Shulou(Shulou.com)12/24 Report--

To survive winter is spring.

Author | Wang Lei Chumen

"you can't see the complete logic of completely self-driving, and you may need to find another way."

He Xiaopeng recently expressed his views on self-driving on Weibo, due to the recent Waterloo of GM's self-driving company Cruise, which led to the departure of CEO and the near collapse of its business.

Before that, many people were pessimistic about completely self-driving, but Xiaopeng, as the first echelon player in the domestic intelligent driving auxiliary camp, the leader was not optimistic about the development of "fully self-driving", which still caused a lot of heated discussion.

Where is the way out for companies that aim at completely self-driving?

01. More than one boss sings bad songs and first confesses the background.

Before the accident, Cruise had just completed the milestone of the global commercial landing of self-driving technology, when the California Public Utilities Commission issued a license for its first operation, allowing Cruise to provide round-the-clock taxi service to the general public like ordinary taxi companies, and to raise fees during peak hours.

However, due to frequent accidents, less than three months later, Cruise was pressed the ESC button, and the star company of autopilot, which was given the hope of commercialization, returned to pre-liberation overnight.

On the defeat of Cruise, he Xiaopeng commented, "A clear road map can be seen for high-level autopilot or someone who is fully self-driving, but there is no complete logic for completely self-driving. I even think it may be necessary to find another way."

In order to avoid the introduction of war, a "personal opinion" prefix is added.

In fact, it's not just he Xiaopeng. As early as 2019, Thomas Sedran, head of commercial vehicles at Volkswagen Group, made a similar view, which sounded more pessimistic.

"solving this problem is as complex as sending humans to Mars. Fully autonomous driving means that all around you are equipped with the latest generation of mobile infrastructure, constantly updated high-precision maps and near-perfect road markings. Do you think this is possible on a global scale? it's obviously not realistic."

It was 2019, autopilot was in its infancy, and investment was hot.

But year after year, there is still an endless stream of "bad talk" about fully autopilot, and even the tough-mouthed Musk admitted on Twitter that it was more difficult than he thought.

Shortly after Musk's pessimistic comments, in July 2021, Su Jing, the former president of Huawei Intelligent driving, even said: "I will not see L5 autopilot in my life" and "companies that do RoboTaxi at this stage will be ruined." And Su Jing also paid the price because of such remarks.

It's not hard to find that the bad guys are often "insiders" who even punch themselves in the face. George Holtz, founder of Comma.ai, a self-driving startup, bluntly says that fully self-driving cars are a hoax!

At the investor communication meeting after BYD's 2023 earnings conference on March 29 this year, when answering investors' questions about the development progress and plans of BYD's smart driving.

Wang Chuanfu said, "self-driving is a new costume of the emperor", "self-driving is bullshit, hoodwinking", "held in by capital", and "ultimately is just an advanced auxiliary driving."

To some extent, this directly denies the future of complete self-driving.

02, the watershed of complete self-driving in 2009, the technology giant Google began the research and development of self-driving project, which also started the research and development boom of self-driving technology. Now self-driving has experienced more than a decade of development, from the global popularity, after several years of high-spirited, came to the "insipid after passion."

Now people are more pragmatic than the pursuit of full autopilot: landing first. After all, by 2023, there will be no company that can be said to have achieved full autopilot.

According to the SAE (American Society of Automotive Engineers) grade classification, autopilot can be divided into six L0-L5 levels.

L0 is the fuel car we drove in the early years, without any autopilot assist function, completely by human driving and control of the car.

L1 and L2 levels can achieve auxiliary driving and simple partial autopilot, such as the current vehicle standard lane maintenance and constant speed cruising functions, at this stage the person is still the main body of the vehicle.

At the L3-L4 level, there is conditional autopilot, and full autopilot can be achieved within some autopilot operation design range (ODD).

Only when it comes to L5 is fully autonomous self-driving in a real and complete sense, without human intervention or human supervision.

In this process, with the development of the industry, a series of concepts such as completely self-driving, high-level autopilot and fully autopilot have been born.

For example, he Xiaopeng mentioned that someone is fully self-driving and completely self-driving.

According to our common sense, the development process of things is from low to high, which is in line with the normal logic. But now some autopilot companies are doing the opposite, starting directly with advanced autopilot at the L4 level.

This also makes self-driving technology in the development process gradually formed two main technology directions, one is self-driving companies such as Cruise and Waymo, from the very beginning is the most advanced, in one step to develop self-driving Robotaxi.

The other is to take the car company as the representative of the step-by-step route, starting with the low-level L2, and then iteratively upgrade.

If Google started the craze of self-driving in 2009, then after 2015, the popularity of artificial intelligence and big data technology pushed autopilot to another wonderful part. Capital poured in frantically, a large number of start-ups were set up one after another, and car companies also hurriedly announced that they would start the journey of self-driving.

Just like people can not always be young, capital will not always have one favorite.

In the past two years, with the deepening of technology and testing, the commercialization of self-driving cars has fallen short of expectations, and the enthusiasm of visible capital for the industry has begun to fade.

At a time when commercialization is a long way off, many companies are still insisting that the investigation and suspension of Cruise's old rival Waymo,Cruise has not affected Waymo's commercial operation of driverless taxis in San Francisco.

Waymo is also understood to have operations in Phoenix and Austin, a pilot self-driving taxi service in Los Angeles, and plans to deploy the next generation of self-driving cars next year, but Waymo also faces huge public mistrust.

Waymo, whose only income is Robotaxi, suspended its self-driving truck program last month after spending hundreds of millions of dollars on it.

In addition to Robotaxi, it is easier to land completely self-driving in ports, logistics, buses and other directions, and some companies aim at these easier scenes from the beginning.

After Robotaxi, the domestic self-driving company Xiaoma Zhixing opened the operation of Robotruck (pony smart card). In July this year, Horse Zhika and Sany heavy Industry formed a joint venture, Yiji Zhika, to carry out the research, development, production and sales of L4 self-driving trucks.

Apollo, owned by Baidu, is a representative of fully autopilot in China, providing both self-driving solutions, self-driving taxi platform "radish run" and L4 self-driving bus Apollo Robobus and other fully autonomous driving scenarios.

Of course, more high-flying enterprises are also starting to turn to the more practical L2.

03. I hope that autopilot, which is still mentioned now, is in fact a very general concept. It can refer to auxiliary driving, autopilot in the case of someone, or full autopilot in an unmanned state.

In other words, autopilot is not difficult to land, what is difficult to land is fully autopilot.

From the "alternative approaches" of those self-driving companies above, we can also see that the full autopilot industry is indeed what he Xiaopeng said-there is no complete logic and may need to find another way.

In fact, these fully autopilot companies have gradually begun to "find another way".

Layoffs, closures and plummeting market capitalization have become three high-frequency words for self-driving companies. Mobileye, a self-driving company, shrunk its valuations again and again last year, from an initial $50 billion to $30 billion, before going public at just $16 billion; Waymo launched three major layoffs this year alone.

Why is full autopilot so difficult to achieve? There are just a few important factors-capital, technology and regulation.

Both low-level autopilot and fully autopilot are inseparable from the four technical systems of perception, planning, decision-making and control, and more than N technical routes have evolved in the first step of perception. From lidar to pure visual perception, and then to the retention of high-precision maps, the popularity of large models of BEV and AI has been debated both inside and outside the industry.

In order to promote, the car companies even did not hesitate to use slogans such as L2+, L2.9, L2.99999, but in fact, we are still at the L2 level, which can not be called the auxiliary driving of self-driving.

Waymo once claimed internally that 99 per cent of the technical problems of unmanned cars have been overcome, but more people care about the 1 per cent you have not yet reached.

The Cruise accident is a good example. Even if you conquer 99 per cent of the technology, the technology that is not realized when the accident occurs is a crisis of public trust in the business.

Once there is a crisis, it will have an irreversible impact, and turning may really be the best choice.

Of course, it is not without good news, although the industry leaders are more and more uncertain about the landing time, but in terms of policy support, there are already relevant guidance documents.

On November 17, the four departments officially issued a notice on the pilot work of intelligent network access and road access.

It is clearly stated that intelligent Internet connected automobile products with self-driving function with mass production conditions can be selected to carry out access pilot.

This means that L3 and L4 self-driving cars have official policy support and can be legally on the road on pilot roads.

Not only that, in this notice, the division of responsibility is made clear for the first time, with regard to the handling of accidents on pilot vehicles, and the attribution of responsibility for traffic accidents is determined according to different circumstances.

When the vehicle has an accident when the autopilot system is not activated, it shall be held responsible according to the current regulations. If a violation or accident occurs when the autopilot system is activated, the pilot enterprise and the pilot user shall provide proof materials to the relevant departments within a specified period of time. If the materials are not provided in accordance with the regulations, they shall be held responsible for the accident.

In addition, the intelligent network car can be used as the object of liability judgment in the accident, and the pilot users should bear the liability for the determination of relevant liability in accordance with the law. If there is an accident fault, the pilot user may recover compensation in accordance with the law. If it constitutes a crime, the relevant responsible persons shall be investigated for criminal responsibility according to law.

For the first time, clear policy support should set up a stove for the self-driving industry in the cold winter.

This article is from the official account of Wechat: Chaodian Lab (ID:SuperEV-Lab), author: Wang Lei Chumen

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.