Text | Angel Lee
Edit | Su Jianxun
A new unicorn in the autonomous driving industry was born.
36Kr exclusively learned that Zhixing has obtained nearly 1 billion yuan of A round financing and a post-investment valuation of more than 1 billion US dollars, becoming one of the few emerging autonomous driving unicorns in China to achieve mass production of intelligent driving.
The investors in this round are Meituan, Hillhouse Venture Capital, and First Trip Holdings. It is understood that the funds raised are mainly used for investment in automatic driving research and development and the construction of talent system.
This is the third round of financing received by Chi Heng this year. On February 10 this year, Zhixing received a financing from Great Wall Motors, the funds are unknown. 12 days later, Zhixing completed hundreds of millions of yuan of Pre-A round of financing, including Shougang, Meituan, Hillhouse Capital and so on.
Such an intensive and rapid financing rhythm is rare even in the current situation when the intelligent transformation of automobiles is in full swing. From the perspective of annual development, Zhixing is expected to become one of the fastest growing autonomous driving companies in the domestic autonomous driving track.
Zhang Kai, chairman of Zhixing, judged: "2022 will be the most critical year for the development of the autonomous driving industry. Competition in the field of passenger car assisted driving will officially enter the second half, and autonomous driving in other scenarios will also officially enter the first year of commercialization. In 2022, the number of projects at the end of the millennium is multiplied exponentially, the passenger car assisted driving project is expected to expand by 7 times, and the unmanned logistics vehicle project is expected to expand by 3 times. Sufficient technical ammunition and funds are now available. ”
Image source: Zhixing
Founded in 2019, Zhixing is independent from the Intelligent Driving Forward-looking Branch of Great Wall Motor Technology Center. Part of the core team comes from the intelligent driving R&D team inside the Great Wall, and part comes from technology companies such as Baidu and Huawei, and the current team size is nearly 600 people.
Born out of the Great Wall, the self-driving startup has a higher starting point than its peers, and even earlier than some new car manufacturers in the release of high-end assisted driving functions.
Previously, the noH of the intelligent pilot assistance system released by Zhixing in October has been mass-produced in Weipai Mocha models. As a high-level driver assistance function, Pilot Assisted Driving can help vehicles realize autonomous driving from point A to point B within the range of high speed and urban expressway covered by high-precision maps. A similar function of the ideal car, the navigation assistance function, was only introduced to car owners in December.
Some time ago, 36Kr experienced the NOH function during the evening rush hour of the East Fifth Ring Road, and successfully realized the functions of automatic up and down high-speed ramp, automatic lane change, automatic follow-up and so on.
In the process of interviewing 36Kr, Gu Weihao, CEO of Miller Wisdom Company, said many times that the scale landing is the most core advantage of Miller Wisdom Bank. On the one hand, the scale landing means that it can accumulate massive amounts of scene data; on the other hand, large-scale mass production also means that it can get more price advantages and suppress the price of the product even lower.
According to official data, by the end of 2022, the assisted driving system of Millima Chi Heng is expected to include more than 40 models that have been/ will be installed, and the orders are all from the Great Wall, covering the Great Wall Wei brand mocha, tank, Haval and other brands. Zhixing previously said that it expects the total number of passenger cars to be carried in the next three years to exceed 1 million units.
The accumulation of massive data is equivalent to "feeding" the continuous feed of the assisted driving system, seeing enough scenes, the more scenarios the system can cope with, and finally achieve complete unmannedness through continuous iteration.
Admittedly, the scale of mass production will always catch up one day. "But the development cycle of the car project is very long, and some of the things that Tier1 is doing today may not be apparent until 10 months later." Whoever can get on the car faster, the system ability is faster to close the loop, and the product force becomes stronger, there is a better chance to eliminate others. ”
The time difference brought about by the landing of the scale gives Zhixing time to think in practice and how to achieve a stronger assisted driving ability system.
In Gu Weihao's view, the evolution to advanced autonomous driving is a complex system engineering, but it is not as "insurmountable between low-level and advanced technologies" as the industry says." ”
He believes that the evolution of autonomous driving is not just an improvement at the algorithmic level. "Sensors must have progress, calculators must have progress, data transmission mechanisms and methods must progress, and the software structure and software algorithms in the cloud must actually have progress, coupled with a large number of complex scenarios." Only after seeing many scenes, the software system has the ability to pick out different scenes. ”
Take the pilot assisted driving NOH of The Ultimate Wisdom Walk as an example. In May 2022, the Company plans to expand the driving range of the Military Navigation Assisted Driving to urban roads. However, the density of traffic participants and traffic flows on urban roads is much greater than that of high-speed sections.
The city-level pilot assisted driving solution will have a richer sensor lineup: 2 lidar, 12 cameras, which will make the system's perception more accurate.
In addition, a related data processing system is also being polished internally. Through the construction of the system architecture, the massive data accumulated by Zhixing in the mass production vehicle can establish a real data backhaul mechanism to achieve rapid iteration of the assisted driving system.
In addition, The Company also launched the ICU 3.0 autonomous driving computing platform with Qualcomm, using Qualcomm's 5nm autonomous driving chip, and the platform has a single board computing power of 360TOPS (computing power units), which can be continuously upgraded to 1440TOPS in the future. The computing power of thousands of TOPS is also expected to support the evolution of Zhixing from assisted driving to advanced automatic driving.
Zhang Kai, chairman of Zhixing, and Gu Weihao, CEO
Looking back, the deep binding of the Great Wall did allow The Wisdom Of The End to quickly open up the market. So, how does Zhixing consider other ecological cooperation?
In this regard, Gu Weihao said: "Our own positioning is still a driverless company. In fact, there are many forms of our products, but today we have seen passenger cars and unmanned logistics vehicles, but next year we will definitely see more product forms. ”
The implication is that Zhixing may enter a new field of autonomous driving in the future, driverless freight or driverless buses.
"It's all possible, but we still follow our own judgment and won't do products that burn money for several years without seeing the point." It is our criterion to be able to commercialize as soon as possible, and to apply our technical accumulation in these areas of passenger car logistics vehicles. ”
It is not surprising that Zhixing has these thoughts.
Because in addition to the mass production of assisted driving in the passenger car field, Zhixing has previously been involved in the field of low-speed driverless terminal logistics.
In this field, Zhixing assumes the dual role of manufacturing and independent operation. On the one hand, backed by the Great Wall's automobile production and manufacturing capabilities for many years, Zhixing is providing manufacturing, factory testing and verification, scale mass production and other services for Ali's driverless logistics distribution vehicle "Little Wild Donkey". Meituan is also one of its customers.
According to 36Kr, at present, Ali unmanned car little wild donkey has completed a total of 350 vehicles, will complete 1,000 units within the year, and will expand to 10,000 in the next three years.
On the other hand, Zhixing also personally went down, and the joint supermarket chain Wumei duoduo started the driving and distribution operation of the terminal logistics unmanned vehicle.
Previously, Gu Weihao proposed a three-law of automatic driving, that is, the development of automatic driving will follow the development law from "low speed to high speed, from carrier to manned, from commercial to civilian". It is also using its practice to verify the applicability of the law.
But at present, intelligent driving as the core direction of car intelligence, its smooth road may only go halfway, how the road ahead has not yet been explored. In the face of the long journey, the end of the zhixing may also need to show more hole cards.
The following is an excerpt from an interview with Gu Weihao, CEO of 36Kr and Mi mo Zhixing (excerpted):
36Kr: At present, there are three major businesses: assisted driving, terminal logistics unmanned vehicles, and intelligent hardware, but assisted driving and low-speed unmanned driving face completely different driving scenarios, how to open up the two services?
Gu Weihao: There are actually several levels involved in this, from the hardware level, the sensors and domain controllers for unmanned vehicles in terminal logistics are converging with passenger cars, and the hardware level is actually interconnected. At the software level, both systems are the same architecture, and data acquisition, storage, computation, and verification are all systems.
The third level is the scene, our logistics car operation or commercial supermarket distribution, is equivalent to driving on the auxiliary road of the city's open road, and the auxiliary driving of passenger cars will gradually enter the city, so the two scenes are partially overlapping.
36Kr: Has your terminal logistics trolley already got the right of way on the open road?
Gu Weihao: We are now cooperating with local governments, which have their own publicity process. If we are past the publicity period, we can test and trial operation in a specific area.
36 Kr: How will The Ultimate Wisdom Walk evolve to more advanced autonomous driving?
Gu Weihao: This is actually a process of systematic development. First of all, the sensor is gradually evolving and developing, from the past 1 million pixels to 8 million pixels, and then to 12 million pixels, the development of the camera itself will make the perception of the machine clearer and more real. The cost of lidar is also changing dramatically every year, the price is getting cheaper and cheaper, and the accuracy is getting better and better, so the sensor itself is making great progress.
This is followed by the calculator. In the past few years, the computing power of the computing platform has increased exponentially, because the accuracy of the sensor and the information carrying capacity are increasing, which will inevitably lead to an appropriate increase in the computing power of the calculator.
There is also data transfer. Because of the development of sensors, data is becoming more and more numerous, the density is also increasing, the entire computing structure system is actually going to undergo a big change, so the transmission also needs to have a great development. After the data is backhauled, more complex training and larger-scale data processing are done in the cloud.
So self-driving software is actually a system of work. Sensors must be improved, calculators must be improved, data transmission mechanisms and methods must be improved, and the software structure and software algorithms in the cloud must be improved, coupled with a large number of complex scenarios. Only after seeing many scenes, the software system has the ability to pick out different scenes, the machine sees more and more, can be trained, and the real L4 automatic driving can come.
36Kr: As one of the few companies in the industry to launch pilot assisted driving, how to maintain the momentum at all times?
Gu Weihao: Scale. The new car-making force has generally reached almost 100,000 this year, but we already have tens of thousands this year. There will be more models on the bus next year.
36 Kr: But there are others in the industry who can do it on this scale.
Gu Weihao: But the development cycle of the automobile project is very long, and some of the things that Tier1 is doing today may not be apparent until 10 months later. Whoever can get on the car faster, the system ability is faster closed-loop, and the product force becomes stronger, and there is a better chance to eliminate others.
36 Kr: But the driver assistance function in the end, may become a standard thing?
Gu Weihao: It will be standard, but the ability is not the same, "yes or no" and "good or bad" are not the same, "good or bad" will determine who will be left in the future finals.
36Kr: Deep binding of the Great Wall is undoubtedly a way for Zhixing to quickly open the market, but will this deep binding have an impact on the company's entire ecological cooperation in the future?
Gu Weihao: Our own positioning is still a driverless company. In fact, there are many forms of our products, but today we have seen passenger cars and unmanned logistics vehicles, but next year we will definitely see more product forms.
36 Kr: More product forms refer to products in more unmanned driving areas? Such as driverless freight or driverless buses?
Gu Weihao: It is possible, but we still follow our own judgment and will not do those products that burn money for several years without seeing the point. It is our criterion to be able to achieve commercialization as soon as possible, and to be able to apply our technology accumulation in the fields of passenger cars and logistics vehicles.
36 Kr: When will it be your profit node?
Gu Weihao: This still depends on the development of our own business, our goal is indeed to be listed on the science and technology innovation board, but I think these are not particularly strong expectations.
36Kr: How much do you think the mileage of assisted driving has accumulated to be a qualitative breakthrough?
Gu Weihao: At present, I think it is all a process of quantitative change, and we cannot see the qualitative change. It's all quantitative change, but this quantitative change is also enough.
36Kr: The current assisted driving system is only on this highway, and then everyone will go to the city assisted driving, how do you think it can be achieved?
Gu Weihao: This is actually a very systematic and complex thing, considering the participants of the traffic flow, the density of the traffic flow, etc. Urban roads and highways are very different, urban roads have more intersections, more traffic lights, more two lanes to three lanes, lane lines are not clear, there may be some cone barrels, obstacles, temporary traffic control, much more complicated than high speed. The traffic participants themselves have become more complex. In addition, the density of traffic flow, the density of morning and evening peaks and normal times is different, and the conditions involved are also different.
We will break through by adding sensors, and the city version of the pilot driver assistance solution released in May 2022 will have two lidar and 12 cameras, which will make the system perceive more accurately.