Text: Shi Yaqiong (36Kr) Ben (36Kr) Meng Jiayue (Hanergy) Huang Meng (Hanergy) Zhang Xinwei (Hanergy) Wu Hao (Hanergy) Yin Shaoyi (Hanergy)
In the just-concluded 2021, on November 25, the Beijing Intelligent Connected Vehicle Policy Pilot Zone officially released the "Implementation Rules for the Pilot Management of Autonomous Driving Travel Service Commercialization in the Beijing Intelligent Connected Vehicle Policy Pilot Zone (Trial)" and issued the first batch of domestic autonomous driving vehicle charging notices to some enterprises.
Beijing has become the first city in China to explicitly recognize the "Robotaxi" commercial pilot for autonomous driving, marking that the domestic autonomous driving track has finally entered the "second half" - the commercial operation stage. Autonomous driving in scenarios such as high-speed trunk lines, last-mile logistics, mining areas, ports, and sanitation will also officially enter the first year of commercialization.
Internationally, leading players in China's autonomous driving field have gradually become leaders in the international autonomous driving industry. According to data from the "2020 Autonomous Driving Disengagement Report" released by the California DMV in 2021, AutoX and Pony.ai ranked in the top five for MPI (Miles per Intervention), closely following Waymo and Cruise.
According to data from Tianyancha, the autonomous driving track entered an explosive period starting in 2016, with financing continuing to rise thereafter. In 2021, related financing reached as high as 94 cases, the highest in the entire automotive mobility sector, with disclosed financing amounts exceeding 43.5 billion yuan, a historical high year-on-year.
Among the more than ten large-scale financings exceeding $300 million in the automotive sector throughout the year, autonomous driving and related tracks occupied five seats, namely: Horizon Robotics, Momenta, Banma Zhixing, WeRide, and Hesai Technology.
From an industry perspective, with the maturity and commercial implementation of autonomous driving technology, the original value distribution pattern of the automotive industry chain will be disrupted. Cars will no longer be driving tools subordinate to humans, but will become autonomous navigation transport robots.
Looking ahead, autonomous driving will profoundly change the automotive industry. Cars are likely to be divided into two categories: one is human-driven cars, and the other is mobility service cars. Traditional automobile manufacturers will gradually transform into operation service providers, offering users one-stop services such as MaaS (Mobility as a Service) / TaaS (Transportation as a Service).
Objectively speaking, autonomous driving is a super track with a long slope and thick snow. It is currently at a critical stage of climbing. Policies need to be improved, technology is still continuously iterating, and business models are undergoing fundamental changes. At the same time, the industry is also filled with many controversial topics.
This report is jointly completed by 36Kr and Hanergy Investment, with more than 50,000 words of research content. It aims to dissect China's autonomous driving field from the following aspects, restoring the current autonomous driving industry, segmented tracks, and venture capital situations in China from different angles, as well as our predictions for the future of autonomous driving and even the automotive and mobility industries. Readers can obtain the full report through the download link at the end of the article.
Since 2009, the National Natural Science Foundation of China has held an annual "Intelligent Vehicle Future Challenge" to develop unmanned vehicles with natural environment perception and intelligent behavior decision-making capabilities, and to test research results through autonomous driving in real road environments. In 2011, the Hongqi HQ3 unmanned vehicle developed by the National University of Defense Technology in cooperation with FAW completed a 286-kilometer fully unmanned highway experiment from Changsha to Wuhan, marking new technological breakthroughs in complex environment recognition, intelligent behavior decision-making, and control for unmanned vehicles in China.
In 2013, Baidu launched its unmanned driving project, conducting road tests in Beijing and California, and opening unmanned vehicle trial operations in Wuzhen. Huawei started in the same year as Baidu, gradually entering the ranks of connected vehicle suppliers through cooperation with car manufacturers. AutoX entered the scene at the end of this stage, completing open road tests in the same year after its establishment in 2016. Pony.ai was founded in December 2016, entering the Robotaxi field. Jingchi, Roadstar, and others successively joined, and the wave of autonomous driving arrived.
During this stage, China's autonomous driving achieved leapfrog development. Baidu released the Apollo plan and, after four years of evolution, realized the commercial exploration of Robotaxi implementation, while L4-level autonomous driving solutions have also been reduced in dimension and installed in mass production; Huawei clarified its market positioning, providing high-quality intelligent vehicle solutions for OEMs, empowering them with core technologies such as autonomous driving and connected vehicles; autonomous driving unicorns and Pony.ai continued to advance their technology, successively launching and improving products and services. During the same period, the autonomous driving industry chain gradually matured, with upstream core sensor manufacturers emerging continuously and downstream scenario solutions beginning to be implemented.
According to data from the "2020 Autonomous Driving Disengagement Report" released by the California DMV in 2021, leading players in China's autonomous driving field have gradually become leaders in the international autonomous driving industry. AutoX and Pony.ai ranked in the top five for MPI (Footnote: Miles per Intervention, the average mileage interval per intervention), closely following Waymo and Cruise.
It is planned that by 2025, the technological innovation, industrial ecology, infrastructure, regulations and standards, product supervision, and network security systems for China's standard intelligent vehicles will be basically formed.
In addition to actively promoting policy and legislation, since 2018, intelligent connected vehicle demonstration zones have also been blooming across the country. In addition to a batch of intelligent connected or autonomous driving demonstration zones promoted in cooperation with the Ministry of Industry and Information Technology, some provinces and cities have successively created intelligent connected vehicle testing scenarios based on their own industrial needs through cooperation with institutions or capital cooperation. Among them, the Beijing High-Level Autonomous Driving Demonstration Zone, with the entire Beijing Economic and Technological Development Area as its core, is the world's first networked cloud-controlled high-level autonomous driving demonstration zone. Since its establishment in 2021, it has cumulatively opened 1,000 kilometers of autonomous driving test roads, with test mileage exceeding 3 million kilometers, and has opened 56,400 5G base stations (data as of January 2022).
1.1.2.2. Technology: Automotive electronic and electrical architecture shifts from distributed to centralized, software and hardware decoupling improves software development efficiency
Traditional cars mainly adopt a distributed electrical architecture, where each vehicle function corresponds to one or more ECUs (Electronic Control Units), and signals are transmitted between ECUs via the CAN bus. ECUs are mainly used to receive information from sensors, process it, and output corresponding control commands to actuators for execution. The main work of vehicle enterprise electronic control system development (software algorithms, matching and calibration, etc.) relies on ECUs.
With the increase in automotive electronification, the number of ECUs in vehicles has reached hundreds, provided by different suppliers, leading to pain points such as inability to coordinate algorithms, redundancy, difficulty in unified maintenance, and unified OTA upgrades. To this end, Tier 1 suppliers such as Bosch, Continental, and Aptiv have launched new generations of electronic and electrical architectures, whose main technologies include gateways, domain controllers, and automotive Ethernet, achieving an upgrade from distributed to domain-centralized architecture, ultimately moving towards central computing to achieve vehicle-cloud synergy. The centralized architecture can reduce computational redundancy and improve utilization, while centralized controllers are more convenient for coordinating multiple sensors to jointly perceive the environment inside and outside the vehicle and make overall decisions.
In recent years, camera, millimeter-wave radar, and ultrasonic radar technologies have become increasingly mature in automotive applications, with prices continuously declining. Currently, the industry is optimistic about a significant reduction in autonomous driving costs. Among them, the unit price of vehicle cameras continues to decline, currently around 150 yuan, with expected relatively small future declines; the market supply unit price of millimeter-wave radar is about 500 yuan, with some room for future decline; LiDAR prices have remained high, mainly used in surveying, mapping, and industrial production in the past. In recent years, with new technology routes such as solid-state LiDAR replacing traditional mechanical radar, process costs have significantly decreased, and with future autonomous driving technology development driving up supply volumes, there will be greater room for decline. In 2019, Luminar released a LiDAR solution priced at less than $1,000. Velodyne plans to reduce its average selling price from $17,900 in 2017 to $600 by 2024. In 2020, Huawei announced that its mass-produced LiDAR unit price would be below $200.
From a technology perspective, the United States is highly advanced in single-vehicle intelligence. As the country with the highest number of artificial intelligence enterprises globally, the United States leads the world in the AI field, with sufficient talent reserves, strong basic scientific research capabilities, and related enterprises distributed across the foundation, technology, and application layers. In addition, the United States has developed integrated circuit technology and has maintained a leading position in high-end chip design, laying a good foundation for the development of high-performance vehicle chips.
In contrast, China's "cloud + vehicle + road" technology route has the opportunity to overtake on curves. China has a large number of 4G and 5G base stations with wide coverage, and the Chinese government strongly promotes the construction of new infrastructure such as 5G networks, the Internet of Things, satellite internet, and data centers, supporting the evolution from LTE-V2X to 5G-V2X, giving it a relatively obvious advantage in vehicle-road coordination technology.
From a commercial perspective, due to a significant labor shortage in the United States, companies are more willing to pay for personnel replacement, with progress more concentrated in Robotaxi and unmanned logistics. In 2018, Waymo took the lead in offering free unmanned taxi services to its early users in Arizona. In October 2020, Waymo One opened its driverless taxi business to the public for the first time in Phoenix. In February of the same year, the National Highway Traffic Safety Administration (NHTSA) approved Nuro to deploy unmanned delivery vehicles first. In contrast, China has a greater acceptance of artificial intelligence and is more willing to pay for autonomous driving from the perspectives of safety improvement and efficiency enhancement, with relatively faster commercial progress in special scenarios.
According to data from Tianyancha, the autonomous driving track entered an explosive period starting in 2016, with a total of 38 autonomous driving-related financings throughout 2016. Subsequently, financing in the autonomous driving track continued to rise, reaching as high as 94 cases in 2021, with disclosed financing amounts exceeding 43.5 billion yuan.
The growth in industry financing is partly due to the spread of ample liquidity to the primary market. More importantly, the development and application implementation of autonomous driving technology have gradually gained capital recognition, and the importance attached by internet giants and automotive leaders has increased.
The realization of autonomous vehicle functions requires the participation of multiple entities such as automobile manufacturers, parts suppliers, vehicle computing platform developers, and mobility service providers. The upstream includes the perception, transmission, decision-making, and execution layers; the midstream is the platform layer, including integrated intelligent cockpit platforms, autonomous driving solutions, and traditional connected vehicle TSP platforms; the downstream mainly consists of vehicle manufacturers and third-party service providers.
The upstream includes the perception, transmission, decision-making, and execution layers. The perception layer consists of vehicle cameras, radar systems, high-precision maps, high-precision positioning, navigation systems, and roadside equipment; the transmission layer provides signal transmission for autonomous driving based on communication equipment and services, mainly including communication equipment and communication services; the decision-making layer includes computing platforms, chips, operating systems, algorithms, etc.; the execution layer executes decision commands, including wire-controlled, electronic drive/steering/braking, system integration, and other automotive parts manufacturers. The four systems are interlinked to achieve vehicle networking functions.
1) The perception layer is used to perceive changes in the external environment and obtain relevant information. It mainly includes intelligent hardware (sensors, RFID, and vehicle vision systems, etc.), navigation (GPS, BeiDou, and inertial navigation systems), and roadside equipment. Intelligent hardware is the "eyes" of intelligent vehicles. The unmanned driving hardware system includes sensors, RFID, vehicle vision systems, etc. With connected vehicles and intelligent interconnection becoming future trends, the demand for related hardware products is also increasing. The navigation system is the compass of intelligent vehicles. The navigation and positioning of unmanned vehicles mainly obtain the vehicle's position, heading, and speed in real time through the Global Positioning System (GPS), BeiDou Satellite Navigation System (BDS), and inertial navigation systems. Roadside equipment is a necessary condition for ensuring the realization of "vehicle-road coordination" in autonomous driving. If autonomous driving only has vehicle-side data, it is difficult to achieve safe and accurate driving; a series of roadside equipment is also essential. The key construction of the connected vehicle roadside includes RSU, roadside intelligent traffic management facilities, MEC equipment, etc.
2) The transmission layer converts environmental information obtained by the perception layer into signals transmitted to the decision-making layer based on communication technology, similar to the transmission nerves of the human body. The transmission layer mainly includes communication equipment and communication services, where communication equipment mainly includes components and information interaction terminals, and communication services mainly include DSRC and C-V2X, two wireless communication technologies serving autonomous driving.
3) The decision-making layer uses the information fed back from the perception and transmission layers to establish corresponding models and formulate appropriate control strategies. Due to the complexity of real road conditions and different solutions adopted by different people for different road conditions, decision algorithms need to cover massive data of most rare road conditions and perfect and efficient artificial intelligence technology. Functionally, the decision-making layer mainly includes core components such as operating systems, chips, algorithms, high-precision maps, and cloud platforms. The current threshold for truly realizing autonomous driving technology lies in the decision-making layer.
4) The execution layer is the bottom layer of the unmanned driving system. Its core task is to enable the vehicle to stably drive according to the trajectory planned by the decision-making part through the coordination of drive, brake, and steering control systems, while also being able to achieve actions such as avoidance, maintaining distance, and overtaking. With the development of autonomous driving, the execution layer, which was previously driven by human force and promoted through vacuum and hydraulic methods, is gradually being replaced by electronic and electrified systems. Wire-controlled technology, which replaces mechanical force with electrical signals, will be fully penetrated in the autonomous driving era.
After technical research and development at various levels, vehicle manufacturers ultimately carry out technology integration and production assembly, completing the final step of unmanned driving product production. Vehicle manufacturers provide application and practice platforms for key technology enterprises, and technology enterprises directly provide solutions and components to vehicle manufacturers.
The assembled vehicles are partly sold directly to consumers and partly become service vehicles for third-party service providers. Such service providers are generally mobility service providers, and the data feedback from their services will assist vehicle manufacturers and technology enterprises in adjusting product research and development. Some vehicle manufacturers are also transforming into third-party service providers or conducting in-depth cooperation with them. The entire unmanned driving industry chain is showing an ecological and networked trend.
In the future, midstream manufacturers that deliver end products or services will have the highest industry value. With the maturity and commercial implementation of autonomous driving technology, the original value distribution pattern of the automotive industry chain will be disrupted. Cars will no longer be driving tools subordinate to humans, but will become autonomous navigation transport robots. In the future, core components will shift from transmission systems that reflect power and driving control experience to intelligent software systems and processor chips that reflect the level of autonomous driving technology, and the value chain will shift from traditional OEMs to technology companies. The profits of non-core components and vehicle manufacturing will be further compressed, while the profits at both ends of the industry chain will significantly increase. In the future, solution providers focused on design and R&D, as well as mobility and operation service providers closer to users, will have higher profits.
The autonomous driving industry chain is complex, with upstream and downstream covering rich and detailed segments. In this report, we select representative scenarios in each link for coverage, focusing on the upstream Tier 1, ADAS, and chip fields, and downstream solution providers for various application scenarios. Among them, due to the wide range of autonomous driving application scenarios, they can be further divided according to their functions into passenger-carrying scenario solutions mainly based on Robotaxi, cargo-carrying scenario solutions mainly based on high-speed trunk lines, last-mile delivery, mining areas, ports, etc., and sanitation cleaning scenario solutions.
In the era of autonomous driving, the application of new intelligent technologies and equipment in the automotive field presents a historic opportunity for China's emerging Tier 1 companies.
According to Roland Berger's forecast, by 2025, globally, it is expected that only 14% of vehicles will not have ADAS functions, 40% will have L1-level functions, 36% will have L2-level functions, and 10% will have L3-level or higher functions.
According to data from Gaogong Intelligent Vehicle Research Institute, from January to April 2020, Mobileye ranked first among front-view perception solution suppliers, with a lead of nearly 10 percentage points over the second and third places, Continental and Bosch. Continental Group benefited from the gradual launch of some Toyota TSS2.0 solutions and the switch from Denso, ranking slightly ahead of Bosch in market share from January to April.
However, because the state encourages independent innovation, in the field of intelligent connected vehicles, a large number of startups have entered from radar, cameras, parking, etc., and some of them have achieved a certain scale.
From the perspective of the distribution and scenario layout of major Tier 1 products related to autonomous driving internationally and domestically, Bosch and Continental have the most comprehensive product lines and layouts among Tier 1 suppliers.
Taking the interviewed company Continental Group as an example, according to its financial report data, between 2018 and 2020, Continental Group received orders totaling more than 9 billion euros from global automobile manufacturers in the ADAS field.
Continental Group's latest generation multi-function camera MFC500 series and 4D imaging radar will enter the mass production and installation cycle. At the same time, its automotive AI chip is expected to begin batch production in 2026. Continental Group stated: "With the completion of the chip link, Continental has taken a key step towards the highly specialized sensor modules and control units required for future high-performance vehicle computers."
According to its latest disclosed information, Continental Group will conduct autonomous driving and unmanned driving road tests under complex road conditions this year, and implement new technologies in multiple fields such as braking systems, 5G connectivity, Ultra-Wideband (UWB) technology, material development, and human-machine interaction, while launching multiple mass production projects.
Currently, more than 70% of L2 and above advanced driving assistance systems use Mobileye's vision solutions. Even many Tier 1 suppliers that previously developed their own vision algorithms have chosen to give up self-development and directly cooperate with Mobileye to quickly occupy the market.
Mobileye is in an obvious advantageous position in the front-loading market. Especially after the release of EyeQ4, with the help of Intel, its goal is no longer to be a simple vision solution supplier, but to become a leading enterprise in the autonomous driving field. Mobileye's solution adopts a black-box model, and traditional Tier 1 suppliers can no longer obtain sufficient service support from Mobileye. Leading Tier 1 suppliers have realized the crisis brought by Mobileye's dominance and have sought other solutions. Bosch, Continental, and Denso insist on using self-developed algorithms, benchmarking Mobileye's EyeQ4 solution.
Domestic Tier 1 suppliers have awakened relatively slowly and have begun to catch up in recent years. Unlike Bosch and Continental, domestic Tier 1 suppliers more often adopt a binding cooperation model of in-depth cooperation with OEMs or the establishment of joint ventures, jointly developing and implementing autonomous driving technologies, helping OEMs expand new high-end brands, and accelerating the transformation of the "New Four Modernizations." For example:
Geely Holding and Baidu formed a smart electric vehicle company, with Baidu fully empowering the joint venture with technologies such as artificial intelligence, Apollo autonomous driving, Xiaodu Vehicle, and Baidu Maps; SAIC and Alibaba jointly established the high-end pure electric vehicle brand "IM Motors," which applies Ali's Banma connected vehicle system and adopts SAIC Group's three-electric core and intelligent driving technologies; Changan, Huawei, and CATL will jointly establish a new high-end intelligent vehicle brand, with the three parties jointly developing the CHN intelligent electric vehicle platform, equipped with Huawei's intelligent cockpit platform CDC, autonomous driving domain controller ADC, and some three-electric components. Zhou Zhifeng, partner of Qiming Venture Partners, believes that intelligence and new energy are recognized as the two major trends in the automotive industry. The coupling of these two major trends at the same point in time has also created the biggest opportunity in the automotive industry in recent years. However, the acceptance and overall implementation speed of automotive intelligence have clearly exceeded market expectations. Traditional Tier 1 suppliers do not have enough time to fully prepare for product upgrades and are currently in a state of accelerating to make up for their shortcomings. But the basic know-how of the automotive industry, such as chassis electronic control, is still in the hands of traditional Tier 1 suppliers, and the market still relies heavily on them.
In addition to the attempts of traditional Tier 1 manufacturers in the autonomous driving field, startups represented by idriverplus and Hongjing Zhijia have also entered the ranks of leading Tier 1 suppliers by virtue of cutting-edge technology research and development and rapid business implementation.
Based on the current situation, two types of Tier 1 suppliers will emerge in the future: one type deeply cultivates vertical field technology, establishing in-depth cooperation relationships with OEMs to efficiently customize development for them; the other type, represented by traditional Tier 1 suppliers, focuses on breadth, providing general overall solutions for a large number of OEMs, while helping OEMs quickly implement solutions in standardized links such as functions, safety, hardware, testing, and verification. That is, at the Tier 1 level, it is necessary to adapt to the standardized cost reduction needs of OEMs and meet the differentiated efficiency enhancement needs of car companies.
It is also worth noting that while international Tier 1 suppliers are achieving multi-level functional implementation, they have begun to restart underlying system research and development. As a bridge between systems and software applications, Tier 1 suppliers are successively releasing middleware products. Through comprehensive sensor product layouts, they centrally configure autonomous driving solutions for OEMs, reducing the complexity of system integration, reducing development costs, and accelerating product implementation.
Previously, ADAS was generally benchmarking against Mobileye
