Autonomous driving technology has never been closer to us than it is now – just as the information disclosed at the China Electric Vehicle 100 Forum at the beginning of this year, the “new four modernizations” of automobiles with the trend of electrification, intelligence, networking and sharing , is giving birth to a “revolution on the road”.
This will bring greater imagination to the automotive-related industry. According to PricewaterhouseCoopers, the development of smart cars will also drive the automotive-based chip, software, information communication, and data service industries to become new economic growth points.
As an important part of automobile intelligence, autonomous driving technology and a series of products and services derived from this technology have been widely valued by industry participants because of the “rich ore” contained in it.
Self-driving cars will create a 2 trillion market in China by 2040
A few days ago, McKinsey & Company of the United States released a report saying that around 2040, autonomous vehicles will account for 66% of the total passenger mileage (PKMT) in China. At this rate, self-driving cars will generate a local market of up to $2 trillion in annual revenue, of which about $1.1 trillion will come from mobility services and $0.9 trillion will come from self-driving car sales. Meanwhile, autonomous vehicles will account for nearly 40 percent of new car sales and about 12 percent of total car ownership.
At the commercial application level of autonomous driving technology, the travel industry will obviously play a leading role.
“This is because cars equipped with such technologies will bring higher operational capabilities (closer to 24/7 operations) and lower labor costs (no driver required),” said Lu Shuai, an automotive analyst at The Beatles Consulting. .
For the same reason, the adoption rate of autonomous driving technology in buses and commercial vehicles is also expected to reach 69% and 67%, respectively.
Lu Shuai said that self-driving cars may change the value dependencies of the current travel market, shifting related value from products to services. In other words, after the widespread application of autonomous driving technology, people’s consumption habits in travel will change from owning a car to paying by mileage.
Today, similar judgments can be found in the Chinese market. “In the current market environment, users’ travel consumption behavior has shown a trend of ‘from buying a complete vehicle to buying by mileage, number of trips and usage scenarios’.” Lu Bin, co-founder and senior vice president of WM Motor When talking about the reasons for the company to take the lead in the field of smart travel, he said recently.
The above-mentioned emerging concept of “Mobility as a Service” (MaaS) will undoubtedly bring earth-shaking changes to the product sales and business models of automobile manufacturers.
In China, fully autonomous driving technology (level 4 or above according to the standards of the American Society of Automotive Engineers) is expected to be fully rolled out within 9-10 years. This time span is close to the average development level of the world’s major automotive markets in the above areas.
“For the current industry players, this period of time for transition is not long, so they must try to integrate into this new business environment as soon as possible.” Independent automotive analyst at the University of Engineering and Technology (UTT) in Troyes, France Liu Rui made the above judgment.
In the entire travel system, the relevant rules of the game will change accordingly. Software and data are gradually replacing hardware, and in the process of manufacturing and operating cars, there is a clear distinction between the good and the bad.
Under such circumstances, a huge industry that brings together automobile manufacturing, passenger and freight transportation, software, hardware and data services seems to have emerged from a genesis explosion.
Perhaps for now, most automakers are still focusing on new car sales, while transportation companies are busy providing related services, and technology companies are silently delivering all kinds of hardware and software to automakers.
In the future, new business models will emerge one after another, thereby changing the value flow of this behemoth industry.
“Technology companies may buy cars from automakers and then provide services directly to end consumers.” Liu Rui analyzed, “Or, automakers may vertically integrate services and software development into the enterprise. This is like At present, several leading companies in the automotive industry are already doing that.”
In 2016, General Motors spent $58.1 billion to buy self-driving car startup Cruise Automation and grow it into a separate business unit. In 2017, Ford announced a $1 billion investment in Argo AI, which focuses on artificial intelligence development, and has since brought it under its umbrella as a self-driving subsidiary. In April of this year, Daimler Trucks announced the acquisition of a majority stake in Torc RoboTIcs, a developer of autonomous truck driving systems, to further advance the application of this technology in the commercial vehicle sector.
Reverse integration also occurs in this area. Apple Inc. confirmed on Wednesday (June 26) that it has acquired self-driving car startup Drive.ai and hired dozens of its engineers to develop related technologies. Uber said on the same day that it had acquired computer vision startup Mighty AI to help advance its self-driving car technology.
“Autonomous driving technology is like a hot soldering iron, melting the seemingly unbreakable barriers between automobile manufacturing, mobility and IT industries, and accelerating their integration.” Lu Shuai commented, “But it also broke the 100-year-old barrier. There are so many rules coming to the auto industry, and automakers have to be well prepared for these changes.”
He believes that in this industry sector, players from different fields must try to dissolve differences in product life cycles and business models in order to eventually form an efficient competition and cooperation relationship.
For example, the life cycle of automotive products may be as long as 4 years or even longer, while for a set of software, the frequency of possible updates can only be calculated in weeks or months. Likewise, the emphasis on product or service should be gradually neutralized and eventually balanced in such relationships.
There is no doubt that many companies have already begun to implement their development strategies for self-driving vehicles, including technology companies such as Baidu, Tencent, Waymo (a subsidiary of Google’s parent company Alphabet that focuses on the development of self-driving technology) and companies such as General Motors, SAIC (25.770, 0.23, 0.90%), the automaker of Tesla.
“But given the dynamic and rapidly changing nature of this industry, such industry players must constantly adjust and update their existing strategies based on actual conditions,” said Lu Shuai.
A survey conducted in May by France’s Capgemini Research InsTItute based on more than 5,000 samples showed that Chinese consumers are more positive about self-driving cars than most countries. In contrast to Chinese consumers, respondents in the U.S. and U.K. were the least interested in self-driving technology, with just over a third of respondents optimistic about self-driving cars.
“Among the global respondents surveyed, 53% of Chinese respondents said that self-driving cars will be their preferred mode of transportation in the next five years. Pay a premium of up to 20%,” researchers at Capgemini Research wrote in the report.
Through a cost cross-analysis, McKinsey believes that autonomous driving technology will reach a mature state around 2023, and its application cost will drop to about $8,000 by around 2025, which also means that the application speed of this technology will be greatly improved by then. .
The “unicorns” behind autonomous driving
However, considering the complex traffic environment in China, the widespread adoption of autonomous vehicles will still face some problems in the short term. For example, such vehicles must first adapt to different road conditions and the relatively intense driving habits of other traffic participants.
However, the key to solving these problems does not lie in vehicle companies. Unicorn companies in the autonomous driving industry chain are making rapid progress in this field.
“So from a technical point of view, there is no fundamental difference between China’s self-driving promotion and other countries in terms of methods,” said Fabrice Provot, who is currently in charge of a software company in Shanghai. Algorithm development related to autonomous driving.
Pouwo said that the computing platforms used to calculate the traffic environment in New York and Beijing are basically the same. “Because existing platforms have enough buffer capacity to handle computational tasks that are more complex than analyzing China’s road conditions,” he added.
Jeffrey J. Owens, chief technology officer and global executive vice president of ApTIv PLC (formerly Delphi), a provider of future mobility development technologies and solutions, introduced to reporters in an interview with Jiemian News earlier. A multi-domain controller developed by the company is similar in shape to a small chassis, but functionally covers the control and information exchange of key components such as radar, cameras, airbags, crash sensors and detection systems.
Currently, in the electrical and Electronic architecture of the car, data is being processed at a speed of about 65M per second, while the multi-domain controller can process 15G of information per second.
“There may be as many as 50-55 pieces of microcontrollers in a self-driving car, but with high-performance processors and large-scale software integration, it is ultimately possible to make information processing more efficient and enable near real-time data interaction, potentially in In the event of a collision, near real-time data exchange is very important,” Owens explained.
As early as 2016, an Audi SQ5 equipped with Aptiv’s autonomous driving solution completed a public road test across the United States without prior route mapping.
On the other hand, the configuration of the sensor does not need to change with the change of the region, because the current sensor can cover all key directions in different driving scenarios.
“The most special thing about China is that its road environment is filled with many highly complex traffic signs. In different regions, traffic lights and signs are sometimes not fully standardized.” Fabrice Pouwo said, “In addition, some Traffic participants do not strictly follow traffic laws, which makes programming self-driving cars more difficult.”
Because of these problems, it takes more effort to optimize decision-making algorithms for autonomous vehicles for Chinese road conditions, Puwo said. Compared with the United States, its development may take about 2-3 years longer.
In view of this, industry insiders generally believe that China’s first self-driving cars will be the first to be applied in specific environments within the next five years, but large-scale applications will not be earlier than 2027. During this period, autonomous driving technology must first adapt to various traffic environments, including urban and rural areas.
“Of course, the core algorithms of self-driving cars do not vary by country. It’s just that China’s complex road conditions mean that self-driving technology needs to go through more tests and obtain more data before it is applied.” Puwo explained.
For example, developers need to collect and input local traffic data to solve the problem of special and inconsistent traffic signs. They must also optimize the motion planning of the vehicle as it moves through the test road, so that the self-driving vehicle can be algorithmically taught how to deal with reckless traffic participants.
“To truly seize these opportunities, industry players need to establish links with end consumers through differentiated services, or master the core components of autonomous driving technology.” Lu Shuai analyzed, “So in this technology, distinguishing Which elements have long-term strategic value, and the latter can be improved over time, will help industry players firmly hold the soul of self-driving cars.”
The burgeoning self-driving car ecosystem differs from conventional car-based mobility solutions because it focuses more on the development of a “technology stack”—a kind of high-tech And the concept that is frequently used in the computer industry, refers to the combination of technologies that use multiple technologies as an organic whole to achieve a certain purpose.
In the technology stack of autonomous driving, the core elements usually include sensors, computing platforms, software algorithms (object detection and analysis, motion planning, etc.), system integration and verification, maps, location-based services (LBS), etc.
In a car, the combination of these technologies will form the core of an autonomous driving system. For automakers, parts suppliers, tech companies and other industry players, they are all key components in creating the self-driving technology itself.
“There is no doubt that in the technology stack of autonomous driving, important elements will continue to be iteratively upgraded over time,” said Puwo. The negative factors of innovation are monitored and assessed in real time, and a quick response is made when necessary.”
At different technology levels, each technology stack has its own unique prerequisites for success, said Puwo. For example, sensors require excellent reliability and safety, and at the same time meet the needs of large-scale mass production, in order to meet the necessary economic requirements through economies of scale.
In the field of software and algorithms, companies must have flexible development skills and the ability to iterate quickly to improve algorithm performance by simulating data on a constant basis. System integration and verification also place extremely high requirements on reliability, safety and mass production capability.
Chinese figure among unicorns
One of the most critical questions for players in the autonomous driving industry right now is how different China’s technology stack is from the rest of the world.
“At present, the final form of this technology stack is still in a highly uncertain state, and the competitiveness of different industry players and the legal and regulatory environment will have a decisive impact on its form development.” Lu Shuai analyzed.
However, many industry players have found through early tests that in China’s autonomous driving technology stack, local and global technology solutions are likely to coexist.
Taking high-definition maps as an example, this technology is considered to be one of the important foundations for realizing autonomous driving. “The industry threshold for high-precision map data collection is very high, requiring professional fleets to collect a large amount of data and requiring strong data processing capabilities, so smaller startups cannot participate in it. Usually, this industry will form a situation similar to an oligopoly.” Technology blogger Shan Yuxiang said.
Currently, in the field of high-definition mapping, mainstream international players include Google, HERE and the Dutch company TomTom. In China, the three technology giants “BAT” are almost “unifying the rivers and lakes”. Baidu, Alibaba’s AutoNavi Maps and Tencent’s NavInfo (16.280, -0.21, -1.27%) occupy absolute advantages in the above fields.
In a typical self-driving car technology stack, certain elements—such as high-definition maps, location-based services, data clouds, etc., are restricted by local regulators for safety reasons. In the industry, such restrictions are referred to as “reverse restriction policies.”
Affected by this, overseas companies cannot participate in the technical aspects of autonomous driving in China.
In China, HD maps are almost always provided by local industry players. The same is true for geolocation-based services and data clouds. However, five technologies, including mobile service interfaces, motion planning algorithms, Internet of Vehicles, central processing and graphics computing units, and sensors, are currently open to overseas industry participants.
Some technology companies with foreign backgrounds are now trying to circumvent the above problems through different technological paths. Nullmax, a technology company founded in Silicon Valley in the United States in 2016, is one such example.
In June this year, the technology company released an autonomous driving system solution called Nullmax Max, the biggest feature of which is that it can achieve L3-level autonomous driving without relying on high-precision maps and lidars.
“Our view is that the smarter the car, the less reliance on HD maps,” said Xu Lei, the company’s founder and CEO. In fact, the Nullmax Max system mainly uses sensors to perceive the external environment, and then uses deep learning to improve the intelligence of autonomous driving, thereby simulating an autonomous driving experience similar to human driving behavior based on a technology called “reference planning”.
The abandonment of lidar is due to the consideration of application cost. “Our set of sensors are all automotive-grade sensors, mainly cameras, and do not rely on expensive lidars. In terms of magnitude, a laser is much more expensive than all our sensors combined.” Xu Lei explained Dao said, “If autonomous driving becomes a technology that everyone can’t afford, it will be difficult to really create value for the public.”
A positive sign for China’s autonomous driving industry players, however, is the Chinese government’s strong support for improving the technology’s local R&D capabilities. The autonomous driving test areas established successively in Shanghai Jiading, Xiongan New Area and other places can be regarded as examples. Such support will undoubtedly play a positive role in nurturing the R&D capabilities of local industry players.
Today, many venture capital groups and mainstream Internet companies have a strong interest in the opportunities that China’s autonomous driving industry holds in catching up with the world.
Over the past five years, local auto and parts companies developing autonomous driving-related technologies have received a cumulative $7 billion in funding. Alibaba, Baidu and Tencent are all investing heavily in these areas and seeking cooperation.
For example, Baidu and Tencent have invested in NIO, while Alibaba has invested in Xpeng Motors. In addition, Alibaba has established strategic partnerships with SAIC, Tencent and many OEMs, while Baidu has relied on its Apollo program to gather many industry players in the autonomous driving value chain.
Abundant disposable funds and high support from the government will have a positive effect on China’s self-driving car industry. The superposition of the two may make the above-mentioned industries show a competitive picture at the fastest speed.
However, some analysts pointed out that China’s autonomous driving industry needs to integrate itself into the global ecosystem to avoid the problem of “reinventing the wheel”.
“For Chinese companies, this kind of integration is very important, because to achieve high-level autonomous driving, the technical requirements are basically very close, so it can be easily transferred between various markets and regions.” Liu Rui said.