In the field of intelligent driving, will it be graded this year?
The above question is one of the things I thought about before this auto show. Recently, in the field of lidar and chips, their technological innovation and iteration are also accelerating under the premise of the rapid development of intelligent driving. The functions of the software of intelligent driving are becoming more and more advanced and mature, and after the product standards at the tip of the pyramid begin to be in the field of intelligent driving, there will be a level of division of intelligent driving configurations and functions one after another.
At this auto show, we focused on domestic brands, and of course there will be some joint ventures that are trying to catch up at the level of intelligent driving, such as Japanese Toyota. However, more information about intelligent driving still depends on domestic brands, such as Huawei, Xiaopeng, NIO and Xiaomi. As a result, we have seen X brands and began to have lightweight solutions for intelligent driving, or with such an idea, to put it bluntly, it is to reduce the number and level of perception equipment on the car.
So, from this year, will there be a clearer classification in the field of intelligent driving?
Huawei takes the lead, and the grading has begun
The Beijing Auto Show caught up with the intensive release period of Huawei's products. In addition to the facelifted model of the M5, the facelifted model of the M9 and the Xiangjie S9, the Zhijie S7 (priced at 249,800 yuan) was also unveiled again. And our focus is on the intelligent driving solution on the Zhijie S7, which is the basic version of the ADS system, without lidar, it is a pure vision solution, and the lidar version, it can also support the NCA function of high-speed and urban expressways.
As a result, Huawei now has at least three driving solutions that can be used.
A set of ADS basic version, pure visual driving can realize the NCA function of high-speed road section + urban expressway, a set of 126-line LiDAR ADS 2.0 intelligent driving system, and a set of 192-line LiDAR ADS 2.0 intelligent driving system. In terms of breakdown, these are three intelligent driving solutions corresponding to different price points. The underlying logic of this development model is to develop a set of intelligent driving, and then reuse it, and start to reduce or replace the configuration on multiple versions to meet the needs of various price points. Especially under the premise that the intelligent driving controller is self-developed, the cost and operation difficulty of this reuse are lower.
Then the ideal L6 is indeed a loss, and the loss is that there is no self-development of lidar, so that the facelifted Wenjie M5 rolled out a set of lidar + high-end intelligent assisted driving at the same price.
In fact, originally this auto show, I also wanted to pay attention to the information of a wave of NIO's sub-brands, but Ledao Automobile did not come to the scene. And the intelligent driving scheme of this car, according to the exposed information, is also a pure visual NOA for high-speed and urban areas, but it will not be discussed in detail if it is not exhibited, and the underlying logic is still algorithm reuse.
Who, still has this ability?
In fact, Xiaomi SU7 used this scheme at the beginning, ready-made, cut the chip computing power, cut the lidar and changed to pure vision + low computing power chip. In the self-developed intelligent driving echelon, Xiaopeng can also be ranked in the head camp. The MONA brand brought at the auto show (no actual car has been released yet) is also a pure visual driving solution without lidar, and the underlying logic is still to reuse the existing lidar driving solution.
Similarly, iCAR V23 is also a product that takes the route of pure visual solutions, focusing on cost performance, no lidar configuration, and the functions to be realized are high-speed NOA and urban NOA;
Of course, intelligent driving has begun to be graded not only by electric vehicles, but also by fuel vehicles.
The Volkswagen Tiguan LPro and the all-new Volkswagen Magotan at the auto show are two fuel vehicle products that use DJI's intelligent driving solutions. What they are currently using is a relatively low-cost intelligent driving hardware solution (about 7000 yuan), which can achieve functions including high-speed NOA + urban NOA.
From a technical point of view, it is very easy to implement, plus a set of intelligent driving controller and perception equipment, as long as the circuit architecture of the whole vehicle allows, you can get on the car. It is predicted that after this year, more and more fuel vehicles will begin to use intelligent driving functions, and now there are also Tier 1 companies that have begun to prepare low-cost chips to meet the needs of fuel vehicles.
After starting to segment, does it work well?
Rather than cutting the performance of the intelligent driving perception scheme, it is better to say that this is a kind of decentralization of intelligent driving configuration and functions. The intelligent driving industry has been developing rapidly for at least 5 years, including the development of chips, lidar and the whole set of driving solutions, which have reached a good level of maturity.
This year, it can be judged that there is already a trend towards subdivision development in this field.
And today's rhythm is to subdivide the intelligent driving market, so that there are suitable intelligent driving products to choose at all price points. A few years ago, the 128-line and 126-line lidars were flagship products, and now there are also products that use a single one to make low-cost lidar driving solutions, and there are quite a few. Then, the current flagship LiDAR products already have 150-line and 192-line LiDAR products.
Answering what consumers are concerned about, can the effect be easy to use?
First of all, we must make a misunderstanding clear, intelligent driving does not reverse, shrink in the existing configuration and function, but split in hardware and function. For example, at this auto show, there are multiple versions of intelligent driving in supply, including the 10+TOPS version that responds to the most basic high-speed NOA, the 80TOPS version that responds to high-speed NOA, and the 128TOPS version of low-cost urban NOA, corresponding to the cooperation of 9 domestic car brands, and the problem of interior car installation should not be big this year.
Let's talk about the pros and cons of this split.
The benefits are obvious, so that more low-priced models have the minimum intelligent driving function, that is, the function of high-speed NOA; the realization of this set of functions, chip + visual perception + map can be done, the difficulty is no longer the high computing power chip, but the cost of the high computing power chip.
From the perspective of technical logic, what is really affected is actually the batch of products with urban NOA functions. These products require high computing power and lidar, and if these two hardware configurations are subsequently refined and graded, then the corresponding experience will be downgraded.
Roughly downgrade,It will be the special accuracy of intelligent perception hardware and price matching,10-200,000、200,000、200,000 and more than 300,000 yuan,Will be equipped with different computing chips to form an intelligent driving solution;Roughly the difference in use,It will be reflected in the sensitivity of the instruction operation、The distance with the car、The number of takeovers,The completion of complex instructions may not be as smooth as the high computing power chip,And the accuracy of recognition may also be affected due to the limitations of the perception device。
Let's talk about the last question, is there a difference between pure visual urban NOA and lidar urban NOA?
At this stage, in the field of intelligent driving technology in the domestic market, there are lidars that are still equated with advanced and easy to use, while the urban NOA made by pure vision solutions is equal to cost-effective and usable (a few like Jiyue, which can also be used to achieve easy use with pure vision). The underlying logic finds the reason, pure vision can indeed save the cost of lidar, but if the cost is not invested in the training of algorithms, neural networks, and large models, it is difficult for the pure vision solution to reach the "easy-to-use" level of equal to lidar.
This year, the classification of domestic intelligent driving solutions will be composed of pure vision solutions< lidar + medium computing power chips< high-performance lidar + high computing power chips.