After half a year, I don't know if anyone still remembers a slogan on the poster of the Great Wall Salon car, "Less than 4 please don't talk." Unfortunately, lidar is not a CPU, and the number of heaps (cores) can be stronger. Algorithms, algorithms, what matters is algorithms!
Of course, today is not to talk about salons, but to talk about the topic of "lidar". As THE NIO ET7 was gradually delivered to its first customers, its volume reached its second wave since its release, and even the "JDM Brain Residue Powder Concentration Camp" - Subaru's new BRZ Dading Group was discussing the car.
Unexpectedly, the most discussed thing is not the workmanship, performance, and endurance that ET7 is proud of, but the three different sizes of "acne" on the roof, that is, lidar.
Because they all feel that "roof acne" is too incongruous and too unsightly.
But what they don't know is that not only the NIO ET7, but also the NIO ET5, the Ideal L9 and the Zhiji L7 all use the overhead lidar scheme. I suddenly remembered the iPhone on the table, how did consumers change from "swearing" to "getting used to it" since the birth of that "bang"?
Good things should be placed in a prominent position?
"I don't have a bag on my head, and no one else knows that my car has lidar." Aside from the ridicule, similar expressions don't really give us an idea of why manufacturers are placing lidar on the roof. What are the benefits of the "overhead bag" that the entire design team can make concessions.
To make this clear, we must first figure out the basic working principle of lidar.
In simple terms, autonomous driving consists of three parts: perception, decision-making, and execution, and lidar plays an important role in the perception link. If you haven't returned junior high school physics to your teacher, you should know that radar works by emitting ultrasonic waves from bionic bats, and the same is true for lidar, which reaches the obstacle in front by emitting a laser beam, then receives the laser beam that turns back from the obstacle, and measures the time difference and phase difference of the laser signal to determine the distance of the obstacle. If you played the game of using echo to determine distance when you were a kid, it should be well understood.
Therefore, if you want lidar to complete the "perception" task better, it must be "standing high and measuring far", and it can send back the maximum and minimum signals at the same time, which is the most ideal measurement result, and the roof is undoubtedly a more suitable position.
In addition, if the lidar is placed too low, it is more susceptible to cover materials such as mosquito carcasses, flying stones, and sludge. If you install a "radar cleaning" device on the lidar like a headlight cleaning, it always feels redundant and the cost is not necessary.
In addition to this, there are several reasons for the cost of repairs as well as pedestrian regulations. In my previous video of the static experience of xiaopeng P5 (review please poke "six advantages and four disadvantages to take you to quickly understand the smart home car Xiaopeng P5"), I mentioned that one of the disadvantages of lower-mounted lidar is that it is too easy to be the first "killed" in a collision. Although the fixed lidar HAP used by the Xiaopeng P5 is an upgrade of Livox based on the Horiz series, and Livox is a subsidiary of DJI, the unit price of HAP has been reduced to several thousand yuan. However, the slightest friction accident will easily incur tens of thousands of yuan in repair (replacement) costs.
Pedestrian protection is better understood, before The Weibo exposure degree SIMUCar 2.0 dual lidar version on the car, from the red box position to see the lidar placed on the cover, Li Xiang also commented that the program does not meet the latest pedestrian collision regulations. Otherwise, the classic flip lamp will not disappear into history.
The above basically explains the main reasons why lidar is placed on the roof of the car, and it is summed up to meet the performance requirements of lidar, the requirements of pedestrian collision regulations, and the requirements of low damage rates.
Smart you must be thinking, aren't the A-pillars on both sides and above the windshield also suitable positions? Aren't Subaru's EyeSight and Tesla's polyocular cameras installed in the rearview mirror position of the car?
This brings us to the final key factor in the placement of lidar: the conditions of installation.
First of all, let's talk about the A pillar, even if we put aside the objective reason that the space inside the A column is too small to install lidar, if the lidar is forcibly embedded in the A pillar, then the car may fail in the crash test. Riders who know a little about body-in-white know that the rigidity of the A-pillar plays a vital role in passive safety, and most automakers also use ultra-high-strength steel in this area. Usually, this area does not have additional designs that affect the safety of its structure, and it is really an inversion to want to dig a hole in the middle to install lidar.
The only problem with the windshield is that the light transmittance is not enough, and the current national standard requires that the windshield transmittance of the new car factory is not less than 70%. However, the light transmittance that meets the requirements for the use of lidar is more than 90%, and the front windshield of mass production and commercial use on the market has not yet been born.
Of course, there are also many suppliers who have thought about other locations, such as XenomatiX, which provides solid-state lidar solutions, and Marelli, which provides LED matrix headlights, which are collaborating on the development of integrating lidar into the headlights. If the vertical angle of the lidar is sufficient, the headlights are indeed a relatively suitable position, and the headlight cleaning function can also be shared. It's just that the price of this headlight may exceed the imagination of many people, and the solution is still in the laboratory.
As things stand, the roof is indeed a relatively ideal placement for lidar. Because of this, I wonder if Musk's reason for saying that "only a fool would use lidar" is because he can't tolerate such an ugly design.
Lidar is not the "ESP" of the smart car era
After talking about the overhead layout of lidar, let's look at the important role of lidar for intelligent driving. Nowadays, when the newly launched smart electric vehicles are generally equipped with lidar, and the individual cost of lidar is gradually decreasing, and it has entered the 200,000-level electric vehicle, is it possible that lidar will become the "ESP" of the future intelligent electric vehicle and become a non-mandatory but default standard component?
The key to this is to clarify whether lidar is a just-needed component of the L2-liter L3 for smart cars, or the L3-liter L4.
The answer is: NO.
Looking back at the "Automotive Driving Automation Classification" issued by the Ministry of Industry and Information Technology, the important transformation from L2 to L3 is that "detection and response of targets and events" is completed by the combination of "driver and system" of L2, and becomes the "pure system" of L3.
With L2 level autopilot assistance, when the assist is turned on, the driver can briefly leave the steering wheel and accelerator pedal, but the driver remains in an observational state, the system is only responsible for driving, and the driver's intervention is required when other situations occur. After entering L3, the system will control the vehicle according to the road conditions, such as changing lanes to overtake, identifying other traffic participants including pedestrians, and steering according to the high-precision map, but the driver is required to intervene in case of emergency.
Compared with lidar, lane-level HD maps are more like the transition from L2 to L3.
The biggest change from L3 to L4 is that the driver does not need to dynamically take over the vehicle, the system is fully responsible for unexpected situations on the road, so as to make a judgment, the only restriction of L4 is that the automatic driving system can only be turned on in areas where the design operating conditions are available.
Whether it is L2 to L3, or L3 to L4, the most important change is whether the system can gradually replace the work that originally required the driver, and does not clearly stipulate the radar detection distance, breadth and other data for turning on automatic driving assistance. Even if the calculation of the braking distance is mentioned in the ISO 22179 standard (vehicle specification level), it means that the vehicle must sense at least about 200m ahead of the target to have enough warning time, which is also a target requirement, rather than specifying lidar to complete this requirement.
Because lidar is not the only component that can accomplish this task. In terms of detection distance and detection range, the camera is YYDS. However, the image quality of the latter is indeed inferior to that of lidar, and it relies heavily on algorithms and computing power. This is also why Tesla decided to abandon the Xavier autopilot chip developed by Nvidia based on GPU in 2020 and turn to the FSD chip with higher computing power developed independently.
However, there are also solutions to integrate vision and lidar, such as Xiaopeng G9, which is mainly visual, and lidar fills the blindness. Therefore, including the Xiaopeng P5, the performance specifications of the lidar they are equipped with are not very high. Both NIO and Ideal will be dominated by lidar in the near future, such as ET7, which uses expensive, ultra-long-range detection capabilities (WEILai said that there are 500m) Tutton Falcon lidar, while using multiple 8 million cameras for assistance, ideal L9 also uses a single Hesai AT128 lidar as the "main vision".
All roads lead to Rome, and the future realization of L4 and L5 comes in many forms. Whether enterprises choose to use lidar only depends on whether lidar is in line with the technical route at that time and cost control.
Master observation
In general, lidar will not become an iconic part of the car like "ESP", it is just one of the solutions to the road to high-end autonomous driving. However, because China's traffic situation is much more complicated than that of foreign countries, and the imaging performance of lidar is indeed much better than that of cameras, it is a relatively efficient solution that does not rely too much on visual algorithms (requiring ultra-high computing chip and long-term AI learning). In the medium to long term, lidar should become a fairly common component in the field of smart electric vehicles.