Scholarship, under the Chinese robe

Without a new concept, if you do n’t touch the porcelain or pull it, you do n’t call the press conference.

Recently, the domestic autonomous driving company, on the AI ​​DAY organized by its own home, not only shouted the slogan of China's mass production autonomous driving, but CEO Gu Weiyi also disagaed again "Backpot Man" Tesla - The NOH (Urban Navigation Assistant Driving) function at the end will drive the former at the speed of landing.

At the time node, Tesla's AI Day will be held on September 30th in North America time. At the end of the nose, Zhixing rushed to be held before, with great gunpowder.

However, at the press conference also affirmed Tesla's achievements. Gu Weizheng proposed that the development of autonomous driving for nearly ten years can be divided into three stages: hardware driver, software -driven, and data driver. From a global perspective, there are only Tesla families that really enter 3.0, and the end is where it is located at the end. On the way to sprint.

In fact, the meaning of holding Gauthram's words is that other car companies must stand side by side.

It is also a unicorn in the field of autonomous driving. Wu Gansha, CEO of Yushuang Technology, believes that "autonomous driving has not yet entered the second half, strictly speaking, it is just the end of the start. Measurement. "" "

From the perspective of Zhixing at the end, as a startup that has just been established for more than 1,000 days, it has weaved a Chinese robe for himself.

Who "enter the city" first?

Enterprises can't wait to appear new products for marketing. And urban navigation auxiliary driving is the fragrance of players in the field of autonomous driving.

On the AI ​​Day, Zhixing said that the model "Mocha DHT-PHEV Lidar Edition", which is equipped with a large-scale mass-produced city NOH, plans to mass production in September and will be released within the year. This can be described as the limelight of a group of car companies.

Prior to this, "Wei Xiaoli", Huawei, Zhi Ho, etc. were planning to pilgrimage their own urban leading driving functions. Among them, the most aggrieved Xiaopeng was Xiaopeng. As early as November last year, Xiaopeng Automobile demonstrated the test version of the self -developed NGP (smart navigation auxiliary driver), and plans to push in many cities such as Guangzhou and Shenzhen in the first half of this year.

Official news shows that Xiaopeng G9 will be officially launched on September 21. One of its important selling points is that it will realize the intelligent assistance driving function of the city's full scene. There is no doubt that it is also grabbed by Mocha, which is equipped with NOH.

Why is the city market scene of car companies such as "Wei Xiaoli" difficult to land? An important factor is a high -precision map. At present, as one of the core sensing techniques of autonomous driving, especially in the city market scene, high -quality maps have become an indispensable part.

Because of this, a lot of car companies have been "tripped":

First of all, the data of the high -precision map is very heavy, but the urban roads across the country are not a small workload;

Second, it is necessary to solve the problem of freshness. At present, the vast majority of autonomous driving systems have high requirements for the freshness and quality of high -precision maps, but the map manufacturers fail to satisfy car companies in their freshness;

Finally, the qualifications of the high -quality map industry are tightening. Recently, the Ministry of Natural Resources issued the "Notice on Promoting the Development and Maintenance of Geographical Information of Surveying and Mapping", further reiterating the relevant provisions and requirements of existing laws and regulations and policies on surveying and mapping activities. Car companies.

In order to avoid the above problems, Zhixing chose the line of "emphasis on, light maps".

At this stage, the technical routes of intelligent driving are divided into three genres: pure vision; heavy maps, light perception; emphasis on perception, light map.

For example, Wei Xiaoli, etc., relies on high -precision maps, follows the so -called so -called heavy map light sensitivity line; at the end, Zhixing uses the line of "heavy perception, light map", that is, to get rid of the restrictions of high -precision maps, rely on laser radar+millimeter Paper+camera's heavy perception ability.

For the city of NOH, the city of Mocha, public information shows that the hardware it is equipped includes: 2 125 -line laser radar, 12 cameras, and 5 millimeter wave radar.

In the official context, in order to get rid of the dependence on the high -precision map, Zhixing chose the same neural network model based on Tesla -based neural network model to make the three dimensions of space, time, and sensor. Accuracy.

The category of technology applications requires markets to verify. In any case, due to the lack of high -precision maps, the security and reliability of the NOH system cannot avoid being questioned by the market.

Perhaps it is to leave a back road for yourself. Previously, Zhixing said that it would not rule out that the data covered in the future would also be used. On September 5th, Zhixing signed a strategic cooperation agreement with Siweitu. Is this cooperation in lack of confidence in existing roads, or change the path?

In order to recapture the volume, Xiaopeng Automobile's official Weibo said on September 17: The first domestic marketing market in China has a high -level intelligent assistant driving driving. The city NGP has begun to pilot it in Guangzhou and push it to some Xiaopeng P5 users. Even if Xiaopeng only opened a city in Guangzhou, it was the first to mass production. I don't know how to think at the end.

In addition, the most praised comments under the Weibo are "Can you recognize all static objects." At present, when car companies pay attention to whether its auxiliary driving technology can become a marketing gimmick and whether it will lag behind other brands, I hope not to forget the standard of safety.

The victory of "progressive"? In the speech on AI Day, Zhang Kai, chairman of Zhixing, said that the progressive route will win the end in the intelligent driving competition. Auxiliary driving is the only way for autonomous driving. "

Zhang Kai's words were almost sentenced to death for leap forward routes.

For a long time, around the evolution of autonomous driving, there have been two major "genres" in the industry: Among them, Waymo and Pony Zhixing are representatives of the "leap forward". It is the representative of the "progressive" route. Its strategy is to first equip the L2/L3 -level assisted driving on the mass production vehicle, and then enter the L4/L5 after collecting the data.

The two parties have not been convinced by both parties before the camp. In the past two years, the situation has changed. Many L4 companies have begun to "reduce dimension" into the L2 field. For example, Qingzhou Zhihang, the Robobus field, launched a high-end autonomous driving solution; Wen Yuanzhi obtained Bosch Investment, which will develop the development of L2-L3-level autonomous driving software applied to passenger cars.

As a comparison, among the "progressive" players, Musk wrote on Twitter on August 15 that Tesla has produced more than 3 million cars, and its auxiliary driving system Autopilot will collect billions of billions. Miles and driving data.

The total mileage of auxiliary driving system has exceeded 17 million kilometers. "Data is the core essential of driving the maturity of autonomous driving, and the progressive route is the best path for data accumulation." Zhang Kai mentioned on this AI Day.

These industries seem to be declared a gradual successful victory.

However, there are different views in the industry, thinking that the current victory in the L2 field is still unknown whether it can gradually reach L4. In July 2021, Wang Naiyan, chief scientist of Tusson, posted an article on Zhihu "Is massive data a silver bomb for high -level autonomous driving? ", Ask a question," Can a large amount of data L2 evolve into L4? "

After demonstrating, Wang Naiyan believes that "if a system's architecture is designed for L2/L3 level (allowing system failures caused by a single point failure), then even more data is added, it cannot automatically complete the failure detection and processing. Complete the level of level. "

Of course, "Can you surrender, why can't you rise?"

During an open day event, the view of Xiao Ma Zhixing's engineer is that the use of data reduction will be relatively easy, that is, it is relatively easy to use the architecture of high -profile L4 -level autonomous driving to do L2 and L3 levels. However, if the low configuration does not even reach the L2 level data, it is difficult to use the test and verification of the entire L4 level system, and can only be applied in more limited areas.

In addition, the product strategy of Zhixing has been questioned.

At the end, Zhixing has proposed the "windmill strategy". The three windows of the "windmill" are passenger cars, low -speed drone ecological platforms and smart hardware, respectively.

This is a bit similar to Momenta, and the latter established the "one flywheel" technical route at the beginning of its establishment. Through the "data -driven algorithm", it continuously accumulated and iterated to promote mass production autonomous driving (MPILOT) and completely unmanned driving (MSD (MSD) ) "Two legs" walk. Because of this, it was evaluated by Wen Yuan Zhixing CEO Han Xu as "the two rabbits instead of one rabbit."

Similarly, according to "Late Latepost", a senior self -driving practitioner expressed his questioning of the strategy of Double Products: He believed that the coordination of mass production of vehicle auxiliary driving technology and the end unmanned delivery vehicle was limited. Technically, the former mainly detects and track the lane line line, and the latter is mainly based on map positioning and tracking. The focus is different. In terms of business, the two business lines belong to different branches of the industry and have different play methods.

As a proof, "Cyber ​​Automobile" in June last year has disclosed that the Tsathang cooperation operates the Wumi Multi -point supermarket located in Shunyi, Beijing, and has never opened unattended delivery. "Without investing in actual operations, it may be because the self -driving ability does not reach the level of operation, or the unmanned delivery vehicle is far from the ability he said."

When will I get out of the Great Wall?

At the end, Zhixing has today's achievements, which is obviously inseparable from the Great Wall's blood transfusion. Counting the We brand Mochaka, which is planned in September, HPILOT, the three generations of passenger vehicle assisting driving products, has been equipped with ten passenger models under the Great Wall.

This is related to the background of Zhixing. At the end of 2019, the Smart Express was independent from the Great Wall Motor Technology Center Intelligent Driving Prospective Division. Zhang Kai, the former director of the Great Wall Intelligent Driving System Development Department, served as the chairman.

This is a bit similar to the model represented by the United States GM and its Cruise, and the host car factory has differentiated a independent technology company to develop unmanned driving.

Cruise was acquired by GM for $ 1 billion in 2016. Great Wall Motor's Wisdom is like a general car to Cruise. Zhang Kai once said: "Smart Xing is a Chinese representative of the" car company+technology company "Cruise model."

In an interview in September 2021, Gu Weiyi described the relationship with the Great Wall. "The Great Wall itself is a very thick leg. Thousands of cars, why don't I serve the Great Wall well? "Gu's thought at the time was that the goal of Zhixing was to achieve the autonomous driving of L5, but the premise was that the company had to survive. In -depth binding the Great Wall, not only can it obtain stable income, but also use opportunities for cooperation to continue to obtain data from low cost and support its gradual way to the L5 route.

Such strategies have changed now. At the beginning of 2022, Smart Bank formulated and launched a flexible cooperation model, which can provide a full set of solutions, or cooperate in cloud services, hardware, software, and modules. It can also submit source code.

Since then, at the first AI DAY in 2022 held on April 19 this year, the "6P Open Cooperation Model" for passenger car driving was launched at the end. According to Zhang Kai, the bottom chose to create a model with the car company, not the supply model.

There may be a variety of factors behind this. For example, it involves IPO matters, and it can be reported that Yitong, which will be news about the IPO news of the US stocks not long ago.

Yicheng was jointly established by Li Shufu and Shen Ziyu, chairman of Geely Holding Group. Since its establishment, the company has provided millions of sets of car networking systems for Geely's brands and models, but it is also true that it is necessary to avoid the risk of affiliated transactions during the listing process. To this end, Yichang set the goal inside, and the total revenue should be 30%from Geely.

Earlier, in order to solve the problem of the order from a single major shareholder, Gu Weizheng pressed the treasure on the unmanned logistics vehicle. "I think it will consider the proportion of major shareholders' related transactions. Our more business forms are solved, not necessarily to enter the second OEM. "

Today, Zhixing has changed its strategy, looking for the second car group partner, and also reflects the performance of its unmanned flow vehicles.

According to Phoenix.com's "New Vision", Schiwang has been emphasizing that there is only cooperation with Great Wall Motors, highlighting its independence, and actively expanding other cooperation OEMs. But from the perspective of public information, breakthrough progress has not been made.

In addition, for Great Wall Motors, which is in the transition period, how to deal with the relationship with the slightest Zhixing is also the factors that the latter needs to focus on the development of fast lanes.

Reference materials:

The car produced "Autonomous Driving" into the city ", but I can't get separated from the map but cannot be separated from the map"

Late LatePost "Shougang, Meituan, Gao Yan participated in the Great Wall Holding Autonomous Driver Subsidiary, this year's industry financing has exceeded 25 billion yuan"

Cyber ​​Auto "To be a Chinese version of Cruise, is there any opportunity for Zhixing?" "

36 氪 "Gu Weiyi, CEO of Zhixing: No need to go out, the Great Wall itself is" very thick legs ""

Phoenix Technology "Can Zhixing become the next ace of the Great Wall? ｜ New Vision "

- END -