Take you understand Lidar in three minutes

Classification：Technical Support

Release Date：2023.02.22

With GM's focus gradually shifting to Cruise Automation, Waymo's self-driving began to be commercialized and BAT's in-depth layout in the field of intelligent driving, it is not difficult to see that self-driving technology has become a hot land for big manufacturers to step into, and autonomous driving is inseparable from the three core logic of perception, decision and control.

These three core logic are also easy to understand. If the car is regarded as a person, the perception is just like the human eyes, which collects and feels the surrounding situation and sends it to the brain. After receiving the information perceived by the eyes, the brain makes planning decisions, and then controls the hands and feet to execute the brain's instructions. These three rings are interlinked, one cannot be absent.

Let's start today with the first link, perception. There are four kinds of sensing components of self-driving cars: camera, ultrasonic radar, millimeter wave radar and Lidar. This time, we will focus on the lidar with the best comprehensive performance.

According to the mechanical structure of lidar can be divided into two types, one is mechanical lidar, the other is solid state lidar.

The biggest feature of mechanical lidar is that it has a mechanical rotating mechanism, which means it can turn itself, so it is larger.

Solid state lidar can actually be subdivided into OPA, MEMS, Flash three lines, the principle is relatively abstract will not expand, let's take a general look, solid state lidar on the structure of the biggest characteristic is that there is no rotating parts, relatively small head.

Several important parameters of lidar are measuring distance, angular field of view, measuring accuracy and measuring speed.

It's easy to understand that the maximum measured distance determines how far the "eye" can see. This parameter is especially important when we are driving at high speeds. How far we can see determines that we can drive too fast and need to leave enough reaction time.

The angular field of view generally has a vertical right Angle field of view and a horizontal angular field of view, and the angular field of view determines how large the "eye" can see.

Measurement accuracy means how clearly the "eye" can see. This accuracy is enough, just like when we drive a car, we only need to see the objects around us that will affect us, and we do not need to see the sand and stones on the ground.

A key value here is angular resolution. The smaller the angular resolution, the higher the relative accuracy. For example, a 64-line lidar with an angular resolution of 0.4 degrees can scan an object 1.7 meters away 50 meters and produce 5 lines of data.

A 40-line lidar with an angular resolution of 0.33 degrees can get six lines of data from the same object.

The measurement speed determines the refresh speed of the information transmitted by the "eyes" to the "brain". The lidar senses that there is a car, and the comparison of the two data before and after can tell whether the car is moving, the direction of the movement, and the speed of the movement, which affects the brain's judgment all the time.

The mechanical lidar arranges the laser beam vertically to form a plane, which is rotated by a mechanical rotating component, and scans the surrounding environment to present a three-dimensional pattern. We often say that 16 lines, 32 lines and 64 lines are the number of vertical wiring harnesses. The more the number, the higher the density and the higher the accuracy, but the amount of information needed to be processed by the computer also increases.

Because mechanical lidar is rotating, it has a horizontal view of 360 degrees and can see all the way around, and the speed of rotation also affects the scanning frequency.

Mechanical Lidar is large and mechanically rotating, so it doesn't fit perfectly into the shape of a car. It just sits on top of the car and doesn't look futuristic.

Solid-state lidar has a very limited horizontal perspective because it has no rotating mechanism, requiring multiple solid-state lidar locations in different directions. The advantages of solid-state lidar are fast response, high accuracy, and relatively small size, which makes it easy to hide in the car body.

So where can I see it?

Lidar manufacturers at home and abroad are more famous, Velodyne, Quanergy, LeddarTech, Datang Juchuang, Beike day picture, He said technology and so on.

Among them, Velodyne is the first and largest, focusing on mechanical lidar. Its 64, 32 and 16 lines of lidar are priced at $80,000, $40,000 and $8,000, respectively. This is not even tax and freight added. I'll leave the price of the new 128 line to your imagination. Another problem with Velodyne is production capacity. With lidar manufacturing demanding precision, the delivery cycle for 64 lines of Velodyne is close to four months, making it hard to buy one.

Baidu's Apollo uses Velodyne's lidar.

The driverless passenger car cooperated with FAW was announced at Baidu World Conference in November this year, and it was announced to achieve small batch production in 2019. Although it has been integrated with the car body in appearance, there is still some distance to be recognized by consumers.

Quanergy specializes in solid-state lidar, releasing what it claims is the world's first solid-state lidar in 2016 with a cost of $200 per sensor.

China's Sagitar is a latecomer, but it has a strong cost control ability, and also provides a complete lidar sensing program.

Compared with millimeter-wave radar and camera, Lidar has obvious advantages in detection accuracy, detection range, stability and environmental adaptability, but the price is very touching. As for whether unmanned driving will take Lidar route or gradually take computer vision + ultrasonic radar route, I think more depends on the price of Lidar in the future.