A few days ago, Tesla CEO Musk attended the TED2022 conference held in Vancouver, Canada, and completed a super-long interview of more than 50 minutes. Talking about the tens of billions of dollars to buy Twitter, talking about the space program, Musk also rarely talked about the progress of Tesla's self-driving project FSD.
He said more than 100,000 drivers, including himself, have participated in the FSD testing program and claimed that L4 level autonomous driving of FSD could be achieved this year.
At about the same time, BMW released its new electric flagship i7, but still can only support L2, and said that it may support L3 in the future.
Talking to a friend about this, he said that he had an illusion that the traditional car companies with a century-old car manufacturing history seemed to be inferior to the new forces in terms of automatic driving.
This is actually not entirely an illusion, according to the current situation, the new forces have indeed done a better job in the technical climbing of autonomous driving. What is the truth of the matter? Let the kung fu car take everyone to see it together.
(1) Traditional car companies are more rigorous? Not necessarily
It is often seen on the Internet that the reason why the autonomous driving technology of traditional car companies is slow to develop is mainly because the verification is too strict. The new forces do not have this trouble, Tesla, Weilai launched are semi-finished products, so it looks advanced. Traditional car companies are not unable to do it, they are disdainful of doing so, and this statement is quite marketable.
In fact, traditional car companies are actually quite aggressive in the promotion of new technologies. At that time, the turbocharged engine was just popular, and the problems were also very many, and then the double-clutch transmission behind it was also a trick for many users. After the ascension of the sixth country, a large number of trucks have recently had the problem of particle trap blockage, in large part because the domestic market uses temporary solutions that are not fully verified, and in the end, most of them are paid by users.
In the field of automatic driving, traditional car companies have never been "polite". The earliest promotion of L2 level automatic driving technology, in fact, most of them belong to traditional car companies, which can realize adaptive cruise control functions and lane keeping auxiliary functions, and can actively brake in emergency situations.
Although the functions are not much, the early traditional car companies in the publicity, especially like to promote their AEB emergency braking technology, but the actual effect is really very bad. Either there is a problem and there is no braking at all, or the overreaction causes the driving experience to decline. This problem has not been solved very well until now, but it was vigorously promoted by car companies as a selling point as early as ten years ago. So it is not realistic to say that any of their technologies will be fully validated before being promoted.
What's more, today's L2 level of automatic driving has been defined as combined driving assistance, and it is called conditional automatic driving until L3, and L4 is highly automated driving. In the case of the delay in the establishment of L3 automatic driving regulations, the car company itself does not have any responsibility, in fact, it is a question of providing functions and allowing users to choose whether to use or not to use.
So, traditional car companies are indeed losing this race. This situation is positively related to their car-making system.
(2) Is the dispute between Mobileye and NVIDIA a microcosm of the route of autonomous driving technology?
A while ago, Kung Fu Automobile talked about the problem of domestic car companies using NVIDIA's automatic driving chips, not only byd bydir such a predator into the game, but also by new forces such as Ideal and Weilai, who also chose to switch to NVIDIA. The Israeli company Mobileye, which has been studying in the field of autonomous driving for nearly 20 years and once occupied 70% of the market share, has become the biggest loser, and the market is constantly being eroded.
Nvidia, as a latecomer, can shoot the front wave on the beach, in addition to computing power and justice, in large part, it relies on a more flexible market strategy.
Traditional manufacturers such as Mobileye and Bosch prefer to make a simple solution that can adapt to more models. For example, Bosch's AEB program, in fact, is composed of front-mounted millimeter wave radar and control system provided by Bosch, the advantage is that the cost is low enough, and it can indeed play a certain role, but the disadvantage is also very obvious, that is, the lack of personalized adaptation, the supply of A home is this set, the supply of B home is also this set.
The ideal car, on the other hand, uses the millimeter-wave radar plus visual observation scheme, and after a series of verifications, its entire AEB completion degree is much higher, and even took the top of the AEB test.
For example, all traditional car companies are adhering to the L2, and it is also the "black box" directly given by Mobileye. Manufacturers don't have to care what is inside, anyway, no matter what kind of car, no matter what the driver's operating habits are, the functions it provides are the same. So what did Tesla do? Its FSD system does not belong to L2 or L3 or L4, but has all L2 functions, most L3 functions, and a very small part of L4 functions. Its logic is also very simple, dilute the meaning of this level, so that users can use it as well as possible.
The role of NVIDIA is that of a core chip provider to ensure the computing power of the entire system. Let the major car companies calibrate and test themselves, so the ideal is very different from Xiaopeng's plan, and Weilai is another routine, but the whole drive is indeed better than that of traditional car companies. This is a test drive for everyone, and naturally it is more clear at a glance.
To put it bluntly, the new forces are really developing, calibrating, and thinking about what kind of solutions can achieve better results. The so-called many years of experience in car manufacturing by traditional car companies is actually integrating the resources of suppliers, A's sensor plus B's chip plus C's ECU, or buy a complete set directly from Mobileye. And these suppliers, can not do perfectly, if they can really get it, not already built their own car.
Just like the previous joint venture brands mocking their own brands, "overtaking in curves" is not only difficult, but also unsafe. Nowadays, some new forces are actually in front of traditional car companies, because they are really developing and testing, and traditional car companies are still touching Bosch to cross the river.
(3) The cost has skyrocketed, and the pure visual solution will become the final mainstream?
At present, the mainstream autonomous driving solutions in the world are mainly divided into four types, pure vision solutions, high-perception solutions, high-precision map solutions and V2X wireless communication technology solutions, which actually represent four different genres.
The logic of the pure visual scheme is simple, humans have no special perceptual ability, and all the information when driving comes from visual perception. Therefore, in theory, the vehicle only needs a camera, it has even driven farther and clearer than ordinary humans, and in theory, through the camera and a set of neural networks that simulate human vision, it can achieve automatic driving. However, theory is theory, and the current difficulty of realization is still high.
The high-perception scheme is through the camera to see and high-performance radar perception, so as to avoid the surrounding obstacles, theoretically the sensor's ability is infinitely powerful, this scheme can also achieve automatic driving; high-precision map is to refine all the road information, the current coverage is still relatively narrow, V2X emphasizes that all vehicles are in the network, then everyone knows where each other is, naturally there will be no problems.
Originally, these routes were all played separately, especially when the funds in the market were extremely abundant before, and they complemented each other perfectly. However, with the economic downturn, the financing difficulties of new forces and the soaring prices of raw materials, those car companies that used to play lidar have said that they will not play, and the high-precision maps of major map manufacturers have also made little progress, and V2X has slowly not heard of it. Pure visual solutions with the lowest cost and highest ceiling seem to have become mainstream. The technological accumulation of traditional car companies in this regard seems to have once again fallen behind the new forces.
(4) Kung Fu shooting
Nowadays, the automatic driving performance of traditional car companies is indeed inferior to that of new car companies. The reason is that the technology of traditional car companies relies on "integration" to provide standardized solutions through suppliers at all levels. In order to adapt to more customers, suppliers will inevitably adopt a relatively conservative approach, and the effect is indeed not worth mentioning. Just like Qualcomm's chips are professional, they will never catch up with Apple, because one is personalized customization, and the other is a general solution.
The current level of autonomous driving suppliers is still very far from Qualcomm. If you are satisfied with the "low insurance" solution provided by Bosch, the technological progress of traditional car companies may always be so slow.