Various enterprises compete to invest, ushering in new opportunities for the automotive industry
2024-05-15
Driving on urban elevated roads or intercity highways, your hands can completely release the steering wheel, and the cameras and sensors outside the car monitor the road surface in real time: even on bends, uphill and downhill slopes, and at the entrance and exit of ramps, the vehicle can drive smoothly in the middle of the lane; Whether accelerating or decelerating or braking suddenly, it can closely follow the preceding vehicle at a specific safe distance; In special situations such as lack of concentration, unclear road markings, and no GPS signal, the driver attention retention system will use multi-level warnings to remind the driver to take over vehicle control and ensure safety.
At the 2018 Asian Consumer Electronics Show, SAIC General Motors launched the Cadillac CT6 equipped with an intelligent driving system. SAIC General Motors defines this intelligent driving technology, which is mass-produced in the industry and can truly release both hands on highways, as L2 (partially automated).
Unlike the cautious approach of traditional car companies, Google's Waymo, shared car rental companies Uber, Lyft, and many autonomous driving startups have directly targeted L4 (highly automated). On July 4, at the Baidu AI (Artificial Intelligence) Conference, Robin Lee, chairman and executive officer of Baidu, announced that the global L4 class automatic driving bus Apollon was mass produced.
How far are we from true autonomous driving? What opportunities and challenges do Chinese enterprises face in the pursuit of technological innovation?
The automatic driving technology routes chosen by traditional car enterprises and Internet enterprises are different
On July 6th, Daimler Group announced that it has become an international automobile manufacturer that has obtained a road test license for autonomous vehicles in Beijing. Previously, SAIC, Chang'an, BAIC, GAC, Weilai and other automobile enterprises, as well as Baidu, Tencent and other Internet enterprises, have successively obtained road test licenses for autonomous vehicles in Shanghai, Beijing, Chongqing and other places. Changan, Baidu, as well as electric vehicle startups such as Xiaopeng Motors and SF Motors, have also obtained testing licenses in California, USA.
However, it is also an automated driving test, but the content of the test for car companies and Internet companies is quite different. Changan Automobile, which has taken the lead in mass production of L2 technology this year, is conducting L3 level autonomous driving tests in California and Michigan, USA, with a cumulative test mileage of 1 million kilometers. Xiaopeng Motors has chosen the super automatic parking system as the entry point for autonomous driving, and is currently conducting optimization tests on scene coverage and user experience for the mass-produced model G3. Waymo's test fleet has traveled over 8 million kilometers in autonomous mode in 25 cities in the United States. Since May 14 this year, three L4 autonomous vehicle on Baidu Apollon's open platform have also started to conduct real road tests in Xiong'an Citizen Service Center Park for several days.
According to industry classification, autonomous driving is divided into five levels: driving assistance, partial automation, conditional automation, highly automated, and fully automated, commonly known as L1 to L5 in the industry. Traditional car companies tend to choose the technology route of iterative driving assistance sequence, and are currently in a critical stage of transitioning from L2 to L3. Most Internet enterprises choose to directly develop L4 automatic driving technology and set up automatic driving teams for road test.
"When you see a company claiming to be conducting L4 level testing or having already achieved L4 level mass production, you first need to look at what scenario it is based on." Xu Wei, Vice President of Zhejiang Zero Run Technology Co., Ltd., said that achieving highly autonomous driving on park roads is only limited to L4 scenarios. To truly achieve a full scenario L4, not only high-precision maps are required, but also V2X (information exchange between vehicles and the outside world), which cannot be achieved without the construction of smart cities and the transformation of infrastructure. It is difficult to achieve it within 10 years. Therefore, the general consensus in the industry is that L3 will be mass-produced in 2020, and in semi closed or limited scenarios, L4 level autonomous driving will appear at the Tokyo Olympics; Around 2030, the full scene L4 is expected to land.
Of course, the seemingly radical technological route of Internet enterprises also has its rationality. The direct development of L4 not only circumvents the high-tech threshold set up by the automobile industry for a century, but also avoids the difficulty of human-computer driving that L3 is difficult to break through. In addition, by accumulating as many field test miles as possible, conducting simulation induction tests in various critical scenarios, continuously accumulating data, and improving the system's in-depth learning ability, it is precisely the strength of Internet enterprises.
However, before the red line of safety, problems still arise repeatedly. In March this year, a autonomous vehicle in Uber crashed into a passer-by, which directly led to the suspension of its road test permit in Arizona. As for Tesla's driver fatalities caused by the Autopilot assisted driving system, there have been more than one incidents.
"Safety is crucial in the development of autonomous driving," said Cadillac CT6 engineer Li Linden. To ensure safety, sufficient redundant design is essential. It is understood that the autonomous driving vehicle being tested by General Motors is equipped with 5 LiDARs, 16 cameras, and 21 millimeter wave radars. In addition, it is also equipped with two sets of computing systems that work simultaneously, two sets of power supply systems converted from high-voltage batteries, a signal transmission system that uses additional redundant paths, and a braking system that uses two sets of execution methods
"The original intention of research and development is to bring safety, efficiency, and convenience to users through autonomous driving." He Jugang, the director of Changan Automobile Intelligence Research Institute, believes that the maturity of autonomous driving technology software and hardware takes time, and many new technology costs urgently need to be reduced. Progressing from L1 to L4 step by step is more in line with the laws of industrial development. However, at the same time, autonomous driving has a long industry chain, and there are many technological breakthroughs that need to be made. Although there are many investors and different technological routes at present, after the big waves of sand washing, the winner will form their own core capabilities, and the industry division of labor will be clearer.
China's autonomous driving can only be solved by the Chinese themselves
——At the end of 2018, Waymo Autonomous Vehicle Company plans to launch autonomous taxi services in Arizona, USA;
——In 2019, the autonomous vehicle developed by the Cruise team under GM will be commercially deployed in the United States on a large scale;
——In 2020, the new Mercedes Benz S-Class sedan will be equipped with L3 level autonomous driving technology;
——In 2020, Changan Automobile's L3 level autonomous driving vehicle will achieve mass production;
——In 2020, Xiaopeng Motors launched mass production of L3 level autonomous driving;
——In 2020, Toyota autonomous vehicle will shuttle between venues of the Tokyo Olympic Games;
——In 2021, Volvo will mass produce L4 level autonomous vehicles
According to the Brookings Institution, the total investment in the field of autonomous driving worldwide exceeded 80 billion US dollars from 2014 to 2017. According to statistics from another independent think tank, the funds invested in the field of autonomous driving technology in 2017 accounted for more than 70% of the total investment in the global automotive technology industry.
Various enterprises are vying to invest in autonomous driving, naturally eyeing its huge market potential. A recent report released by Intel and a research company predicted that the global market size of autonomous vehicle would reach US $7 trillion by 2050. Boston Consulting said that by 2035, the global sales of autonomous vehicle will reach 12 million, more than a quarter of which will be sold in China.
Can the huge market dividend of autonomous vehicle turn into an opportunity for China's automobile industry to come from behind?
"Compared to Google, which started developing autonomous driving around 2005, China started relatively late and has less accumulation in data, algorithms, talent, and methodology. However, I believe that autonomous driving in China can only be solved by Chinese people themselves." Dr. Junli Kugu, Vice President of Autonomous Driving at Xiaopeng Motors, said that unlike other technologies, the internationalization of autonomous driving has many limitations.
Gu Junli explained that due to the high traffic density in China, the mixing of pedestrians, non motorized vehicles, and cars, as well as the presence of numerous obstacles and the high degree of randomness in traffic behavior, the existing autonomous driving algorithms in the West almost need to be rewritten. In order to build an autonomous driving solution that adapts to Chinese regulations and traffic scenarios, Xiaopeng Motors is building a vertical system that includes big data extraction for driving, scene drawing, algorithm development, verification and evaluation, and algorithm iteration evolution. Firstly, deep customization of vehicle sensors is required - through multi-dimensional cameras, multi-dimensional ranging radars, ultrasound, and other sensors, heterogeneous multi-sensor fusion is carried out to achieve multi scene, full range, and dead angle coverage. "On the upcoming Xiaopeng G3, we are equipped with nearly 20 sensors, and the vehicle currently under development with L3 level autonomous driving function is equipped with as many as 30 sensors," said Gu Junli.
However, despite having a "home advantage", facing a huge market "cake", whether Chinese enterprises can enter the industrial chain and enjoy technological discourse and rich industrial profits depends on whether they can achieve breakthroughs in core technologies such as sensors, artificial intelligence chips, high-precision maps, and algorithms.
"International mainstream chip manufacturers such as NVIDIA, NXP, TI, and Mobileye are all vigorously developing artificial intelligence chips. However, in the field of autonomous driving, apart from Mobileye, there is no mature solution, which is a good opportunity for domestic integrated circuit design manufacturers." Xu Wei said, this is why Zero Run not only makes its own sensors, but also self-developed chips. With the deep accumulation of Dahua Technology in the field of security cameras, the cost of self-developed chips is only half of that of external procurement. More importantly, the efficiency will be higher when the algorithm architecture is integrated and optimized by the same internal team.
In the view of Zhang Yongqian, General Manager of Horizon's Intelligent Solutions and Chip Division, due to the fragmented demand for artificial intelligence, it is almost impossible for traditional industry giants to use scale, funding, channel capabilities, and even crowd tactics to defeat startups. This also provides opportunities for Chinese artificial intelligence chip design companies like Horizon to grow and develop.
"In the fields of LiDAR, high-precision mapping, and artificial intelligence chips, Changan has cooperation with domestic and foreign suppliers." He Jugang said that China's industrial advantage is having a full industry chain from raw materials to primary and secondary suppliers, then to host factories, and finally to travel service providers. Although the technical threshold is high, domestic enterprises in this chain are very competitive, catching up quickly, and the gap is constantly narrowing.
"However, domestic enterprises often do not do enough in the fields of basic science and multi-dimensional technology that need to be deeply explored." Gu Junli said that in the trend of autonomous driving, startups and capital companies can only abandon the old tricks of hyping up concepts, attracting attention, and telling stories, sink down, make up for the missed lessons, and fully explore the underlying technology that needs to be deeply explored, in order to truly seize the huge business opportunities.
