Lidar provokes a war of words, E'an, Xiaopeng, ideal each insists, who is the best solution?

Recently, Baidu's intelligent car company brand "Jidu" exposed its latest car robot concept car, which pioneered the layout scheme of the front cover to dual lidar, achieving good results. Unexpectedly, it attracted the industry's famous "big mouth" Li Xiang complained, he said that there is no difference between putting two lidar on the cover or bumper, and putting one on the roof, and even the roof single purpose program performance will be better.

He Xiaopeng soon joined the discussion, saying that two are definitely stronger than one, and the future city NGP can see the difference, Xiaopeng even considered the plan of arranging two on the roof, and finally chose the current plan.

Combined with the recent WM LiDAR model Road attempt to expose, Aeon and NIO lidar models went on the market. It seems that overnight, lidar has become a hot spot in the current new energy automobile industry. What is the difference between these technologies? Who can grasp the direction of the future direction? Let the kung fu car take everyone to see it together.

(1) Binocular vision and lidar, in fact, are not contradictory?

Before talking about lidar, let's talk about the development of intelligent driving cars. As we all know, the earlier private cars actually had no perceptual ability.

In the 90s, the "old three" Santana, Jetta, and Fukang actually belonged to the "three noes" cars without radar and cameras, and everyone was also happy to drive. People drive them, relying on the human eye, two "cameras" of less than 10 million pixels.

Then everyone began to "add weight" to the vehicle, such as with a reversing radar, you can perceive the distance from the obstacles behind you when parking, plus a 360 ° or even 540 ° image, you can see the surrounding area and even the bottom of the car, greatly reducing the scratching. The radar around the car can sense the vehicles near the lane, and when changing lanes, it is more spectral, and the vehicles in crisis situations can also take the initiative to brake and stop to avoid accidents.

Recently, Tesla and Xiaopeng have been able to realize automatic car following in the city, and they can directly let go of their hands and feet on the high speed, and many times everyone feels that the vehicle is better than their own driving.

High-precision positioning units, high-definition cameras, surround view cameras, ultrasonic sensors, high-precision luxury wave radar, these intelligent driving configurations are gradually "getting on the car" during this period. It is only in recent years, when major car companies have begun to introduce lidar, that the route has begun to diverge.

The first to stand up against Tesla, Musk believes that blindly stacking hardware will not solve the problem. Since humans can complete the perception of the surrounding environment with their binocular eyes, in theory, cameras with higher accuracy and farther vision can also complete this action, provided that the vehicle's AI processor is powerful enough to think like a human and be more accurate than human judgment. To do it, Tesla even shielded the vehicle's own radar on some of the new measurement models, using only the pure visual scheme of camera perception. But judging from the results, this road seems to be more difficult to walk, and the difficulty coefficient is even more geometrically increased. Most of the other opposing car companies believe that lidar is too expensive and are not optimistic about its popularity prospects.

But in any case, after Tesla fell to the pure vision scheme, it was frequently frustrated in the evaluation of major intelligent driving tests, and the blessing of lidar for intelligent driving was still obvious. In the view of Kung Fu Automobile, the binocular vision scheme and the lidar scheme are not contradictory, and the former is not mature enough, and the latter is indeed a better solution at this stage.

(2) The roof monocular is a pretext, and the "three eyes" of the Erlang god is YYDS?

Compared with ordinary millimeter-wave radar, lidar uses a laser beam, the working frequency is higher, the anti-interference ability is stronger, and the angle, distance and speed resolution have been greatly improved, which can indeed better "assist" the intelligent driving system.

At present, the main disagreement of car companies is actually focused on how to use this "artifact".

At present, the set degree is arranged on both sides of the front cover, xiaopeng is arranged on both sides of the bumper under the headlight group, and the ideal and Weima are arranged in the middle of the roof, and only one is used.

In fact, the real set of the big one is GAC Aeon, its previous flagship model AION LX Plus, which has a range of more than 1000km, is equipped with 3 second-generation lidar. It uses a layout of 1 overhead + 2 on both sides of the front fender, which is a bit like the three-eyed position of the "Erlang God". The official propaganda is more "honest", claiming that it can cross-cover the 330° field of view around the body, which should be the limit of what can be done at present, but there is still a 30° "blind spot" that requires other sensors to complete.

From the perspective of actual performance, AION LX Plus can not only achieve high-speed NDA intelligent piloting, but also realize the unique urban NDA pilot assistance of the same level, automatic turning around, dodging "ghost probe" These functions can be completed autonomously, so that users can rest assured that the "driving right" will be handed over to the intelligent system, and it can also realize the summoning parking beyond the visual distance, which is also relatively leading at present.

To put it bluntly, the current intelligent driving, than how to combine these high-performance hardware with high computing power intelligent driving platform, the range that "three eyes" can measure is indeed wider than monocular or binocular, which cannot be escaped no matter how to play word games.

(3) The development of intelligent driving technology has stagnated, and it is imperative to cancel the classification?

In the past few years, almost all car companies have claimed to have achieved L3-level intelligent driving. But a few years have passed, only Mercedes-Benz's L3-level "conditional automatic driving" system has passed the approval of Germany's local technical regulations, and to put it mildly, this certification is not recognized in the two major markets in China and the United States, where new energy vehicles are the most popular.

Why is L3 so hard? In fact, it is still strongly related to the essence of intelligent driving.

Let's first look at the classification of automatic driving, starting from L3 level, it is recognized as highly automated driving, and the obligation to observe the environment becomes the "responsibility" of the system, and the driver only needs to intervene in the fierce driving. In fact, at present, Xiaopeng and Aeon are close to this level, but they dare not "promise" to achieve L3 level intelligent driving. Because once such a "commitment" occurs, it means that the problem is the responsibility of the car, after all, the manufacturer needs to ensure the absolute safety in this case.

At present, several mainstream intelligent driving implementation paths have inevitably encountered bottlenecks. First of all, the pure visual program mentioned above, it needs to have a highly intelligent AI, can be like a person "thinking", the difficulty coefficient MAX; the coverage of the high-precision map scheme is limited, there is no map manufacturer that has the strength to achieve high-definition map measurement nationwide; V2X vehicle networking program is also very far away, after all, let tens of thousands of dollars of cars or cars sold more than ten years ago are networked, the current difficulty is also relatively large, since most of the cars on the road can not be networked, The guiding significance of V2X for intelligent driving is naturally greatly weakened.

Comparatively speaking, lidar is indeed a better solution at present, its implementation is quite good, and the cost is constantly decreasing. Previously, Huawei promised to reduce the price of car-mounted lidar to $100, which is acceptable for many car companies.

But it has to be said that with the current technical level, it is still difficult to achieve a complete L3, after all, car companies still cannot guarantee 100% safety. Even if there is an accident in 1 million kilometers, multiplied by a staggering number of users, it is impossible for car companies to bear. Therefore, a more reasonable solution is actually to cancel the restrictions of the grading system, which can achieve most of the L3 functions, and can also realize some L4 or even L5 functions, so that driving becomes more relaxed and comfortable, but it is not classified as highly automated driving.

(4) Kung Fu shooting

Judging from the performance of the Aeons, the blessing of lidar for intelligent driving is still very large. At least for now, it has more advantages than pure visual solutions, and the price is still decreasing, and it has basically the possibility of promotion. Judging from the current degree of completion, the more "lower" Aion LX Plus "Erlang Shen" three-eye solution should be the best done.

But the real automatic driving is still far from us. During the morning rush hour, users who often see Tesla or Xiaopeng "catching up" are actually testing car companies with their own safety to some extent, which is not compliant. Technological progress often doesn't happen overnight, so let's wait and see.