After GAC invested in Didi’s automatic driving, where is BYD Road, a self-developed ＂smart driving＂？

Source | AUTO core ball

Author | Kai Kai

On October 12, Didi Autopilot announced that it had been invested by Guangzhou Automobile Group. Guangzhou Automobile Capital, a wholly-owned subsidiary of Guangzhou Automobile Group, and Guangzhou Development Zone Investment Group will jointly set up a special fund in the same proportion, and the investment in Didi Autopilot will not exceed 149 million US dollars.

This undoubtedly adds another warmth to the autopilot track that has been silent for a long time.

Car companies are choosing their own self-driving roads. As an auto giant, BYD used to seem to be in a swing.

Two months ago, according to the South China Morning Post, insiders revealed that BYD had decided to abandon the cooperation agreement with Baidu on autonomous driving technology and focus on its own smart car software development, but as of the date of writing, neither party had responded to this matter.

In fact, Wang Chuanfu, the founder of BYD, has always had a very subtle attitude towards autonomous driving. He once said at BYD’s 2022 financial report exchange meeting that driverless driving is a vain thing, but at the same time he also emphasized the practicality of high-order assisted driving, thinking that assisted driving is equivalent to a god watching people driving, which is a real thing.

It is worth mentioning that Baidu, which cooperates with BYD in autonomous driving, started with L4 unmanned driving, which is also what Wang Chuanfu called "a vain thing". So, there may have been differences between the two in the concept of intelligent driving.

Although there are differences in ideas, as Wang Chuanfu said, "Intelligentization is the second half of the automobile revolution", and it is no suspense for BYD to enter the intelligent era. Just whether to "sell your soul" to seek outsourcing or stick to the road of self-research, BYD’s choice will probably change the pattern of the entire autonomous driving industry.

01 NOA suddenly "blossomed everywhere"

Since last year, the autonomous driving industry has further entered the cold winter, or the capital chain has broken, closed down and shut down, or scrimped and saved. In the final analysis, unmanned driving can’t really be "unmanned". There is still a gap between scene simulation and real road conditions, which makes it difficult to achieve business landing and capital naturally loses patience.

However, when the automatic driving of L3 and above is in trouble, the automatic aided navigation driving represented by NOA in the city is developing in full swing. According to the insiders, NOA is the entry threshold for unmanned driving. Only after NOA is widely popularized in cities will unmanned driving appear.

From a technical point of view, if autonomous driving is to become a reality, it is bound to be inseparable from a large amount of data accumulation and experience more road tests. Urban NOA can just become the entrance of data accumulation.

From the perspective of car companies, although consumers still have reservations about advanced assisted driving, it is a process from quantitative change to qualitative change. Once consumers get used to the intelligent driving function, it is not only the traditional factors such as control, chassis and battery that affect the purchase decision. The layout of NOA city is "preparing for a rainy day".

Therefore, major new energy vehicle companies began to enter the arms race in NOA, with Tucki and Huawei aiming at 50 cities, while Ideal and Great Wall directly shouted out the goal of 100 cities. As the champion of new energy vehicles, BYD obviously does not intend to give up this enthusiasm.

According to Time Finance, BYD revealed that it will propose a system-level solution for intelligent driving through the full-stack self-developed advanced intelligent driving assistance system "Eye of God". Zhao Changjiang, general manager of BYD, also said in Weibo that OTA push high-speed NOA function will be pushed one after another this year, and urban NOA function will be put on the agenda one after another.

Source: Weibo screenshot

At the same time, BYD has also raised the priority of intelligent driving. On the one hand, BYD has made frequent investments. In 2021, it successively invested in AI chip manufacturer Horizon and laser radar supplier Sagitar Juchuang, and also established a joint venture company Dipai Zhixing with Momenta, a supplier of autonomous driving software solutions, seeking cooperation in high-level intelligent driving.

On the other hand, BYD also adjusted the structure of its intelligent driving team. Huan Wang, the former head of intelligent driving research and development, left, and Han Bing, the former director of electronic integration department, took over.

During Huan Wang’s tenure, his plan for intelligent driving was to walk on two legs, including recruiting Tier 1 suppliers and self-research. However, Huan Wang’s point of view is to give priority to the Tier 1 supplier’s product solutions to help BYD quickly shorten the gap with its friends, but BYD’s intelligent driving ability has never broken through.

According to the "Red Star" report, BYD insiders said that Huan Wang was replaced by Han Bing because Wang Chuanfu was not satisfied with the progress of Zhijia.

After Han Bing took office, BYD Planning Institute, which is in charge of intelligent driving research and development, began to recruit talents vigorously. Li Ke, executive vice president of BYD, once said that BYD has recruited a software team of 4,000 to 5,000 people for autonomous driving.

If BYD continues to cooperate with Baidu in autonomous driving, it is obvious that such a huge technical team is not needed, and BYD has long planned to go it alone.

Source: boss direct hire screenshot

BYD doesn’t want to "sell its soul"?

Judging from the development speed of NOA in the city, the intelligent turning point seems to come sooner than expected, and it is reasonable for BYD to become more anxious.

However, the problem is that the cooperation between BYD and Baidu did not begin in 2022. As early as 2015, when Baidu established the autopilot division, Wang Chuanfu became Baidu’s autopilot consultant. In the following years, BYD has been cooperating with Baidu as a vehicle manufacturer, including building unmanned experimental vehicles for Baidu and adopting Baidu’s navigation and automatic parking technology.

But after so many years, the cooperation between the two is still "tepid". BYD’s intelligence level has not been widely recognized by the market; Between BYD and Baidu, there is no cooperation to launch self-driving production models.

The two sides have been "hand in hand" on the road of autonomous driving for so many years, but after all, there is no satisfactory result for each other. In fact, it has also laid the foundation for this "breakup".

On the one hand, BYD may not want to "sell its soul". As early as 2021, Chen Hong, the chairman of SAIC, publicly opposed Huawei’s car-making scheme, thinking that it would make suppliers such as Huawei become the soul, while the OEM would become the body.

There are similar "non-competition clauses" in the investment cooperation between Baidu and Weimar. Weimar will accept Baidu to empower its intelligent driving while winning Baidu’s investment, and Weimar will no longer be able to carry out self-research. For BYD, this is an unavoidable risk problem.

Especially at the moment when the "Huawei model" is accelerating, Huawei’s car-making partners are changing in a flowing way, but the brand of the world has always belonged to Huawei, and the traditional OEM has become more passive.

In addition, in the long run, self-research can further improve the enterprise’s ability to resist risks. Wu Xinzhou, a former vice president of Xpeng Motors Autopilot, said that only self-research can help enterprises to run in small steps, make more bright spots and differences, reduce a lot of uncertainties, and most importantly, reduce costs by integrating the industrial chain.

Finally, for BYD, which has come all the way from the era of fuel vehicles, its ability to become the sales champion of new energy vehicles in China has a lot to do with its ability to control in the automobile industry chain. "Self-research" has obviously become a muscle memory, which has a lot to do with BYD’s "technical fish pond" management system.

BYD advocates internal competition. Although different departments have different R&D emphases, they will also compete with each other, and whoever has a better scheme will be adopted in the end. Although such an organizational structure may lead to repeated waste of resources, it does provide a "wolf soil" for technology self-research.

BYD’s idea represents the attitude of some OEMs towards the self-research of intelligent driving, but there are also many companies that can’t wait for self-research and just want to get on the bus first.

For example, SAIC, which once put forward the "soul theory", invested in Momenta, an autonomous driving startup, as early as 2021. At present, its brand Zhiji has been equipped with Momenta’s L2+ solution. Geely’s brand Krypton has also expanded its strategic partnership with Mobileye, Intel’s autonomous driving department, saying that it will jointly develop a pure electric vehicle with L4 autonomous driving capability.

While not wanting to sell the "soul", we continue to cooperate with autonomous driving enterprises, which shows that the intelligent competition of new energy vehicles has begun to accelerate. Some insiders said that at present, autonomous driving has entered an inflection point, and mass-produced urban auxiliary driving ability and full-stack closed-loop ability are the "admission tickets" for enterprises to enter the second half of the competition.

At present, new energy car companies are facing two choices. One is to insist on self-research from scratch, just like BYD’s electrification process, which takes a long time to climb the peak. However, there is some uncertainty in this way, and the loss of admission tickets for enterprises means that consumers’ minds may be occupied by other brands.

Another option is for car companies to temporarily put down their attachment to the "soul", so that OEMs can get the packaging plan quickly and at low cost, exchange time for space, and constantly improve their own intelligent capabilities. However, the disadvantage of this method is that the control of technology by car companies will be relatively reduced, and the intelligent driving system may lack innovation and personalization.

What is the winning rate of self-developed car companies?

"Self-research" turns to the left and "outsourcing" turns to the right. How to choose the optimal solution? Let’s take a look at the challenges that car companies will face once they choose self-research.

From a technical point of view, intelligent driving is a large-scale dynamic system with hardware as the carrier and driven by algorithm software, including environmental awareness, independent decision-making, high-precision maps, AI chips, big data and cloud computing, large-scale model development and training and other technical difficulties.

In recent years, the investment and upgrade of car companies in hardware and software have never stopped. Take the hardware as an example. In the early years, the car camera was only a million pixels. Now, most of the models with assisted driving as their main selling point have been equipped with laser radar. With more and more targets in the car’s field of vision, the perception algorithm and prediction algorithm also put forward higher requirements for chip computing power.

However, hardware is only the foundation. If you want to develop from assisted driving to automatic driving, the core is still based on deep learning algorithm, learning and optimization through massive data training, forming a closed loop of obtaining data, processing data and optimizing data, which requires software to be powerful enough.

In these two aspects, some links can be solved by "money-making ability". For example, BYD has started self-research on core components such as laser radar, millimeter-wave radar and autopilot chip, and at the same time actively joined hands with chip and laser radar enterprises such as NVIDIA, Horizon and Sagitar Juchuang to speed up the research and development progress through large-scale investment.

In addition, Han Bing said that BYD has established a research and development fleet with more than 300 vehicles and accumulated more than 150PB of data. At present, it is developing a large data-driven model, and will build a BYD Supercomputing Center, and its data training closed loop will soon be opened.

Source: Beijing Zhiyuan Conference

But beyond that, there are still some problems that may not be solved by money. On the one hand, it is time, referring to Tesla, which proposed the whole data closed-loop capability based on BEV perception as early as 2020; Tucki also launched the first edition of the city NGP as early as Xpeng Motors Science and Technology Day in 2021. If the data is training iterations every minute, then BYD is at least two years behind.

On the other hand, compared with the new car companies, the traditional OEMs have to face a more complicated management structure. The development of intelligent driving often needs the cooperation of multiple departments, but a traditional OEM like BYD has many product lines and is responsible for its performance. So how can BYD adjust its R&D progress to ensure the differentiation among different product lines? If this competitive and collaborative relationship is not handled well, it is likely to slow down the progress of intelligent research and development.

Therefore, although BYD has sufficient confidence in financial resources and manpower and rich accumulation in electrification, it may not be too easy to realize full-stack self-research of intelligent driving, and it may even be more difficult than new car companies.

Therefore, it doesn’t seem to be the optimal solution of intelligent driving to simply research the whole stack or rely entirely on suppliers. Because in the background of more and more complex intelligent driving functions and higher integration, it is difficult for both car companies and suppliers to be competent for all development work. In the future, the division of labor in the new energy vehicle industry will be further refined, and car companies will gradually change from brand owners to service providers.

It is true that stabilizing the core technology will make car companies feel more secure, but quickly occupying the market and improving customer satisfaction are also important dimensions that determine the future development of the brand. Therefore, car companies must weigh their own strength and cooperate with suppliers in necessary links to win-win, which may be a better solution.

Although BYD has chosen self-research, it is also well aware of this truth. At present, it still insists on walking on two legs. Low-end models will continue to use external suppliers’ solutions, while mid-to high-end models will adopt BYD’s self-research solutions, first win the "admission ticket" for intelligent assisted driving, and then continue to seize time to shorten the gap.

Next, if BYD can really change space through time, as the largest and most appealing OEM at present, its self-research results may once again subvert the development of the industry, just as BYD is now leading by virtue of integrated technology, and even has the opportunity to redefine the service and charging model of intelligent assisted driving.

Of course, the premise is that BYD can successfully fight for its position in the intelligent campaign, and it has both the right time and the right place. Now BYD has made up for the harmony between people and see if it can do so.

References:

Red Star: BYD’s Intelligent Driving Revolution

Tan Qing said AI: BYD held hands with Baidu, and the new car "body and soul theory" was defeated by "life and death speed"?

HiEV Garlic Grain Vehicle Research Institute: "Car enterprises develop their own intelligence and drive the domain control, a high-end bureau with a few players participating"

* The pictures in this article are all from the Internet.

Focus on smart cars to help key decisions.