Definition: The history of lidar
Before the discovery of lidar, our definition of radar was (Radio Detection and Ranging) is a radar with millimeter wave, microwave, and ultrasonic waves as carriers.
When with the development of laser technology, we talk about the combination of laser technology and radar technology, and then invent the corresponding lidar, the earliest application in the aerospace field, and then, applied to surveying and mapping, security, transportation and so on.
At the same time, many types of lidar have emerged, such as limeographic radar, liber radar, micropulse lidar, laser synthetic aperture radar, and lidar. At present, it is the most widely used in the military and is also the driving force for the development of its laser technology.
The application of its lidar in autonomous driving is relatively late. The first to propose this plan was in 2004, when the U.S. Defense Advanced Technical Research Agency (DARPA) conducted a self-driving car challenge to see that the vehicle could travel more than 200 kilometers without control.
Due to the harsh environment of the competition, the prize money has become higher and higher over time, but there is still no team that can handle it. By the third time, the founder of Velodyne had built a 64-line radar with a base that could rotate, greatly improving the perception and reconstruction of the surrounding environment, and achieving better results in the subsequent unmanned car challenge
After this, major automobile manufacturers, suppliers began to increase the development and application of lidar, until today, the research enterprises of lidar, that is quite hot, the development of engineers which dick, salary can be comparable to BBA.
Definition: What is lidar?
Definition: Lidar is a radar that operates in the optical band, with a laser as a carrier, an optical device as an antenna, and an optical detector as a receiver, and an environmental detection device, which can be defined as LIDAR (Light Detention and Ranging) or LADAR (Laser Detection and Ranging).
If you know more about millimeter wave radar, you can follow suit, the wavelength of lidar is about 1550 nm, you can find that the laser emitted by lidar is within the infrared range, the wavelength range is located in the infrared (0.76-1000 μm, 760 nm-1 mm), and the millimeter wave wavelength is at the millimeter level (1mm-10 mm), its wavelength is far greater than the laser of the lidar, so in essence, it is a wave, the perception of the surrounding environment, nothing more than the difference in wavelength, All are through the propagation time, determine the distance, the time difference between the two times, and find the speed, as shown in the following figure:
Overview 1: Lidar pros and cons
First of all, from the transmitter end of the analysis, is the wavelength is different, as can be seen from the above figure, it and the biggest difference between millimeter wave is that the wavelength is relatively short, so there is a good resolution relative to the millimeter wave, at the same time, because the wavelength is short, it is easy to receive the influence of external weather and working environment, such as sandstorms, smog, lidar reflection point clouds, it is very easy to be interfered with, so the relative accuracy is affected.
The second is the receiving end, that is, the baud properties of the laser reflection of the color are different, so that the corresponding road signs, lane lines, etc. can be identified! But if it is a millimeter-wave radar, the corresponding results cannot be obtained. This is also where the NB of lidar comes in.
Overview 2: How LiDAR Works
Let's talk about the principle of nagging lidar, and when it comes to the principle, let's see what this thing is made of. Transmitter + sensor + optical components + scanning = lidar (mechanical), about other forms of radar, we will talk about later, the specific form, as shown in the following figure:
What does the launcher do? Emitting lasers, 600nm-1000 nm of laser rays, optical components and scanning parts are receiving reflected light. Sensors are used to analyze reflected light (e.g. information about light intensity) to estimate the type of object.
Through these introductions, we can roughly understand what is inside the lidar? Let's talk about how it works.
Overview 3: Measurement principles of lidar
Ranging principle: According to its pulsed light after encountering obstacles, the turnback time, calculate the relative distance between its target and itself, and other radars are not essentially different, as shown in the following figure:
At the same time, many friends must have heard that laser phase difference ranging, the principle of this ranging, and millimeter-wave radar is somewhat similar, are two high-frequency signals, into an intermediate frequency signal, and then calculate their phase difference, so as to get distance information.
What are the advantages and disadvantages of these two ranging methods? Pulsed light, fast speed, low resolution, phase ranging, slow measurement speed, high resolution, but for automatic driving, the requirement is fast, the resolution is cm level, and most of them use pulsed light for ranging.
Speed measurement and angle measurement principle: This is relatively simple, speed measurement, that is, the multi-frame point cloud information can be differential processing, measuring angle, because it is a mechanical radar, the angle of rotation can be estimated through the base motor rotation angle and multi-frame point cloud.
Height measurement principle: For lidar, there is another dimension that can be measured, that is, the height of the object, this height is actually the line of the lidar to measure, single-line radar, only can detect the radar at a certain height, multi-line radar, it is equivalent to cutting watermelon, the object is divided into multiple parts, so as to get its height. At present, the most NB is the 64-line and 128-line radar.
Principle of object characteristic measurement: Because different objects have different spectral characteristics of laser reflection, the contour, color, heat and other relevant information of the object can be seen through the reflection spectrum.
Therefore, it can sometimes be said that the information measured by lidar is the reason for the 4D information.
Overview 4: Description of the measurement of lidar
In what form is the measurement principle described above expressed? Millimeter-wave radar is represented by a bird's eye view, lidar is represented by point clouds, what is a point cloud?
A point cloud is a massive collection of points that express the spatial distribution of a target and the spectral properties of a target in the same world coordinate system. So how is this point cloud formed?
Since the lidar is not stationary, in the process of driving, the lidar rotates at a constant speed at a certain angle at the same time, in this process, the laser is constantly emitted and the reflection point information is collected in order to get all the environmental information, the single-threaded laser can get the distance of the reflected point, the time and horizontal angle, because each laser emitter has a fixed number, angle, and the lidar will rotate a week to collect all the reflected point coordinates, this set forms what we call a two-dimensional point cloud, as shown in the following figure:
At the same time, if the lidar is multi-threaded, it can form a three-dimensional point cloud, and at the same time contain the corresponding height, resulting in its three-dimensional structure. As shown in the following figure:
According to the above introduction, a general understanding of its working principle, at the same time, lidar and millimeter wave radar, have the corresponding parameter information, I find a map from Baidu, you can refer to the following, take a look. Specifically, there is not much to nag.
Overview 5: Classification of lidar
At present, there are about two types of automatic driving, one is mechanical lidar, one is solid-state lidar, mechanical lidar as described above, the following important introduction, the capital market is more popular solid-state lidar, all lidar classification as shown in the following figure:
Solid-state lidar, which we can understand as lidar without rotating parts, can be divided into three types according to its principle, FLASH, MEMS, based on phased arrays.
The so-called solid-state lidar, the general understanding is that it is a solid-state lidar that does not rotate. It is usually divided into three types, which are implemented based on phased array, Flash, and MEMS.
Phased array radar As shown in the figure below, it can actually be understood as an enhanced version of millimeter-wave radar, with a launch array that scans the surrounding environment according to the laser phase difference of each emission point.
Lidar in the form of Flash, which belongs to the non-scanning program, mainly emits area array light, with two-dimensional or three-dimensional images as the output content of the lidar, as shown in the following figure, one is scanning, similar to millimeter wave radar, and the other is direct scanning radar.
MEMS is scanned by changing the angle of a single emitter by means of a micro-vibrational environment, thus forming a surface array light effect.
These three forms of lidar are relatively small in size, but there are not many mass production at present, most of them are Demo stages, which can be understood as the stars of tomorrow.
SLAM application of lidar
What is SLAM? SLAM is Simplaneous Localization and Mapping, that is, the construction of synchronous positioning and maps, which can be understood as the vehicle moves from an unknown unknown in an unfamiliar environment, and in the process of moving, it locates itself according to unknown estimates and maps, and builds incremental maps on the basis of its own positioning, so as to realize the autonomous navigation and positioning of vehicles, which can also be understood as automatic driving above L3+
As shown in the image below, it is a map built by SLAM.
How is the SLAM algorithm implemented?
Step 1: Data filtering, culling problematic data, the common algorithm is Kalman filter or Bayesian algorithm. Step 2: Match the filtered data with the high-precision map, and then update the corresponding location. Step3: Map fusion, this round of information from the lidar, spliced into the original map, to complete the map update.
What is the role of SLAM in autonomous driving?
The biggest advantage of SLAM is that it can generate high-precision maps, high-precision maps can reach the centimeter level, and at the same time can be on the shoulder of the road, the slope of the road, the distance between the car and the road, which makes the autonomous vehicle not only able to easily control its own position, but also make its own path planning, with more information input, so that the vehicle's automatic driving level reaches L3++ or above.
Lidar-related companies
Lidar enterprises, under the tide of capital, like mushrooms, as many as cattle feathers, but there are a few that really make products, let us count them one by one.
Tall enterprises:
Local Enterprises:
These enterprises have developed well, but let the current form, it is estimated that the following enterprise development will be more NB, currently unknown, because the current L3, L4 technical solutions, there is no final conclusion, such as the Tesla that hangs the sky, he does not use lidar, purely with cameras and millimeter-wave radar algorithm iteration.
Summary
At present, lidar is still on the edge of the blue ocean, although it has obvious advantages, but its shortcomings are still more obvious, so that the public accepts that there is still a period of time, such as its high price, relatively low security (no encryption code, while easy to receive interference from adjacent radar), harsh environmental adaptability (especially the adaptability of smog weather), large amount of computation (the corresponding algorithm and hardware are more complex) and so on.
This also inhibits the difficulty of its practical application, but there will be challenges when there are difficulties, there will be opportunities for challenges, and only in this way will there be a chance to get rich, or it is estimated that it will not be the darling of capital.