Speed at the end: China's self -driving driving fastest 1000 days

Jia Haonan sent from the Gafei Temple

Quantum position | Public account QBITAI

1000 days, a transcript:

The intelligent auxiliary driving system is equipped with nearly 10 mass -produced passenger cars.

The user uses auxiliary driving mileage for more than 15 million kilometers.

The end logistics unmanned delivery vehicle operates nearly 90,000 orders.

The high -level leading assistance function covering the urban road is about to mass production.

Native

This is the model and speed of China's mass production autonomous driving unicorns. In the process of autonomous driving exploration, it is a new speed from technological research and development to commercial landing and refreshing the industry record.

Exposure will be exposed inside

Recently, there was a internal meeting at the end of the day when it was about to be established for 1,000 days.

Entrepreneurship from 0 to 1, employees from 0 to nearly 1,000 people.

Inside the slightest Zhixing, we made a comprehensive summary and thinking on the technical, organization, business, product, etc. since the establishment of 1,000 days:

1,000 days were established, and the problem of delivery from R & D to mass production from 0 to 1 broke through multi -model production autonomous driving.

How to proof? At the end of Zhixing, there are several different dimensions of evidence.

The first is product progress.

Three directions, HPILOT, unmanned flow vehicles, and intelligent hardware.

In each direction, there are records of refreshing the industry.

比如HPilot已经在近10款车型上量产交付，3.0版本能力拓展到城市复杂道路场景；无人物流车小魔驼3.0实现10万元级零售价，同时还具备L4级功能；智能硬件上， The self -developed single -board computing power of the autonomous driving computing platform, the small magic box, reached 360TOPS computing power.

What supports mass production delivery is the autonomous driving technology system.

The end will summarize their technology as the "emphasis" line. The core is to use Transfomer to fuse the data of different camera data before, forming a mapping BEV (View View perception) space, and naturally has the integration capabilities of the laser radar point cloud map.

This is a similar technical solution to Tesla, the purpose is to solve the core goal sense of autonomous driving.

At the organization and business level, Zhixing summarized a systematic methodology. For example, the principle of 6P open cooperation proposed by Zhang Kai, the chairman of Zhixing: Zhang Kai:

△ Zhang Kai, chairman of Zhixing

In addition, there are 12 principles of leadership, Amazon's reverse work method, open cooperation ecology, and so on.

Judging from the tone of Zhixing, this summary is a qualitative description: what is doing and how.

The focus that is easy to be followed is "breakthrough", "multi -model", "delivery" and so on.

But the real point is 1,000 days. This is the most familiar label since its establishment: fast.

Many industries "first" and "first" appeared in a 1000 -day transcript.

Curvit

The most intuitive thing about the "fast" of the industry's "fast" in the industry is the mass production of autonomous driving systems.

The HPILOT intelligent driving system has now covered tens of thousands of models under Great Wall Motors, including tanks, WeChat, Haval and other popular brands.

Therefore, the first "first" of Zhixing is the scale of mass production coverage. Whether it is Tesla, new forces, or other autonomous driving companies, the scale of tens of thousands of installed capacity in about 3 years is unprecedented.

According to the production schedule that has been finalized, at the end of this year, HPILOT will cover more than 30 models, and the installation volume will reach a million levels within two years.

The "first" of scale brings the "first" of speed.

HPILOT, from the 1.0 version of the mass production production in May 2020, has accumulated more than 15 million kilometers of user mileage in more than 1 year. 10,000 kilometers of high -speed and urban expressway.

In the process, the actual road test data of the user has helped HPILOT to complete three large version iterations. The latest version (Great Wall Weizai Mocha PHEV) is the most important update of immediately production delivery is the city's NOH.

Being able to respond to the complex roads of the city, including traffic signals, complex intersections, circles, mixed people and vehicles.

This ability is also a highland for the leading driver of the head this year. Other powerful players, including Xiaopeng NGP, Tesla NOA, and so on.

Therefore, the speed of mass production autonomous driving system is "first", which is the result of comparison with the industry's head.

In addition, there are several "firsts" and "first" on other business lines on the self -driving business line:

Behind the scale and speed of mass production, what supports these is MANA, the first domestic autonomous data intelligent system established by the self -developed self -developed.

MANA contains many sub -modules, including data acquisition, transmission, perception, calculation, verification, etc., which constitute the data driving capacity of the slightest Zhixing.

In addition to the data contribution of the actual road test mileage of HPILOT on the car, another important core is the huge amount of learning, simulation, and verification of the system under the system under the Mana system.

At present, MANA's study time has exceeded 300,000 hours, and virtual driving age is equivalent to human driver driving for 40,000 years.

The progress of mass production and data closed -loop constitute a transcript of "the industry's fastest 1,000 days". How to evaluate?

At the level of financing, the progress of Zhixing has been recognized by high -quality capital in the industry.

As of April of this year, it has completed 3 rounds of nearly 2 billion yuan in financing. Investors include Meituan, Shougang Fund, Qualcomm Venture Capital, Gaoma Venture Capital, First Varling Holdings, Jiuzhi Capital and so on.

At the level of the autonomous driving industry, Zhixing has also built a new model.

At the time of the birth of Zhixing, there were already three models in the global autonomous driving industry.

The first is the Waymo owned by Google, holding high, and directly aiming at L4, trying to iterate the system to "unattended" through road testing, simulation and other means before promoting commercialization.

Waymo, as a pioneer of autonomous driving, explored the iron law of "data -driven" for the entire industry, but was subject to the efficiency of data collection on mass production.

As a latecomer, Musk deeply understands the importance of data -driven, but put aside Waymo's concerns about safety in practical means, and achieves large -scale mass production and data collection of autonomous driving systems by selling cars.

The value provided by Tesla's exploration is to make the industry clearly recognize the first principle of autonomous driving (sensor scheme), and the only way to reach the final (data -driven).

But Tesla's radical strategy also caused the shortcomings of the autonomous driving system at this stage to cause huge controversy.

The third model is the CRUISE of the GM. In terms of business, it is also promoted to the intelligent driving of passenger cars and the L4 autonomous driving two routes in the same scenario. It tries to mass production projects through passenger car production projects to feed the landing of L4.

The core behind is the universality of autonomous driving technology at the core sense of recognition.

At the time of the birth of Zhixing, it was regarded as the Chinese version of Cruise because the business line was similar.

But the speed of iteration, research and development, and mass production is obviously much faster than Cruise.

The fundamental reason is to absorb the essence of the pioneers of various industries. Under the premise of mass production, it has seized the essence of autonomous driving and built a set of its own high -efficiency data closed -loop iteration system.

Behind the transcript

Within 1000 days, what is the factors of driving the first echelon player in mass production of autonomous driving?

Zhang Kai, chairman of Zhixing, has given the "formula" at the end:

(Leading data intelligent*stable mass production capacity*security) ecology.

The formula does not fall from the sky, but stems from Zhixing's correct judgment on the industry.

The essence of this formula is based on data intelligence as the core, and promotes mass production from three product lines: low -speed unmanned flow vehicles, passenger vehicle autonomous driving products, and intelligent hardware.

After mass production, the three business lines also nurtured data to data intelligence to form a positive iteration of autonomous driving capabilities.

Therefore, at the beginning of the establishment, the direction and end point of Zhixing had already clarified the direction and end point, and on the top of this, "reversing" the leadership 12, reverse work method, customer service method, and other specific methodological and organizational principles.

Similarly, "starting for the end", Smart Xing is on the technical route, and also focuses on the cutting -edge development of AI. For example, the rapid changes of computer vision technology from CNN convolution networks to attention models, as well as the large -scale transformer originally used in language models The huge potential in the field of vision and so on.

These are the "driving factor" that Smart Traveling is seen.

What is not so obvious is the team and model of Zhixing.

Born in Great Wall Motor, independent operations, independent financing, and given high autonomy.

The business model gets rid of the long decision -making chain and conservative thinking of traditional car manufacturers, and manages operations by technology companies.

The other aspect of fully stimulating creativity and enthusiasm is that Great Wall Motors did not give any restrictions on the slightest wisdom in terms of business expansion and financing listing.

This makes the genes of the slightest Zhixing completely different from other autonomous driver players. It has both AI technology companies and the engineering strength of the OEMs.

Gu Weiyi, CEO of Zhixing, is one of the old Baidu autonomous driving veterans, which brings a deep understanding and forward -looking of autonomous driving technology.

△ Gu Xing CEO Gu Weiyi

Zhang Kai, the chairman of Great Wall Motors, represents the ability of Zhixing to productize and engineering in autonomous driving technology.

Therefore, Zhixing has gone out of a new autonomous driving mode:

Technically returning to the first principle of autonomous driving, taking a heavy perception data -driven route.

In mass production, relying on the channels of the Great Wall Motors of the Great shareholders, the scale of large -scale landing is achieved quickly, forming a closed -loop data.

Although this model has a unique advantage, it has also been proof by business progress, investors' recognition, etc., and it has indeed become open.

But outside Great Wall Motors?

At the 6th at the end of September, the 6th at the end of the John AI Day, it is said that it will also officially announce the new technology and product results.

In addition, the technical background of Zhixing has always been the most watched part of previous AI Day. This activity has gradually developed into an AI conference focusing on the frontier of autonomous driving.

It is said that the world's top AI scholars, including Zhang Yaqin and Jia Yangqing, will participate.

China's autonomous driving technical forward -looking, commercialization process, and industry model exploration ... AI Day in September is a window that must not be missed.

-Finish-

- END -