Recently, in the capital market, the "three fools" of new energy have made a fuss on Wall Street, and the Internet is also very lively. Musk he Xiaopeng and the two founders quarreled on the social platform.

XPeng Inc. unveiled a new P7 at the Guangzhou Auto Show on Nov. 20, and his CEO, he Xiaopeng, revealed that the world's first mass-produced smart car equipped with lidar will be launched next year. That night, Musk mocked XPeng Inc. for imitating Tesla, Inc. on Twitter, and then he Xiaopeng responded harshly to Weibo Corp, claiming that he would beat Tesla, Inc. so hard that he couldn't find the east.

Last year, Musk also complained that "lidar is like a human body with a bunch of appendices, and the existence of the appendix itself is basically meaningless."any company that relies on lidar may die without a disease." In theory, the lidar is no longer in Musk's field of vision, why do you personally attack XPeng Inc. because of the lidar at this time?

Interestingly, Huawei, which is outside the auto circle, has also begun to release news about mass-production specification-level lidars. Wang Jun, president of Huawei's smart car solution BU, told the media during this year's Guangzhou auto show that Huawei's lidar has become an explosive product snapped up by auto companies, and is considering making an overall release of lidar products in the near future.

Why is Musk engaged in a war of words over lidar? why are the giants staring at lidar? It all starts with the intelligent driving of the car.

01 what does smart driving do?

In Fast and Furious 8, the villain Charlize Theron hacked into the car intelligent driving system with high technology in order to seize the nuclear weapon launcher, making thousands of unmanned cars form a huge army of "zombie cars". The car waves surged forward in the street, and the picture was shocked.

So how on earth can smart driving be realized?

To put it simply, it is to add an intelligent driving module to a conventional car and connect it to the communication network. This module has three core functions:Environmental perception-computational decision-making-control execution corresponds to human "eye-brain-nerve" respectively.

Their functional executive logic is also very simple, very human-like: when walking on the road, the eyes first notice a person coming from the opposite side, then the brain decides how to avoid the person, and finally the nerves drive the limbs to bypass the person.

The eye of intelligent car is mainly composed of camera, lidar, high-precision map, GPS positioning and so on. It is mainly responsible for collecting the information around the car body and determining the location of the vehicle.

After receiving the data information from the eyes, the brain of the intelligent car identifies the traffic condition through the calculation chip + perception algorithm, and then analyzes and calculates the optimal route.

Based on the decision-making results given by the brain, the nerve of the intelligent car gives instructions to the braking system, steering system, motor system and so on, controls the running state of the vehicle and executes the driving route.

The argument between Musk and he Xiaopeng is: what should be mainly used as eyes, lidar or cameras?

Who will be the eye of the smart car?

The visual scheme imitates the human eye, mainly by "seeing", which is equivalent to a two-dimensional camera. The radar scheme mimics bats and is a three-dimensional "scan", which is equivalent to a three-dimensional camera.

The advantage of visual solutions is that they are relatively mature. Camera-led, combined with millimeter wave radar and other low-cost components, constitute pure visual computing, can meet the current L2 self-driving needs, the typical representative is Tesla, Inc..

However, after all, the two-dimensional image captured by the camera is more difficult to mine than three-dimensional information, which requires more powerful algorithms, training of a large amount of data and longer-term research and development investment, and there will be defects in accuracy, stability and field of vision, which can not meet the requirements of self-driving above L3 level for the time being.

The radar scheme can create clear 3D images, usually see farther, see more clearly, and have stronger anti-jamming ability, which can meet the requirements of autopilot above L3, but at a higher cost. Usually dominated by lidar, coupled with camera, millimeter wave radar and other components, the typical representatives are Alphabet Inc-CL C Waymo and domestic Baidu, Inc. Apollo.

According to the structure, lidar can be divided into three types: mechanical, solid and mixed solid, which can be said to be traditional version, high-configuration version and market version.

At present, the technology of mechanical lidar is the most mature, which collects information by mechanical rotation 360 °scanning, so the faster the rotation, the more information is collected. However, because there are too many mechanical parts, the debugging and assembly is complex, the production cycle is long, the cost is difficult to reduce, and the mechanical parts are easy to damage in the real road conditions, so it is difficult to meet the requirements of vehicle regulations.

Solid-state lidar belongs to the "high-rich and handsome" version, which is divided into OPA and Flash, which do not need mechanical parts, but the current technology is not mature. The technical principle of the former is similar to the phased array radar of fighter planes, and the cost of mass production is very high. The principle of the latter is flash, which can complete the 3D drawing of the whole scene at once. Both are the stars of tomorrow, and cars are still out of reach.

At present, hybrid solid-state lidar is the most promising solution to land quickly. It integrates all the mechanical parts into a single chip and controls rotation through circuits, so the radar itself does not need to rotate greatly, and it is not easy to be damaged under real road conditions. mass production capacity has been greatly improved, and with the help of Moore's Law of the chip, the cost of mass production is expected to be reduced to hundreds of dollars in the future.

With the further popularization of autopilot technology in the future, the market scale of lidar will be further expanded. Yole, a consultancy, estimates that about 4.7 million lidars will be shipped worldwide in 2025, with sales of about $6.19 billion.

Lidar looks so beautiful, why are car companies not necessarily so enthusiastic about it?

03 Tesla, Inc. didn't like it, XPeng Inc. came out to do it.

To put it bluntly, now that the car is equipped with lidar, the account is still a little difficult to calculate.

Musk once said, "he doesn't need lidar for free." he wants to solve the problem of visual information processing algorithmically.

His reason is also simple. Musk believes that cars and roads are designed for humans, and that the mode in which people's vision enlarges the brain to process information can be driven safely, so can self-driving. Radar is a sense that people do not have, why install, this is not normal people on crutches, superfluous?

But what is more likely to dislike Musk is that lidar is now more than ten times more expensive than visual solutions. Velodyne's lidar, for example, costs about $4, 000 for a 16-wire lidar and $80, 000 for a 64-wire harness, compared with a few hundred dollars for camera autopilot hardware.

A lidar costs more than two Model3 cars. Musk is naturally not interested in this kind of expensive stuff.

XPeng Inc. is considering playing the safety card to compete with Tesla, Inc.. XPeng Inc. claimed that he would use the fusion mode of camera + millimeter wave radar + ultrasonic radar + lidar to do autopilot. To put it simply,Fill up the equipment and run faster, more stable and safer.。

But even if the lidar can meet the autopilot requirements above the L3 level and the high cost, how should XPeng Inc. deal with it?

XPeng Inc. 's own answer is that he has been training his own suppliers, by increasing the number of orders can greatly reduce manufacturing costs, but also enhance their bargaining power. But apart from that answer,Obviously, Huawei's entry is an external variable that can not be ignored.。

At a time when XPeng Inc. and Tesla, Inc. are at odds with each other, Huawei also revealed that it is studying lidar with the goal of developing 100-line products in the short term and reducing the cost to $200 or even $100 in the future. Xu Zhijun, Huawei's rotating chairman, also claimed that "we can do whatever Tesla, Inc. can do now" and is applying Huawei's corps operations to lidar.

To put it simply, it may be a bit dangerous for XPeng Inc. to challenge low-cost lidar directly, but Huawei's entry has changed the debate.

In addition, in August this year, DJI announced that it could mass-produce 1000-yuan-level automotive lidar, which is suitable for L3 and L4 self-driving. The prices of Horizon and Tele-15 are only 6499 yuan and 9000 yuan respectively, which is very cheap compared with tens of thousands and hundreds of thousands of yuan of Velodyne, which may also be the source of XPeng Inc. 's strength.

If the price of lidar falls sharply in the next year or two, the "free and no use" Tesla, Inc. still insists on his vision plan, which may be enough for Musk to drink a pot.

Of course, lidar is not perfect, and there are some deficiencies in its application.For example, heavy fog and heavy rain will reduce the intensity of the laser signal, and the current lidar point cloud is based on geometry, although it can distinguish the shape, but can not distinguish the color and texture, etc., which means that the lidar can not tell the difference between paper bags and stones.

The overall scheme of lidar with other sensors may be the direction of application.

04 investment outlook: intelligent driving industry chain

In the three levels of intelligent driving industry chain, environmental perception mainly corresponds to camera and radar (including ultrasonic, millimeter wave, lidar); calculation and decision-making mainly corresponds to computing chip, perceptual algorithm, high-precision map, communication technology and so on; while control and execution is brake system, steering system, motor system and so on.

Overseas manufacturers have a first-mover advantage, and domestic forces are also gradually rising. Velodyne in the United States has taken the lead and has been successfully listed this year, and the domestic market is also in full bloom, competing to prepare IPO.

Smart car is a combination of hardware and software, and the number one player in the world is China and the United States.The competition between Tesla, Inc. and China's new forces is also the competition of hardware and software scientific and technological personnel between China and the United States, as well as the competition between the opening up and cooperation of the United States and the battle of the Chinese legion.

It is fun to watch a direct CEO war of words, but the dispute over the technological route behind it will bring more profound changes to our lives.