Author | Yu Bin
Editor-in-charge | Tu Min
Exhibiting | CSDN（ID：CSDNnews）
Looking back at the past 2021, the impact of the new crown epidemic on the world is undoubtedly the main theme of the world, and the epidemic has amplified the dilemma of the "lack of core" in the automotive industry and brought great challenges to the intelligent car supply chain. As an industry investment outlet, what happened to autonomous driving in 2021? What will be the future of autonomous driving? What kind of pain points are you facing?
Based on personal review and reflection, the author briefly describes the following views and does not represent any business or group.
Set the rules of the game and the game "gunshots" officially begin!
On August 19, 2021, the "Automobile Driving Automation Classification" standard, which was jointly compiled by more than ten domestic and foreign automobile enterprises at home and abroad by the National Automotive Standardization Technical Committee and lasted for three years, was officially released, which clearly defines the six levels from emergency assistance, partial driving assistance, combined driving assistance, conditional automatic driving, highly automated driving to fully automated driving, and will be officially implemented on March 1, 2022.
The watershed between driver assistance and autonomous driving is still in L3, according to the definition in the table below, the standard requires that within the design operating conditions and in the case of activation, the level 3 automatic driving system can perform all dynamic driving tasks including lateral control, vertical control, target and event detection and response; when the design operating conditions are not met or the design operating conditions are about to be met, when the automatic driving system fails, when the user's takeover ability is about to fail to meet the requirements, the automatic driving system can issue a takeover request; after issuing a takeover request, Continue the dynamic driving task for a certain period of time so that the driver takes over, and if the driver does not respond, execute the risk mitigation strategy at the right time.
For the "Automotive Driving Automation Classification" standard, personal understanding can independently complete the horizontal control, vertical control, target and event detection and response and other dynamic driving tasks, but also does not meet the requirements of level 3 automatic driving, at least need to achieve a variety of monitoring functions, such as: DMS, steering wheel monitoring, environmental perception monitoring, etc., timely detection and identification of events that do not meet the design and operating conditions, active alarm, and reasonable disposal, in order to meet the definition of level 3 automatic driving.
Although the classification standards for autonomous driving have been issued, laws and regulations related to autonomous driving have not yet been established, and mass production models of autonomous driving (level 3 or above) cannot be put into operation on the road. While studying and establishing autonomous driving regulations, industry regulatory authorities encourage local governments to carry out demonstration operations of autonomous driving in limited areas, on the one hand, to verify and evaluate the level of autonomous driving technology, and on the other hand, to explore the experience of autonomous vehicle operation.
On October 15, 2021, the Beijing Intelligent Connected Vehicle Policy Pilot Zone officially issued the first batch of unmanned road test licenses to Baidu and Xiaoma Zhixing; Shanghai, Guangzhou, Shenzhen, Wuhan, Changsha and other places have successively issued similar policies to guide the demonstration operation of autonomous vehicles.
There is still a big gap from demonstration operation to mass production on the road, how to identify automatic driving? How to define the insurance liability of autonomous vehicles. These issues also require the establishment of laws and regulations related to autonomous driving in order to reach a social consensus.
On July 27, 2021, the German Federal Government officially issued the Autonomous Driving Law, which will come into force on July 28, 2021, according to which the German Federal Motor Transport Authority can issue driving licenses for vehicles with autonomous driving functions, while allowing L4-level autonomous vehicles to drive in specific areas of public roads in specific scenarios,
At the end of 2021, Mercedes-Benz obtained the first L3 level automatic driving system road permit issued by the German Federal Motor Transport Administration, Mercedes-Benz plans to release two mass production models equipped with automatic driving systems in 2022, in the highway, 60 km / h design and operation conditions to achieve L3 level automatic driving, if it can be successfully delivered as planned, Germany will become the first L3 level automatic driving large-scale mass production on the road.
Based on the above information, global governments are gradually completing autonomous driving policies and legal preparations, providing clear technical guidance for the classification of autonomous driving, and providing a legal basis for the large-scale application of automatic driving.
The rules of the game have been formulated, whether it is a joint venture, or an independent brand, whether it is a new car-making force, or a traditional car factory, all players are fighting hard and sprinting to automatically drive mass production models. Head players, such as: Tesla, Xiaopeng, Mercedes-Benz, has been in a state of "gun start", the curtain of automatic driving mass production competition has been opened, as long as the national regulatory authorities complete the legal system, infrastructure construction, automatic driving mass production models will "spray out".
The crowd is racing against the deer, and each is good at winning
Although China's autonomous driving laws and regulations have yet to be improved, and the regulatory authorities have not yet approved the license for autonomous driving (level 3 or above) mass production models, the market size of 10 trillion yuan still attracts many players to participate in autonomous driving (including assisted driving) technology competitions, in 2021, Baidu, Xiaomi, 360 and other technology companies invested heavily in the already relatively "crowded" car-making "track", and the automotive industry presents a competitive situation of competing with each other.
Compared with personal computers and smartphones, autonomous driving presents a different competitive landscape.
Due to the relatively moderate competition, the Wintel model of "chip (Intel) + operating system (Windows)" has been established in the initial stage of the personal computer market;
In the early stage of smart phones, Qiao Gang was far-sighted, independently controlled the "chip (ARM) + operating system (iOS)", almost single-handedly defined the ecology of smart phones, showing a trend of domination. In the face of this first-mover advantage, Google was forced to acquire and open source Android systems, and joined hands with chinese and Korean smartphone brands to establish an AA model of "chip (ARM) + operating system (Android)" that competes with Apple's self-developed route;
The competition in the automatic driving track is extremely fierce, the starting gun has not yet started, and the technology factories and start-ups have flocked in, forming a situation of today's crowds of deer, and the competition for automatic driving at this stage is mainly concentrated in the field of "chip + algorithm".
Traditional automotive controllers usually choose operating system solutions that meet the AUTOSAR (OSEK/VDX compatible) standard, such as VECTOR, ETAS, etc., but in the field of automatic driving, due to the need to deal with a large number of AI parallel computing, the traditional AUTOSAR standard can not be supported, so there is a situation of co-emergence, Tesla chose to customize based on Linux, Baidu chose based on ROS customization, of course, there are also traditional car manufacturers choose AUTOSAR ADAPTIVE.
Since the operating system suitable for automatic driving has not yet formed a standard, each car factory develops chips and algorithms for automatic driving based on a customized operating system, so the algorithm has the opportunity to replace the operating system and match the appropriate automatic driving chip, becoming the focus of competition. At this stage, technology companies rely on the technical capabilities accumulated in the field of general AI to cross-border into the field of automatic driving, hoping to establish new technical barriers and redefine the automotive industry through the collaborative design of software and hardware of chips + algorithms.
Chip: Domestic chips to achieve a breakthrough
Most self-driving players choose the model of self-developed algorithm + cooperative chip, resulting in explosive growth in demand for autonomous driving chips, in addition to the traditional AI chip giants NVIDIA, Mobileye, Intel, Qualcomm and other consumer electronics chip giants have "kicked the ball", while horizon, black sesame, xinchi technology and other startups in China's rapid rise.
When it comes to autopilot, Tesla is the number one player that can't be avoided. From the early "Mobileye chip + self-developed algorithm", "NVIDIA chip + self-developed algorithm", to the final choice of "self-developed chip (FSD) + self-developed algorithm", Tesla has copied Apple's success story in the field of smartphones in the field of autonomous driving.
Personally, I think that the player closest to Tesla in the field of automatic driving is Huawei. Like Tesla, Huawei also has the ability of "self-developed chip (Ascend chip) + self-developed algorithm". Due to the Sino-US trade war, Huawei's self-developed self-developed autonomous driving technology development route has been affected, although HUAWEI INSIDE has been used as a highlight of AITO high-end electric vehicle sales, but there is still no official announcement of the car, perhaps in the near future, this technology giant or will personally go down and directly challenge Tesla.
In addition to Tesla and Huawei, other players basically need to rely on partners' chips to participate in industry competition, and are more or less limited by chip manufacturers in rapid product iteration and software and hardware collaborative design, affecting development efficiency and quality. In 2021, the autonomous driving research of several automakers has been affected by the postponement of NVIDIA's ORIN chip.
Also in 2021, the ideal ONE model for the first time equipped with the domestic automatic driving chip "Horizon Journey 3", with the self-developed NOA navigation assistance driving system, breaking the monopoly of the traditional chip giant on the front-loading market of the automatic driving chip, to achieve the first front-loading mass production of domestic automatic driving chips, which is of historical significance, the current Horizon Journey series chips have been widely equipped with a number of mass-produced models, and the shipment volume has reached 800,000 pieces.
Algorithm: Pure vision vs lidar
On July 10, 2021, Tesla officially launched FSD Beta Version 9.0, the first self-driving car without a radar system and relying solely on cameras as sensors. In the future, Tesla's Model 3 and Model Y models delivered in the North American market will no longer be equipped with millimeter-wave radar, and Tesla will begin to enter the era of "pure vision".
Personally, I believe that Tesla's choice of pure vision sensor route means that the "bet" AI vision algorithm can evolve rapidly. In 2015, in order to break the basic framework of AI controlled by Google, Facebook, and AWS, Musk led the establishment of openAI, a non-profit artificial intelligence research company, hoping to completely get rid of the AI technology dependence on technology giants (how like getting rid of NVIDIA's chip dependence and the joke of self-developed chips). At the same time, Tesla is indeed constantly evolving and even rewriting the autopilot algorithm, from the earliest 2D visual perception, to the top view of multi-camera fusion perception, to the 3D environment modeling and 3D dynamic and static object detection, Tesla is "burying its head" on the road of pure visual perception.
At the same time, with the mass production application of lidar, 2021 is defined as the "first year of lidar". A large number of Chinese manufacturers, such as: Sagitar Juchuang, DJI, Huawei, Hesai Technology, Tuttong, etc., rapid entry, high-intensity investment, may quickly reduce the cost of mass production of lidar, provide low-cost solutions for automakers at the same time, and bring a huge impact on Tesla's pure visual solutions.
At the 2021 Guangzhou Auto Show, the latest models officially announced by the new domestic car-making forces, such as: Xiaopeng G9, Weilai ET7, Ideal X01, etc. are all equipped with lidar, as if lidar has gradually become the "standard" of a new generation of autonomous vehicles. From the information analysis of many channels, although equipped with lidar, the core algorithm of the new domestic car-making forces is still based on visual perception, supplemented by radar perception, and some teams have proposed a multi-sensor pre-fusion solution, but there is still a lack of mass production models. Now the new forces are building cars equipped with multiple laser radars, which is suspected of "stacking materials".
Due to the different perception technology schemes will lead to huge differences in the automatic driving algorithm, if lidar becomes the standard of high-end autonomous vehicles in the future, Tesla's perception algorithm needs to be greatly adjusted, whether it will affect Tesla's participation in high-order automatic driving competition is still unknown, let's wait and see!
Technology evolves and the future can be expected
In 2021, the competition in the field of automatic driving is becoming increasingly fierce, new car-making forces, traditional OEMs, smart parts suppliers and other players are striving to move forward on their respective tracks, in addition to the continuous evolution of vehicles, chips, sensors, and algorithms, electronic and electrical architecture (SOA) related to automatic driving, body chassis (skateboard chassis) is also constantly innovating, after 200 years of history, the automotive industry has once again shone with dazzling light, and the future of automatic driving can be expected!
Author: Yu Bin, CTO of Lianyou Technology, bachelor's degree in automotive electronics, master's degree in industrial automation, more than 25 years of experience in the automotive industry, long-term engaged in enterprise information planning, design, development and operation and maintenance work, has presided over the preparation of Dongfeng Motor Group intelligent network vehicle strategic planning and information planning.
The New Programmer 002: The New Database Era & Software-Defined Cars, created by more than 60 experts. The book is accompanied by "2021 Database Panorama V1.0" and "2021 Automotive Technology and Industry Ecology V1.0", as well as "2021 Database Development Research Report" and "2021 Software-Defined Automotive Research Report", which are presented with graphics and video multimedia.