In the past few years, the autonomous driving industry has been shouting and killing two products, one is high-precision maps and the other is lidar. It is indeed imperative to get rid of the dependence on high-definition maps, but there is no need to shout and kill for lidar.
01
Cost reduction and efficiency increase is the main theme of contemporary society, in the price war wave after wave, volume price is more value of the car circle, cost reduction has become a more urgent need than efficiency increase.
There are not too many tricks to reduce costs, but more of the boring of picking a little here and there. Gather less into more, gather sand into a tower, and the cost will drop little by little.
There are two cost control freaks in the electric vehicle industry: the eldest brother BYD and the second brother Tesla.
According to Musk's tricks, the first effective means is to eliminate unnecessary parts, and the second effective means is self-development. This idea is most thoroughly embodied in Tesla's FSD self-driving system.
Computing chips and sensors are two major cost items for autonomous driving systems:
In terms of computing chips, Tesla came up with its self-developed first HW chip as early as 2019.
In terms of the hardware composition of the sensor system, Tesla did not use lidar at the beginning, and later, Tesla, which was addicted, successively canceled millimeter-wave radar and ultrasonic radar.
According to Cao Xudong, the founder of Momenta, the hardware cost of Tesla's FSD is currently about 4,000 yuan, which is roughly half of the cost of the high-end intelligent driving system with the same performance level of domestic competitors.
In sensor systems, lidar is the largest cost item, and there is no one.
In 2019, the year Musk issued the famous lidar appendix theory, the price of vehicle-mounted high-beam lidar was about $10,000.
At that point in time, Musk's remarks were extremely correct.
However, there is no need for us to beg for a sword, because, in the past few years, the price of lidar has been cut in half many times, directly to the surface of the foot.
At present, the price of high-line lidar has dropped to $500 per unit, and this price can be reduced by more than half, Suteng Juchuang recently released medium and long-range lidar MX in April, and the mass production price is less than $200, and after the product is increased, Suteng Juchuang will continue to reduce the price of MX to 1000 yuan.
Such a low price can basically put an end to the debate over whether to go to lidar, right?
02
The great writer Leo Tolstoy, who was forced to run away from home by his wife, once said something heartbreaking: Happy families are all similar, and unhappy families have their own misfortunes.
The misfortune of the family mainly comes from the quarrel between the husband and wife. In the past few years, the two camps have been fighting over whether to take the pure visual route or choose a multi-sensor fusion scheme with lidar.
If you really want to talk about first principles, human drivers do rely on human eyes and human brains to drive, except for Wuzhu in Qingyu Nian, no one's eyes can emit lasers.
However, the first principle grocery store is not opened by Musk, and the other first principle is that the biggest difference between humans and animals is that humans are good at using tools and using tools to improve their productivity. What does the old man Xunzi say, the gentleman is good at faking things!
In fact, there is no need to discuss the respective characteristics and related physical characteristics of lidar, camera, and millimeter-wave radar, and more importantly for consumers, under the full publicity of several car companies, such as NIO, Xpeng, Li and Huawei, LiDAR has become a must-have option for active safety. This perception has been formed and deeply rooted in the hearts of the people.
Whether it is the vigorous debate between Huawei and Xpeng Motors in November last year, the addition of high-end AEB functions to the major update of NIO's Banyan system in January this year, or the elaboration of the role of LiDAR when Li Auto launched Mega in March this year, consumers have been instilled again and again with the view that there is no high-end active safety without LiDAR.
Of course, in addition to the hardware of lidar, a lot of work needs to be done at the level of software algorithms to ensure that the AEB system can play the expected active safety role.
However, without hardware, it is equivalent to the difficulty of a clever woman to cook without rice, and without lidar, it directly eliminates the possibility of a higher level of active safety (based on braking speed as the standard).
Let's think about it, in the AEB horizontal evaluation test carried out by various media, is Tesla, which is purely visual, basically at the bottom every time?
03
During this time, various companies are laying off employees, and employees who are in the limelight take the initiative to work overtime, do the work of two people, and take one person's money, in order to avoid being on the boss's layoff list.
In order to continue to stay on the car, LiDAR also tried its best to roll itself into a mess.
In order to prove its value to car companies, LiDAR is constantly looking for new uses, and a penny is broken in half, are you embarrassed to take it down?
In addition to high-end intelligent driving and active safety, lidar can also be used for road surface perception of intelligent chassis and precise shielding of high beams.
The concept of intelligent chassis is not very clearly defined, but it is roughly understood that the information of the road surface in front is collected through the perception sensors originally used for the high-end intelligent driving system, and the damping of the CDC and the height of the air suspension are adjusted in advance according to the potholes, bulges, uphill, downhill, and ups and downs of the road surface, so as to improve the smoothness and comfort during driving.
There are three key words here, one is perception sensors, which in this case refers to cameras and lidar.
The second is the road surface ahead, the road surface in front of the vehicle can only be collected by the perception sensor, and the road surface under the wheels can be collected by the vehicle sensor.
The third is CDC and air suspension, which is the executive component of the intelligent chassis.
However, just as there are two technical routes of pure vision and multi-sensor fusion in high-end intelligent driving systems, some car companies only use cameras to scan the bumps and bumps of the road surface, which cannot meet the requirements of all scenes and all-weather, and is easily affected by weather and environment.
Lidar, which is more robust to the weather, can directly provide 3D information, which is best used to determine the degree of potholes in the road surface when combined with the texture information of the camera.
It is worth mentioning that some models advertised as magic carpets only use vehicle sensors (without using perception sensors such as cameras and lidars) to collect information about the road surface that the vehicle is currently driving, which does not conform to the concept of an intelligent chassis that can predict the road conditions ahead.
In addition to scanning the road ahead in the intelligent chassis, when turning on the high beams in low-light conditions, the lidar can help track the oncoming car and achieve accurate shielding of the high beams.
When it comes to lidar, some people will definitely mention Tesla, and I have to admit that Tesla's FSD's autonomous driving experience is indeed very amazing, but it is precisely because of the lack of lidar that it has been beaten by Huawei on the active safety and intelligent chassis!