.
· II ·
The use of lidar is controversial
While the mass production of lidar has triggered competition, there is still controversy in the industry about the technical route of autonomous driving. Some views believe that cameras are insufficient in data form and accuracy, and that lidar needs to be supplemented at the perceptual level of autonomous driving; others believe that lidar is unnecessary and too expensive. Tesla CEO Musk has said that "lidar will not be used for free" and wants to solve the problem of visual information processing from the algorithm.
I combed through for you, first of all, Tesla's autopilot is to achieve commercialization at low cost, and the visual solution can achieve his goal, so tesla's current publicity caliber is no problem, and Tesla is also secretly testing lidar.
The original intention of the new domestic car-making forces is to drive higher levels of automatic driving, so they directly use the most high-end sensor - lidar. Although the current scheme is divergent, but the target detection of lidar, positioning and mapping are visually irreplaceable, to L3, L4 level of automatic driving, lidar is inevitable to carry, so the future of the same destination is inevitable.
· III ·
LiDAR: The cultivation of self-driving incense
2021 is known as the first year of lidar, and now after two years of development, more and more brands have been realized, such as Weilai, Xiaopeng, Weima, Zhiji, Jihu and so on. Lidar can be said to be a good way to achieve higher levels of autonomous driving (L3 level and above), as well as higher safety, compared to millimeter-wave radar, lidar has higher resolution, better stability, and more reliable three-dimensional data.
Most of the current L2 level autonomous driving hardware consists of camera + millimeter wave radar + ultrasonic radar. The advantage of the camera is that the recognition is clear, but the human eye encounters strong light, darkness, blindness and other problems, the camera also has; ultrasonic radar is what we often say reversing radar, the cost is very low, but the measurement distance is short, easy to be affected by weather interference, only in the slow speed of the car has an advantage; millimeter wave radar is a radio wave radar, strong ability to penetrate smoke, can make up for the lack of camera, more for blind spot monitoring, lane change assistance, not afraid of strong light, adapt to bad weather, resolution up to 5cm, However, the accuracy of the judgment is relatively poor, and the effect is greatly reduced when driving (Tesla insists on using this combination of radar + camera).
Lidar (Lidar) beam range is very narrow, so more longitudinal beams are needed to cover a large area, so the wiring beam determines the size of the picture, scans and then measures the distance through the return time, and accurately and quickly builds a model, which is much stronger than other radars at present, so it is more suitable for automatic driving systems, but it is also equally susceptible to weather and higher cost.
· IV ·
LiDAR: Why Autonomous Driving Solutions?
Lidar can provide accurate 3D measurement data and do a better job in relatively harsh weather and lighting conditions. Lidar can be combined with sensor data such as millimeter-wave radar, cameras, etc., to provide reliable identification data for static and dynamic objects in the driving environment, helping to detect obstacles, avoid collisions and safe navigation, and lidar is a highly available, mass-produced solution.
Specific advantages:
(1) Separation rate and accuracy: in the automatic driving of the vehicle, a large number of reliable measurement data is generated, accurate to the centimeter level, and the object is clearly identified;
(2) Working environment: less disturbed by changes in the external environment, able to adapt to bad weather and normal work under light, to ensure the safety of vehicles during automatic driving;
(3) Reaction speed: The emitted light pulse calculates the distance through the round-trip time, and the surrounding pedestrians and objects can be detected, and the surrounding environment can be perceived in real time to make timely judgments.
The traditional mechanical rotation lidar technology is the earliest and most mature, due to the large size, so the equipment is difficult, in the automotive application but almost unable to achieve vehicle specification level verification, and now reduced to the test vehicle dedicated; pure solid-state lidar is the future development direction, is not yet mature, mass production is more difficult; semi-solid-state lidar can achieve scanning at a lower cost and higher accuracy, is relatively mature, is a technical solution that quickly passes the vehicle specification certification, semi-solid-state radar is currently very suitable for loading needs.
· Five ·
Semi-solid-state lidar is mainstream: classification/how it works
Between vehicle-mounted lidar, according to the difference in structure, it can be roughly divided into: mechanical rotation lidar, hybrid semi-solid-state lidar and all-solid-state lidar (Flash flash and OPA phased array, collectively referred to as non-scanning):
1. Mechanical lidar is large in size, high in cost and difficult to assemble. It achieves horizontal 360-degree coverage through rotation, vertical angle coverage through internal lenses, and has more durable and stable characteristics year-on-year, so most of the autonomous driving road tests we see use this type, and the radar is constantly rotating on the roof to complete the lateral scanning, relying on the increase of the laser beam to achieve longitudinal and wide scanning.
Mechanical lidar
2, semi-solid-state lidar is currently the mainstream product of the car. According to the scanning method, there are the following classifications: silicon-based MEMS, rotor, galvanometer + rotor, rotary transmission prism.
Semi-solid-state lidar classification, representative manufacturers, models, prices
(1) Rotating mirror: The rotating mirror is divided into one-dimensional rotating mirror and two-dimensional rotating mirror. The one-dimensional rotating mirror reflects the laser to different directions through the rotating polyhedral mirror; the two-dimensional rotating mirror, as the name suggests, integrates two rotating mirrors, one polygonal prism is responsible for lateral rotation, and the other is responsible for longitudinal flipping, realizing a laser beam covering horizontal and vertical two-way scanning. Rotor lidar is small in size and low in cost, consistent with the effect of mechanical lidar, but the mechanical frequency is also very high, and the life expectancy is not ideal.
Two-dimensional rotor lidar
(2) MEMS: Adjust the laser reflection angle through "vibration" to achieve scanning, and the laser transmitter is fixed, but it is a test of the receiver's ability, and the life is also a major challenge in the industry. The cantilever beam supporting the galvanometer has a limited angle and a small coverage, so it requires multiple radars to work together to achieve large viewing angle coverage, which will cause uneven distortion and overlap at the edge of each lidar scan, which is not conducive to algorithm processing. In addition, the cantilever beam is very thin, and the mechanical life needs to be further improved.
MEMS galvanometer
(3) Galvanometer + rotor: on the basis of the galvanometer, the galvanometer is responsible for the transverse, the galvanometer is responsible for the longitudinal, to meet the wider strafing angle, the frequency is higher, the price is more expensive than the first two, but it also faces the problem of life.
Galvanometer + rotating mirror type lidar, we can understand that the two-dimensional rotor is responsible for longitudinal scanning of the rotor, replaced by a galvanometer
(4) Prism: Prism lidar, also known as double wedge prism lidar, consists of two wedge-shaped prisms inside, and the laser is deflected once after passing through the first wedge prism, and deflected again after passing through the second wedge prism. Controlling the relative speed of the two prisms controls the scanning pattern of the laser beam.
The cumulative scanning pattern of prism lidar is shaped like a petal, the center point is scanned intensively, the edge of the circle is relatively sparse, and the scanning time is long to enrich the image, so it is necessary to add multiple lidar work to achieve higher results. Prisms can achieve high precision and long-distance detection by increasing laser wiring beams and power, but the structure is complex, the volume is more difficult to control, and the risk of bearing and bushing wear is greater.
Prism lidar illumination
3, all-solid-state lidar is ideal, it is difficult to mass-produce at present. As the name suggests, this lidar does not have any mechanical oscillating structure, and naturally does not rotate. Mechanized lidar chipping, smaller, better performance, more reliable life, but can not escape the orbit of Moore's Law, there are currently two ways.
How OPA solid-state lidar works
(1) OPA solid-state lidar does not swing firmware at all, using multiple light sources to form an array to synthesize beams in a specific direction to achieve scanning in different directions. It has the advantages of fast scanning speed, high precision, good controllability, small size (Quanergy lidar is only 90x60x60mm), the disadvantage is that it is easy to form side lobes, affecting the beam range and angular resolution, and the production difficulty is high.
How Flash solid-state lidar works
(2) Flash solid-state lidar, can also be said to be non-scanning, it can directly emit a large area of laser covering the detection area in a short period of time, using a light array to build an image, just like a camera, quickly record the entire scene, reduce the lack of rotation and lens wear, relatively more stable, but the defects are also obvious, such as the detection distance is closer, the processor requirements are higher, and the corresponding cost is also high.
· VI ·
Is lidar a shortcut to L3 autonomous driving?
Automatic driving is the future trend, the current mass production on the market is still L2 level automatic driving, there is no L3 level of automatic driving vehicle on the road, L3 level does not need the driver to monitor the road conditions in real time, only need to take over the vehicle when the system reminds, L3 is the real hands-free technology level.
Lidar has many advantages such as higher resolution, small size, light weight, relatively strong anti-jamming ability, and good low-altitude detection performance, which can bring qualitative improvement to automatic driving. Compared to visual routes and other radar sensors, lidar can be said to be a shortcut to higher level of autonomous driving.
· VII ·
LiDAR market outlook
Although the foreign lidar market has developed earlier, the current domestic lidar market has begun to show a situation of a hundred flowers, and many start-ups are more advanced in mechanical radar and develop to solid-state radar.
Under current technical conditions, lidar is a shortcut to L3 autonomous driving technology. The L3 intelligent driving car will be released in batches in 2021, which is expected to drive lidar to achieve large-scale mass production in 2022. In the future, with the reduction of the cost of mass-produced lidar, lidar is expected to become the standard for mid-to-high-end vehicles for safety reasons, and the scale of the on-board lidar market is expected to further expand.
According to relevant consulting companies, the global lidar market size will be 13.54 billion US dollars by 2025 (breaking 10 billion), which will achieve an average annual compound growth rate of 64.5% compared with 2019. Among them, the size of China's lidar market will reach 4.31 billion US dollars, achieving an average annual compound growth rate of 63.1% over 2019.
· VIII ·
Deficiencies and improvements in lidar
Lidar also has shortcomings in applications. For example, heavy fog and heavy rain will weaken the intensity of the laser signal, the accuracy will decrease, and the point cloud of the current lidar is based on geometry, although it can distinguish the shape, but can not distinguish the color and texture, which means that the lidar can not distinguish between paper bags and stones.
In order to improve the defects of lidar, the vehicle also added cameras to the perception level of autonomous driving. Some analysts believe that the high resolution of the camera and the accurate ranging of the lidar just complement each other. Compared with the pure visual perception that relies on the camera, the integration of lidar and the camera brings about the improvement of information acquisition ability, the algorithm requirements can be reduced, and the driving safety will be more guaranteed.
Autonomous driving must be the mainstream of future travel, and lidar is a pair of "eyes" for self-driving cars, which will become the standard of every smart car in the near future. If lidar wants to become a "vehicle specification" grade product, it must meet the three important conditions of "low cost", "reliability" and "mass production".
· Nine ·
A summary of models equipped with lidar
Xiaopeng P5, NIO ET7, Alpha S Huawei HI Edition, Zhiji L7, BMW iX, AITO Q&I, WM M7
1
● Xiaopeng P5
【LiDAR brand】DJI Livox
【Quantity】2
【Longitudinal wiring harness】Equivalent to 144 wires
2
● NIO ET7
【LiDAR brand】Innovusion (Tudtong)
【Quantity】1
【Longitudinal wiring harness】Equivalent to 300 lines
3
● Polar Fox Alpha S Huawei HI
【LiDAR brand】Huawei
【Quantity】3
【Longitudinal wiring harness】96 lines
4
● Zhiji L7
【Longitudinal wiring harness】Expected equivalent to 300 lines
5
● BMW iX
6
● AITO asks the boundary
【Longitudinal wiring harness】Expected equivalent to 96 lines
7
● WM M7
【LiDAR brand】Innoviz
【Longitudinal wiring harness】Equivalent to 504 wire
More models equipped with lidar:
···
Great Wall WEY Mocha, SAIC Feifan R7, Nezha S, NIO ET5, Aian LX Plus, Salon Mech Dragon, Audi A8, Mercedes-Benz EQS, Changan Boat Architecture, Honda Legend, Volvo XC90, Lucid Air, etc.
· Written at the end ·
Edit the summary
In any case, the current lidar is the best choice for automatic driving advancement, semi-solid-state lidar is the present, pure solid-state lidar is the future, do not rule out that with the development of technology there are better alternative products, but at present this is the margin we can see. As for which of the current semi-solid-state lidar is the main force, it depends on who can better meet the three important conditions of "low cost", "reliability" and "mass production".