Shulou(Shulou.com)11/24 Report--

In 2016, self-driving startup Argo AI made its debut, catching up with the east wind of self-driving, and Argo AI was valued at more than $7 billion at one point. To everyone's surprise, however, Argo AI announced its dissolution in October this year.

After 13 years of research and development of self-driving technology, Waymo,2018, known as the "self-driving industry leader", was valued by Morgan Stanley at $175 billion, surpassing the market capitalization of all car companies except Tesla and Toyota. After several rounds of financing, it was valued by the Financial Times at just $30 billion, a full 80 per cent decline.

Aurora, which claims to share the world with Waymo and Cruise, has adopted layoffs, salary cuts, asset sales and other ways to cut expenses. It is clear that the research and development and testing of autopilot technology is still in full swing in 2022, but these enterprises are all struggling. What happened to the autopilot industry?

01. There has always been no return on investment, and it is difficult for autopilot companies to look back on 2022. The development and popularization of autopilot technology has never stopped, especially in China, where various policies to promote the popularization of autopilot technology come one after another.

The Automation Classification of Automobile driving and the Digital economy Development Plan of the 14th five-year Plan issued by the Ministry of Industry and Information Technology two years ago have laid a good foundation for self-driving. In June this year, Shenzhen issued China's first intelligent Internet car management regulation, Shenzhen Special Economic Zone Intelligent Network United vehicle Management regulations, which laid a sound legal foundation for the development and commercial use of self-driving technology.

At the same time, a number of domestic cities such as Guangzhou, Beijing, Chongqing, Wuhan, Changsha and Shenzhen have issued policies to allow self-driving cars to be put into trial operation on the road in specific areas and specific periods of time. Compared with the laboratory, road testing is undoubtedly closer to the real road conditions, and self-driving companies can also collect more reliable data.

Touyuan: Lei technology photography seems to be developing in a good direction, but self-driving companies are facing more and more serious problems, that is, they have been unable to make a profit for a long time, and their R & D investment is like a bottomless pit.

According to Zhiyan, it is estimated that by 2025, the global penetration rate of L2-level assisted driving smart cars will reach 53.99%, but the penetration rate of L3~L5-class self-driving cars will only reach 1.36%, and annual sales are expected to reach 1.19 million.

Waymo employees have said the company spends more than $1 billion a year to develop autopilot technology. Obviously, the penetration rate of 1.36% will not be achieved until 2025, and it is impossible for self-driving technology to pay back, let alone make a profit in the short term.

In the face of this situation, it is normal for investors to be discouraged. In fact, there is another alternative for autopilot companies, which is to devolve some of the L3~L5-level autopilot technology and first cooperate with car companies to increase revenue capacity. In fact, car companies are developing L2-level auxiliary driving, while self-driving companies tend to directly develop L3-level self-driving technology.

Photo Source: pony Zhixing and even ambitious Apple will directly develop L5 full autopilot as soon as they appear on stage, but such a high level of autopilot technology can not be realized at all for the time being. As a result, it has recently been revealed that Apple has decided to give up L5 fully self-driving in order to promote the launch of electric vehicles as soon as possible.

It can be predicted that L2-level auxiliary driving will become the mainstream in the next 10 years, L3~L4-level autopilot will be popularized at a relatively slow speed, and L5-level autopilot will probably not be realized within 10 years. Self-driving companies blindly pursue cutting-edge technology research and development, but ignore the short-term balance of payments. Investors' money is not blown by strong winds. Seeing the financial reports of long-term losses of these companies, they will naturally reduce or stop investing.

2022 is a pragmatic year for self-driving enterprises, successively abandoning the unreachable L4~L5-level self-driving and returning to L2-level auxiliary driving, which has a real impact on the user experience. Self-driving enterprises no longer have high expectations and low skills, and begin to comply with the actual needs of users, so that the technology can be put into commercial use as soon as possible.

If the technology can not be commercially available, it is inevitable that the income will not be able to make ends meet, and the financial report will lose money for a long time. What's more, even with the above use of their autopilot technology, it is difficult to obtain the approval of consumers before solving the accident rate.

02. The continuous autopilot of accidents cannot convince consumers that even if L3 and above are autopilot, there is no need for a driver at all if they reach level L5. At present, some countries and regions have allowed L3 and L4 self-driving cars to operate on the road, but drivers are still required to pay full attention when using autopilot technology to prevent traffic accidents caused by system problems.

However, since the commercial use of autopilot, there have been a number of accidents caused by autopilot in many parts of the world. On March 18, 2018, a Uber self-driving taxi in the United States knocked down a woman who was taken to hospital and died, making it the first self-driving fatal accident in the world.

According to NTSB survey data, there were a total of 37 accidents in Uber's self-driving system in the 18 months before the accident. Not only Uber, but also self-driving cars of companies such as Tesla and Waymo have had traffic accidents, which have led some investors and consumers to lose confidence in self-driving technology.

The more serious question is, whose fault is it when there is an accident in a self-driving car? The regulations on the Administration of cars in Shenzhen Special Economic Zone clearly stipulates that if there is an accident in a L3~L4-class self-driving car, the responsibility still lies with the driver. Only when an accident occurs in a L5-class car, the responsibility will be attributed to the car company and the self-driving technology company.

Picture source: pixabayL3 and L4 belong to conditional self-driving technology. Enterprises will require users to still concentrate on preventing accidents when using them, but they all buy cars that support self-driving, even pay extra for self-driving technology, and still have to bear invisible traffic pressure. As a result, they still have to be held responsible for accidents, so what is the use of L3 and L4 self-driving?

In short, the current commercial autopilot technologies have limitations. Autopilot companies do not like L2 and below auxiliary driving technology, while L3 and L4 autopilot technology is difficult to convince consumers until the accident is completely solved. A higher level of L5 autopilot is almost impossible in the short term.

That is why, after years of losses and disappointment, investors abandon these self-driving companies, so that their valuations and market capitalizations are falling. But it is self-driving companies that are falling, which is not necessarily a bad thing for the industry as a whole.

03. What punctures the bubble is that the foundation is the same as that of other industries. At the beginning of its birth, autopilot technology belonged to the blue ocean market and flooded into numerous enterprises, but an industry moved from green to mature. There are bound to be many enterprises falling down in the process of progress.

In 2022, many self-driving companies encountered problems, and even Argo AI was forced to dissolve, which is only a hurdle that the development of self-driving industry must overcome. Although the era of self-driving is coming, the market capitalization of any technology research and development company can surpass that of so many established car companies, and it is obvious that the bubble has been blown so big that it is only normal to burst. Under the punctured bubble, we see that the foundation of the whole industry has been consolidated.

In 2022, autopilot technology laid a solid foundation. First of all, the major companies found their own areas, subdivided the application scene, and achieved specialization and refinement.

Self-driving technology is mainly divided into three application scenarios, namely, special work vehicles, large trucks, and passenger vehicles. Commercial and special-purpose models mainly include unmanned express vehicles, mine unmanned vehicles, etc., on behalf of the enterprise Xidi Zhizhi, the company's unmanned mine cars have been put into use in Inner Mongolia. With its excellent technical strength, the company recently won the second prize and popularity award in the second Intelligent Manufacturing Innovation Competition.

The think-tank predicts that the scale of self-driving technology transportation services in Chinese mining areas will increase to 391.2 billion yuan by 2030. It may not match the size of the passenger car market, but it is also a piece of fat.

Large trucks, mainly unmanned trucks, are also the main research and development direction of Xiaoma Zhixing, a famous self-driving technology company in China. In early November this year, Horse Zhixing and Sany heavy card joint venture company Yiji smart card the first batch of vehicles offline delivery. According to the Beijing Business Daily, the market for self-driving heavy trucks will reach 1 trillion yuan in 2030, and the market is huge enough to support many self-driving companies.

Picture source: compared with unmanned work vehicles and unmanned trucks, ordinary consumers are naturally more concerned about self-driving passenger cars. For the safety of passengers, domestic self-driving cars above L3 are not allowed for commercial use, but some areas of some cities allow self-driving taxis on the road. For example, residents of Huangpu District of Guangzhou can use App to book Baidu Apollo self-driving taxis.

The rules vary in different cities. For example, the main driver of a self-driving taxi in Guangzhou still needs a safety guard to take over the car in case of emergency. For self-driving taxis in Chongqing and Wuhan, the security officer sits in the copilot. Although self-driving taxis are only tried out in a small number of areas, there is no doubt that they have a comprehensive commercial basis.

With the gradual maturity of technology, self-driving taxis will get rid of the safety staff in the future and truly realize self-driving. As for the fact that without a security guard, the car encounters complex road conditions, such as traffic accidents ahead and it is difficult to drive on the road, the car will be taken over by a remote "online driver" to remotely control the car through the low-delay characteristics of the 5G network.

Photo Source: although Lei Technology has not yet allowed L3 and above self-driving vehicles for commercial use in China, and only a few areas can be tested on the road, domestic car companies and self-driving technology have never forgotten the research and development and promotion of self-driving technology. Take Wei Xiaoli as an example, the biggest move in the field of autopilot in 2022 was the unveiling of the autopilot research and development platform (NADP).

Wei Lai said that each car will generate 55PB data (1PB is equivalent to 1048576GB) every day, so it is necessary to analyze and extract valid data from huge amounts of data and apply it to the upgrade of self-driving technology. The role of NADP is to provide one-stop platform management to improve the development efficiency of various travel scenario models by 20 times, which will greatly shorten the research and development cycle of automatic driving technology.

Xiaopeng's focus in 2022 is mainly on high-precision maps and urban NGP. At present, the scope of application of urban NGP is auxiliary driving, not autopilot. The principle of this scheme is to draw a more accurate map model with the help of high-precision map, so as to achieve more perfect assistant driving. The technology can also be applied to autopilot, but requires higher-precision maps.

Ideal didn't explain too much technology this year, but it reached a cooperation with NVIDIA and Desai Xiwei. NVIDIA provides it with a chip up to Orin, based on which Desai Xiwei produces autopilot domain controllers for ideal cars and ideally develops its own autopilot algorithms. It is ideally expected that the L5 self-driving car will be realized by 2025. However, in order to achieve the large-scale popularity of autopilot, there is another problem that needs to be solved, and that is the cost of hardware.

The second point is to consolidate the foundation in 2022, the supply chain is becoming more and more mature, and lidar and vehicle chips have stronger performance, higher reliability and lower price.

There are mainly two mainstream schemes for self-driving environment data collection. 2022 is the year when car companies have decided to choose the scheme one after another. For example, Tesla, a global leader in new energy vehicle companies, chose a relatively low-cost pure vision scheme to completely cut off the radar. rely on the camera to collect data, the algorithm to calculate and draw three-dimensional scene.

Most domestic automobile enterprises choose the fusion vision scheme, that is, radar + camera to collect data, and radar is divided into lidar, ultrasonic radar, millimeter wave radar and other types, among which lidar can collect a wide range of three-dimensional information. it is the key to realize self-driving in the future. But the lidar used to rely on imports, the cost and price are more expensive.

In 2022, it is known as the collective rise of domestic lidar enterprises and the desperate year for overseas manufacturers. A few days ago, domestic Hesai Science and Technology announced that the monthly delivery of lidar exceeded 10000 for the first time, and the lidar produced by Suteng Juchuang has also been used in a number of high-end domestic cars. At the same time, domestic lidar enterprises have been trying to find ways to develop lidars that are cheaper and more stable. Last year, Hesai Technology and Suiteng Juchuang launched a semi-solid lidar to replace the original mechanical lidar.

Image source: some time before Lei Technology filmed, Wesai Technology released an all-solid-state lidar, which can not be used as the main lidar, but can be used as an auxiliary lidar, and the affordable price makes it more likely to be popular. In 2018, a 64-line Velodyne lidar sold for as much as $80,000 (about 560000 yuan). Today, a semi-solid lidar costs only about 10,000 yuan.

Hosai Technology has said that the price of lidar will fall to less than $1000 (about RMB 7000) in the future, while Huawei, a rising star in the automotive industry, said that the goal is to bring the price of lidar down to less than $200 (RMB 1400). The fierce competition from domestic manufacturers has greatly accelerated the development of lidar.

Domestic lidar companies applauded, but overseas manufacturers were gloomy. In September, Ibeo, the originator of lidar, filed for bankruptcy. In November, two lidar companies, Velodyne and Ouster, announced their merger. Not long after, Quanergy lidar company was forced to withdraw from the market. Generally speaking, the lidar industry is still developing in a better direction, but domestic enterprises are gradually replacing overseas enterprises.

Another important piece of hardware in self-driving technology is the chip, with Mobileye and Nvidia in a leading position. Mobileye's main market is L2 auxiliary driving, while NVIDIA's is more advantageous in L4 self-driving. Some time ago, NVIDIA launched a Thor with computing power as high as 2000TFOPS, setting a new computing record for on-board chips.

The third point of tamping the foundation in 2022 is that high-precision maps cooperate with vehicles and roads, which is not the key to achieve high-level self-driving.

In the face of complex road conditions, the car's own radar and cameras may not be able to collect all environmental data, especially in situations such as ghost probes, where radar and cameras can not bypass obstacles to detect objects behind. In order to solve this situation, vehicle-road coordination arises at the historic moment.

Self-driving technology companies will install monitoring equipment in areas with complex traffic environment and send the collected data to vehicles located in that area through a high-speed, low-delay 5G network. This function will also be assisted by high-precision maps, which can provide centimeter-level road data for cars and reduce the pressure on cars to draw scene models.

Domestic car-road coordination and high-precision maps are mainly done by Baidu Apollo. The test results in Huangpu District of Guangzhou are good, but they can only be used as a supplement to autopilot, not as a major reliance. Because of the limited cost, it is difficult to install too much monitoring equipment, so it can only be installed at the intersection where the traffic environment is more complex.

Although the bubble of the self-driving industry was punctured in 2022, it is a good thing for the whole industry. Whether it is the software development of autopilot technology or the upgrading of hardware technology, it is progressing steadily in 2022, and the application market division of enterprises is more obvious, all moving in the established direction.

04. In the next year of self-driving, the key to creating a commercial space is the same as that of the new energy vehicle industry, and the development of self-driving technology will also go through the end of the era from bubble blindness to pragmatism and drawing a big pie. It is inevitable that many enterprises will fail to keep up with the trend.

In October 2022, China took the lead in the release of the first international standard in the field of self-driving test scenarios, which also shows that China's self-driving technology has taken a key step towards commercial use. In the face of fierce competition, if self-driving companies want to survive, they must understand one thing, that is, how to create commercial space. Only by making up for R & D spending through revenue can we avoid being abandoned by investors.

Obviously, L3 and above autopilot technology is temporarily unavailable for commercial use, so autopilot companies must keep a low profile and apply their own technology to L2 auxiliary driving to gain stronger revenue capability. avoid persistent negative profits.

Life is relatively easy for hardware manufacturers. After all, advanced self-driving is not commercially available, but car companies must reserve hardware that can support self-driving, so lidar and high-performance chips are not afraid of no sales. Under the competition of domestic lidar enterprises, the price of lidar is declining rapidly. in the future, more than one hundred thousand cars may be equipped with lidar as standard.

The construction of vehicle-road coordination and high-precision maps is also continuing, laying a solid foundation for the future self-driving comprehensive commercial use. At present, vehicle-road coordination and high-precision map can only be used to assist driving, achieve lane-level navigation, constantly collect data and improve themselves.

As for the date of the popularity of self-driving, it can only be said that it depends on whether the development of car companies and self-driving companies can reduce the accident rate to a tolerable range in the coming period of time to ensure the safety of passengers. Until safety can not be completely solved, autopilot can not be widely used.

This article comes from the official account of Wechat: Feng Xingxing (ID:fengchuxing2021), author: lost Soul.

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.