Baidu’s self-driving technology achieves the goal of getting on the bus, and may need to make up lessons

On November 15, Changan Automobile’s high-end brand Avita was officially released. This is the second car company partner officially announced by Huawei’s HI brand.

According to previously disclosed information, GAC will be Huawei’s third customer to adopt this cooperation model.

Not only that, Great Wall’s high-end brand Sharon Motor also recently announced that its first model’s intelligent driving solution will use Huawei’s MDC platform.

The Huawei HI brand was officially released on October 30, 2020. Looking back on one year’s transcripts, plus Celis, Huawei has won 5 car brands.

One month after the release of Huawei’s HI brand, on December 8, 2020, Baidu also released a major decision at the Apollo Ecological Conference.

Baidu Apollo reorganized its business and released ANP (Baidu Pilot Driving Assistance System) products for the mass-produced passenger car market.

It’s almost the same time dimension, and Baidu’s ANP products are the only partners who can disclose its brand to the outside world, but there is always only WM Motor, which is still its investment holding company.

This can’t help but raise a question. Also as a technology giant in the cross-border auto industry, why is Huawei’s autonomous driving technology landing so smoothly, while Baidu is a little difficult?

Behind this question, there is actually another question, that is, what is the current ADAS market structure under the background of the increasing penetration rate of intelligent driving and the pursuit of intelligent driving ability? And what are the notable trends?

Baidu is a veteran player in the domestic autonomous driving track. It has been in this track for 8 years since 2013, and has formed a rich product line.

On December 8, 2020, the Baidu Apollo Ecological Conference was held in Guangzhou. Baidu Apollo re-organized its business and divided its products into 4 major series: Smart Driving, Smart Cabin, Smart Map, and Smart Cloud, and called it Lego-style car smart solutions.

Compared with the progress of the intelligent cockpit, the intelligent driving business is struggling.

According to data released by Baidu in April this year, in the smart cockpit business, it has cooperated with more than 70 car companies in 600 models, and the Xiaodu in-vehicle OS has been mass-produced for more than 1 million vehicles.

In terms of intelligent driving, Baidu only stated that it has cooperated with GAC, Weimar, Great Wall and other brands in AVP (Autonomous Parking System) for mass production, and only Weimar W6 has achieved ANP.

So far, Weimar W6 can only implement the AVP function, and the ANP function has not been opened, and users have not yet been able to experience the effect of this function.

Baidu belongs to the first echelon player on the autonomous driving track. According to normal logic, ANP products derived from its L4 autonomous driving technology should have a certain market competitiveness. Why do car companies not buy it?

Let’s first explain the ANP function. ANP (Apollo Navigation Pilot) is translated into Baidu Pilot Assisted Driving, which refers to the automatic driving ability that can be turned on in specific scenarios of expressways and urban roads.

According to the automatic driving classification standard, this function can be called L3 level automatic driving, that is, automatic driving can be achieved under limited conditions, while requiring human drivers to keep taking over at any time.

However, L3 is a sensitive area in terms of regulatory supervision and the realization of technical functions. Therefore, domestic autopilot track players have tacitly avoided the concept of L3, calling Pilot Assisted Driving L2++ products.

At present, the models that claim to be able to achieve such capabilities mainly include Tesla Model 3, Xiaopeng P7, Weilai ET7, Zhiji IM, and Jihu Alpha S Huawei HI version.

Regarding the difficulty of Baidu’s self-driving car, there has always been a speculation in the industry, that is, the precautionary psychology of car companies for Internet giants. Car companies are worried that in the era of software-defined cars, Internet giants will take away the “soul” of car-making.

This kind of voice has actually existed since the beginning of Baidu’s Apollo plan. The industry generally believes that no car company will sincerely cooperate with Baidu, and Baidu will inevitably embark on the road of building its own cars.

However, without contrast, there is no harm. In the past two years, Huawei has frequently made swords in the automotive industry, and its firepower is obviously stronger. If Huawei can be accepted by car companies, why can’t Baidu.

In other words, the argument that car companies are unwilling to cooperate with Baidu out of precaution is not very tenable.

This points to another way of saying that the mass production of Baidu’s autonomous driving technology may not be enough.

A CEO of an autonomous driving solution start-up company once said to Cyber ​​Auto that car companies are very concerned about the experience of heavy production when choosing suppliers.

Internet giants do have resources and capital advantages, and they can take the lead in the technology research and development stage with money.

To be able to enter the OEM supply chain, not only must have strong technology, but also ensure strong enough engineering capabilities, supply chain management capabilities, assembly line assembly efficiency, and comprehensive considerations such as product safety, power consumption, and cost control.

He believes that ADAS companies have a first-mover advantage. When products from L1 to L2 are gradually delivered to car companies, a lot of experience in large-scale delivery and mass production has been accumulated. Moreover, after the products have been assembled and used in large quantities, the safety and reliability have been verified, and these capabilities cannot be piled up by burning money.

The person in charge of production and investment with a background in a car company also believes that most OEMs do not actually have such a strong awareness of prevention. They do not cooperate with Baidu because they are worried that Baidu is too strong. It may be because the technology itself is not enough.

He mentioned a phenomenon that Baidu spends a lot of money to poach people from car companies to do business, hoping to open up the supply chain system of car companies, but the actual effect is not obvious.

Others believe that it is an enterprise mechanism problem, saying that Baidu is becoming more and more like a state-owned enterprise with a heavy system.

In contrast, Huawei has an extremely efficient response mechanism. A supplier who has cooperated with Huawei told Cyber ​​Auto that, no matter when a demand is reported, basically within 5 minutes, the relevant staff will sit in front of the computer to solve the problem.

The market that Baidu encountered resistance is one of the most fiercely contested markets for autonomous driving players.

If the lack of mass production experience, immature technical polishing, and heavy enterprise mechanism are internal reasons, and the extremely fierce market competition is the external reason for Baidu ANP’s difficulty in getting on the bus.

Intelligent driving has become a new growth point in the automotive industry. With the gradual stabilization of the commercial vehicle ADAS market, the passenger vehicle ADAS market has become the focus of competition among autonomous driving technology companies.

If roughly divided according to the background, the main players in the ADAS market can be divided into four categories: global component suppliers, such as mainland China and Bosch; cross-border technology companies, such as Huawei and Baidu; local component giants, such as Neusoft Group and Desay SV; And a number of innovative start-up companies such as Freetech, Momenta, etc.

At present, Global component suppliers still occupy the main body of the ADAS market, especially in joint ventures and luxury brands with an unshakable position.

These enterprises have continued the advantages of in-depth support of OEMs, including cost advantages brought by scale, as well as experience advantages such as system integration and functional safety.

However, the products provided by Global component suppliers to OEMs are often delivered in whole packages, also known as black box delivery in the industry.

The possible drawbacks of black box delivery are that the vehicle matching effect is not good, the functions do not meet expectations, it is difficult to evaluate problems, and it takes a lot of time to improve the functions after problems are found.

At the same time, global component suppliers often have more complex mechanisms, and their feedback efficiency cannot keep up with the rhythm of the domestic market to a certain extent.

“The OEMs have long relied on Tier 1 to carry out strong delivery in the form of a whole package of software and hardware. This brings a long delivery cycle, high delivery costs, and pain points that cannot be fully self-controllable”, Yang Ke, vice president of Guoqi Intelligent Control Said to Cyber ​​Auto in an interview.

Car companies are becoming more and more aware of product self-development, and powerful car companies want to be more involved in the development of underlying technologies.

Yang Ke introduced that the participation of car companies is not only limited to the draft planning, but also deeply involved in the planning and design of specific functional scenarios, the implementation of core codes, and the mastery of software upgrades.

Therefore, in general, for domestic passenger car customers, the competitiveness of Global parts suppliers is weakening, which brings opportunities to domestic parts suppliers.

For a specific example, BYD’s earliest supplier of ADAS functions is mainly Bosch, which can implement some L1-level ADAS functions.

It may be that the cooperation between the two parties did not meet expectations, and BYD subsequently replaced the supplier with Venier. With the support of Venier, BYD’s new cars can reach the level of L2 ADAS.

But Venier still seems to be unable to fully meet BYD’s needs. At present, BYD is considering teaming up with the domestic autonomous driving supplier Momenta and the domestic chip leader Horizon to build the next-generation autonomous driving system. At the same time, BYD has begun to form its own autonomous driving team.

BYD’s actions represent the general direction of independent brands in the layout of the autonomous driving business. In the long run, full-stack self-development of autonomous driving led by car companies must be the goal pursued by capable car companies.

Full-stack self-research requires huge investment. In order to keep up with the pace of market launch, car companies often choose to cooperate with suppliers in the early stage. The idea is to break up the technology, start with what they can do, and then gradually supplement it. .

To give another example, the ADAS technology solution on the Ideal ONE 2021 version was developed in cooperation with the autonomous driving technology company Hongjing Zhijia.

However, Shen Yanan, president of Lili Automobile, once said that Lili Automobile’s next-generation car will develop its own central domain controller and make its own software on this basis.

He explained that the logic of the ideal car layout for autonomous driving is to first master the capabilities of the hardware and then the capabilities of the operating system, and then it is possible to make a good chip. The chip should be optimized according to the application layer and operating system. Follow this path to preparation and layout.

At present, it can be seen that there are three main modes of automatic driving in the layout of OEMs:

One: Tier 1 cooperates with chip manufacturers to package and sell integrated solutions to OEMs.

Second: some powerful OEMs have self-developed full-stack software and hardware, including core software and hardware such as chips, operating systems, middleware, and domain controllers.

Third: The main engine factory cooperates with Tier 1, the former is responsible for the autonomous driving software part, and the latter is responsible for hardware production, middle layer and chip solution integration.

The first is the model provided by Global Tier1, but at present, it is becoming more and more difficult for OEMs to accept this cooperation model.

The second is the route taken by a few car companies at present, represented by Tesla and Xiaopeng. This model has outstanding advantages, which is conducive to controlling product functions, improving R&D efficiency, and higher security, but at the same time, investment is also Huge, there should be a small number of powerful companies taking this route in the future.

The third mode may become the main mode of cooperation, with car companies controlling the parts that can differentiate the user experience, while the basic software and hardware development work is left to suppliers.

The automotive Electronic and electrical architecture is evolving towards centralization, from ECUs to DCUs, DCUs to central calculators, and then to cloud computing and collaborative computing. From the perspective of the layout of car companies, the next generation of products generally enter the domain control stage.

The continuous evolution of technology will bring opportunities for innovative companies, and domestic component giants are benefiting from the development of autonomous driving business.

Looking at the 2021 semi-annual reports of listed companies such as Desay SV, Joyson Electronics, Thundersoft, and Neusoft Group, it can be found that the intelligent driving business-related sectors of these companies have grown significantly.

At present, the penetration rate of my country’s ADAS market is still relatively low, and it is expected that it will continue to grow significantly in the next few years. The market acceptance of ADAS functions has been verified in the new energy vehicle market.

According to the “Monthly Observation Report on Intelligent Connected Vehicle Products and Industry Development” released by Cyber ​​Auto, in September 2021, the number of new energy vehicles equipped with L2 intelligent driving was 136,732, with a penetration rate of 42.17%.

In the face of a good market, how should players seize the opportunity?

For innovative start-up companies, follow the direction of technological evolution to make forward-looking layouts, and at the same time deeply bind OEMs to hone the ability to implement technology, not greedy for more and faster, but to be stable and solid.

For traditional supplier giants, if they want to continue to maintain market competitiveness, they may need to be more flexible in their models and more agile in feedback on the basis of reliable technology.

In a nutshell: the technology should be hard, and the figure should be soft.

And these two points may be what Baidu lacks. Judging from the progress, it seems that Baidu’s self-driving technology still needs to make up lessons.

As the industry has previously judged, Baidu must embark on the road of building its own cars. At the beginning of this year, Baidu announced the formal establishment of a smart car company, Jidu Automobile, to enter the automotive industry as a vehicle manufacturer.

Can the construction of Jidu Automobile help Baidu to open up the two veins of Ren and Du of technology? Let’s stay hopeful.

The Links:   CM300DXD-24A QM50TX-H