The new forces of car manufacturers have been developed in the past two years. Early new forces are starting from new energy electric vehicles because the autonomous driving threshold is higher. With the development of technology and customer trend, autonomous driving and lidar have to mention the schedule. Even though Tesla is disdainful to the lidar, there are most manufacturers use lidar as the main sensor. Tesla mainly focuses on Visual algorithms + radar sensors. This article interprets the parameters of the lidar.
The role of lidar in autonomous driving is mainly 3D / 4D ambient perception, detects road conditions and obstacles during vehicle driving, transmitting data and signals to autonomous driving brains, and then makes the corresponding driving action. Lidar can be said to be invisible eyes in autonomous driving.
Detailed introduction to lidar parameter
There are many lidar parameters: laser wavelength, detection distance, FOV (vertical + level), ranging accuracy, angular resolution, point number, harness, safety level, output parameter, IP protection, power, power supply voltage, laser emission mode (Mechanical/solid-state), service life, etc.
Six major parameters: detection distance, ranging accuracy, channel, FOV (vertical + level), angular resolution, and the number of points.
1 Lidar detection distance
The detection distance is well understood that the lidar can detect the range or radius. The ranging capacity of the lidar is related to the reflectance of the object to be tested. The reflectance is the ratio of the laser that is incident to the target to be reflected back.
The higher the target reflectance, the more effective echoes that the radar can detect, so the farther the distance can be measured.
Therefore, the detection distance is generally in general, and reflectance, such as 150 m @ 10%, means that the detection distance is 150 meters in the case where the target reflectivity is 10%.
2 Lidar detection accuracy
Detection accuracy refers to the accuracy of the detection distance, generally in centimeters. The higher the accuracy of the detection, and the deeper the 3D view is drawn.
3 Lidar channel
Lidar is divided into single-channel and multi-channel.
The single-channel lidar has only one laser emitter. As the radar rotates to form a horizontal scan line, it can only detect that there is no obstacle in front.
The multi-channel lidar has a plurality of laser emitters in the vertical direction. As the radar rotates to form a plurality of horizontal scan lines, one plane can be scanned.
As shown below:
For example, the single-channel is like we touch objects with a finger, and the multi-channel is like we use the whole palm or even touch objects.
Obviously, the more channel, the more detailed the portrait of the target, the more expensive of course the price.
4 Lidar FOV (horizontal vertical)
This is the detection of the field of view, including the horizontal and vertical direction, just like we open the flashlight as a wall, the light energy coverage range.
The mechanical lidar can be rotated 360 °, so the horizontal FOV is 360 °. The level of solid-state lidar will be small, such as 120 ° is a big perspective. The larger the horizontal FOV, the wider the range can be detected.
Vertical FOV is only useful for multi-channel lidar. It is an angle of the uppermost laser and the lowermost laser formation.
The vertical FOV field is generally biased underground, such as 0 °, then 15 °, down 25 °, so vertical FOV is 40 °, as shown below. This benefit is to allow the vehicle more to detect ground vehicles and pedestrians.
5 Lidar angular resolution
The angular resolution is also divided into horizontal and vertical directions.
Horizontal resolution refers to an angle of an angle formed by two scanned laser points. Since the laser turnover is rotated, the laser emitter is a pulse, so it is a point of the target.
The laser pulse is a fixed frequency, so the resolution in the horizontal direction is only related to the radar rotation speed, as long as the speed is slow enough, the resolution can be high, and it is also normal to do 0.01 °. However, the scanning speed will also affect the rate of information acquisition. Therefore, the horizontal resolution should be corresponding to the scan speed, the scan speed is generally represented by frequency, that is, how many times the 1-second scanning is scanned back and forth.
The vertical resolution refers to an angle of the laser point formed by the upper and lower channel. The channel is not evenly distributed in the vertical direction, but the intermediate intensive, sparse up and down, as shown below. This is also very well understood because the middle is more likely to detect pedestrians or obstacles.
The general manufacturer claims vertical resolution, apparently the resolution of the most intensive part. For example, the 64-channel product, vertical FOV 40 °, if the channel is evenly distributed, the vertical resolution is 0.625 °. It can actually be partially calculated as part of the intensive, claim 0.2 ° vertical resolution.
6 The number of points
Also called cycle acquisition points, an example, a 64-channel lidar, a horizontal FOV is 120 °, and the horizontal resolution is 0.2 ° at 10 Hz scanning frequency.
We can know that the laser has 64 points at a time, scanning a 120 ° can play 64X120 / 0.2 = 38400, 1 second to scan 10 times, a total of 384000 PTS / S.
Obviously, the more the number of points, the better the scan, which is the same as the machine gun.
Summarize
Master the six core parameters above, and master the essence of lidar.
Another example will explain the whole, take a student’s learning to compare:
The detection distance is the depth of learning, the deeper learning, the better the knowledge of the mastery;
The accuracy of the detection is the good and bad learning, the better, the better the test results in the future;
The channel is the ability to learn several disciplines at the same time, and the physics and chemical history of language mathematics is more harmful;
The FOV perspective is the breadth of learning, the better, the more knowledge is rich;
The angular resolution is the fineness of learning, the details of learning, the less the exam is missing;
The number of points is the result of learning, the more points, the better the test results.
