Behind the hot stock market of car, road and cloud: the trillion-level track is surging and the problems to be solved
2024-08-12
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
During the summer evening rush hour, Baidu Maps shows that the traffic conditions in Beijing are all red. But in many areas of Beijing Yizhuang Economic and Technological Development Zone, the residents feel a little different - a touch of green or yellow appears in the red.
As the only national-level economic and technological development zone in Beijing, Yizhuang has a significant separation of work and residence, forming a typical tidal traffic phenomenon during peak hours in the morning and evening. There are many industrial parks here, and large transport vehicles travel frequently, accompanied by pedestrians, non-motor vehicles and motor vehicles, which greatly increases the complexity of traffic.
Nowadays, when walking in the Yizhuang high-level autonomous driving demonstration area, you can see smart integrated boxes and renovated multifunctional integrated poles on the roadside. The demonstration area has deployed roadside perception systems, traffic signal control systems, and vehicle-road cooperative communication systems, and has achieved vehicle-road interconnection through laser radar, millimeter-wave radar, and artificial intelligence cameras.
In an area covering about 60 square kilometers in the main urban area, not only are there many autonomous driving companies operating, including Luobo Kuaipao, Pony.ai, and WeRide, but more than 300 intersections have been transformed into intelligent ones, which is like installing a "brain" and "eyes" for the roads, effectively directing traffic and significantly improving road traffic efficiency and safety.
"Smart cars" need "smart roads", and the expansion and exploration of business operation service models are currently being accelerated. After the expansion is officially completed, it is expected that more than 1,000 new smart intersections will be added, which is expected to alleviate traffic congestion to a greater extent.
The Yizhuang pilot is just a microcosm of the national vehicle-road-cloud integration process. Since January this year, the Ministry of Industry and Information Technology, the Ministry of Public Security, the Ministry of Transport and other relevant ministries and commissions have successively issued notices to promote the "vehicle-road-cloud integration" of intelligent connected vehicles. Since July, unmanned driving and vehicle-road-cloud concept stocks such as Volkswagen Transportation (600611.SH) and King Long Automobile (600686.SH) have soared nearly 200% in 20 trading days. Whether from the perspective of top-level design or market enthusiasm, "vehicle-road-cloud integration" has entered a critical period of industrial scale construction and application.
According to reports from the China Society of Automotive Engineers and other organizations and estimates by industry insiders, the cost of transforming China's more than 5.4 million kilometers of roads into "smart" roads will be around 2.7 trillion yuan. For this reason, vehicle-road-cloud integration is considered to be my country's next trillion-level "new high-speed rail" plan.
However, a reporter from the First Financial Daily recently conducted on-site investigations in Beijing, Shanghai and Guangzhou and found that although the vision of a trillion-level track is beautiful, there are still problems that need to be solved, such as some places acting independently, lack of unified industry standards, large differences in construction plans among different companies and local governments, and low effectiveness of road-side information.
Yizhuang Practice: Smart cars need smart roads
If you drive from Rongjing East Street to Kechuang Fifth Street, your vehicle can pass through up to 9 intersections in a row without stopping when the light turns green.
At the intersection of these two streets, the reporter from China Business News saw that there were more than a dozen cameras and other sensors installed on the multifunctional poles at the intersection. According to relevant sources, the multi-sensor detection camera also serves as an electric police/checkpoint law enforcement device, with multiple illegal monitoring functions such as not wearing seat belts, running red lights, and driving out of lane, as well as detecting traffic and vehicle license plates. The fisheye camera supplements the blind spot, and the millimeter-wave radar and laser radar on it can more accurately detect the speed and flow of vehicles, as well as the specific contours of vehicles, providing real-time and accurate data support for traffic management.
The equipment box at the intersection is equipped with edge computing equipment, which can process the above traffic data in real time and intelligently adjust the timing of traffic lights according to actual conditions to optimize traffic flow. Previous data showed that the time interval of traffic lights at each intersection in the demonstration area can be dynamically adjusted 110 times a day on average according to the actual situation of road traffic.
Han Guohua, general manager of Baidu's traffic business department, told the First Financial reporter that because these scenarios have high latency requirements and serve many scenarios, the computing power used for an intersection is about 200TOPS. The cost of an intelligent intersection, depending on the medium, low, and high configurations, is in the order of hundreds of thousands of yuan. As the scale expands, the cost is also decreasing.
He described the "vehicle-road-cloud integration" as helping single-vehicle intelligence achieve a "God's perspective", that is, to realize digital twins through the full digitization of road infrastructure to make up for the blind spots of perception by vehicle-side sensors alone.
Due to work needs, Han Guohua often has to run from the Ark Building outside the North Fifth Ring Road to Yizhuang on the Southeast Fifth Ring Road. He has also witnessed the changes in Beijing's vehicle, road, cloud and network in recent years.
He told the First Financial Daily reporter that Beijing's intelligent networking is basically based on three main lines: serving L4, serving L2, and serving urban governance.
At the beginning, the construction of intelligent network was for L4, but in fact there are too few L4 level vehicles. After investment and construction, if it cannot be replicated on a large scale, the cost-effectiveness of investment and construction is too low. After several years of exploration, the industry has included L2 level vehicles in the service scope; so far, Yizhuang full-area signal control covers more than 300 intersections in Yizhuang, with the morning rush hour, from leaving home to the destination, and the total travel time of all people as the goal for optimization, coordination and control of all intersections, and more efficient traffic.
Especially with the capability of large models, we can also use all-round perception data to analyze the possibility of congestion, and superimpose simulation and deduction to predict where congestion will occur, why it will occur, how to solve the congestion, etc. in a few hours, and automatically optimize traffic light timing plans, etc.
According to official disclosure from Beijing Yizhuang, the cloud control platform of Beijing's high-level autonomous driving demonstration zone has achieved access to data from more than 420 intersections in Beijing Economic Development Zone, with a total of more than 800 autonomous driving test vehicles. It can receive and process about 300T of data every day. On the large screen of the cloud control platform, the road conditions and their digital twin scenes, intersection intelligent signal control data, etc. are displayed in real time, and the operating status of the autonomous driving vehicle can be queried.
In addition, the cloud control platform will dynamically publish real-time information covering more than 300 intersections. Intelligent facilities at intersections can push information such as traffic light status and green wave recommended speed to the map business navigation app and connected vehicle terminals in real time, and provide driving assistance to drivers through voice prompts such as "The recommended speed is 60 to pass multiple intersections with green lights."
Especially in terms of traffic control optimization, official data show that with the cloud control platform as the data base, 67 roads within the 60 square kilometers of Beijing Economic and Technological Development Zone have achieved green wave traffic and 257 intersections have been dynamically optimized. In the past year, the average delay at intersections has been reduced by 33% and the average vehicle speed has increased by 45%.
According to the reporter, in the future, the demonstration zone's cloud control platform will serve as the city's unified cloud control platform, and will provide support services to Beijing Economic and Technological Development Zone, Tongzhou District, and Shunyi District. It will expand from the original 60 square kilometers to nearly 600 square kilometers, and further explore support capabilities in areas such as deep data open operation, comprehensive smart transportation management, and artificial intelligence industry services.
The 10 billion-level project was launched
"China's electric vehicles are already leading the world. If it wants to maintain its lead, intelligent networking is the second stage. So everyone is paying attention to the car-road-cloud track and considering how to do better," Han Guohua told the first financial reporter.
An industry insider told the First Financial Daily that in January this year, the company's focus was not on intelligent networking, but with the release of the list of pilot cities for the application of "vehicle-road-cloud integration" of intelligent networked vehicles and the bidding in multiple cities, the establishment of relevant departments has also given the industry greater market space. Securities firms are also conducting intensive research. Currently, the industry is working hard to win projects in these 20 cities. The industry expects that projects in these pilot cities will be released in the second half of this year and the first half of next year.
Regarding the specific construction situation, according to data from the National Intelligent Connected Vehicle Innovation Center, my country's vehicle-road-cloud integration industry can be divided into four parts: intelligent connected vehicles, intelligent roadside infrastructure, cloud control platform, and basic support. In 2025, the output value of roadside infrastructure is expected to reach 22.3 billion yuan, accounting for about 26.4%.
Han Guohua told reporters that behind the construction of "vehicle-road-cloud integration", close cooperation among multiple departments and enterprises was involved, forming a relatively close division of labor system.
The government is responsible for policy formulation, fund raising and project supervision. The government has issued relevant policy documents to clarify the goals and requirements of the vehicle-road-cloud network construction, and invested a large amount of financial funds to promote the implementation of the project. At the same time, the government has also introduced excellent companies to participate in the construction through bidding and other means.
As an important force in infrastructure construction, operators are responsible for the laying and maintenance of communication networks. Operators use advanced communication technologies such as 5G to provide low-latency, high-reliability data transmission channels for vehicle-road-cloud networks, ensuring smooth communication between various sensing devices and cloud platforms.
There are also technology companies such as Baidu and Alibaba that provide core technologies and solutions. Technology companies use their own technological advantages to provide algorithms, cloud platforms and data support for the construction of vehicle-road-cloud networks. For example, Baidu provides vehicles with accurate path planning and obstacle avoidance warning services through its intelligent driving and intelligent transportation solutions.
The general integrator is responsible for the purchase, installation and system integration of equipment. According to project requirements, the integrator purchases various sensing devices, computing units and software platforms, and installs and debugs them to ensure the stable operation of the system. At the same time, the integrator is also responsible for the subsequent maintenance and upgrade of the system. At present, operators and technology companies often play the role of "general integrator".
At present, the cost of converting a road into a vehicle-road-cloud network is affected by many factors, including the length of the road, the type and quantity of equipment required, the complexity of the technical solution, and regional differences, making it difficult to calculate a fixed cost figure. From the perspective of road-side transformation alone, "vehicle-road-cloud integration" covers a complete set of digital solutions including perception units (sensors), computing units (chips, etc.), and communication units. The cost of the transformation may currently be millions per kilometer.
In recent months, many places have launched intensive demonstration projects of vehicle-road-cloud integration. In July this year, Changchun City announced plans to invest 12.7 billion yuan in the next three years to carry out the construction of "vehicle-road-cloud integration" and comprehensively create typical demonstration application scenarios such as transportation hubs, urban roads, expressways, and highways. In June this year, the latest data from the Hubei Provincial Investment Project Online Approval and Supervision Platform showed that Wuhan's major "vehicle-road-cloud integration" demonstration project has been approved by the Municipal Development and Reform Commission, with a registered amount of 17 billion yuan.
Taking the "Beijing Vehicle-Road-Cloud Integrated New Infrastructure Construction Project (Preliminary Design, Construction Drawing Design) Tender Announcement" issued by Beijing as an example, the investment scale of this project is 9.939 billion yuan. The project is planned to be constructed in Tongzhou District, Shunyi District, Chaoyang District and other areas, covering 6,050 road intersections within an area of about 2,324 square kilometers, and includes the construction and renovation of the dual-intelligence private network center.
Previously, Guotai Junan Securities had disassembled the above-mentioned bidding announcement and found that the value of a single intersection of roadside intelligent infrastructure was as high as 434,000 yuan, among which the value of edge computing nodes MEC, smart terminals RSU and lidar accounted for relatively high proportions, reaching 13.38%, 5.33% and 1.49% respectively.
According to the "Forecast of Incremental Output Value of the Intelligent Connected Vehicle Industry with Integrated Vehicle-Road-Cloud" released by the China Society of Automotive Engineers, under a neutral expectation scenario, the incremental output value of the intelligent connected vehicle industry with integrated vehicle-road-cloud is expected to be 729.5 billion yuan and 258.25 billion yuan in 2025 and 2030, respectively. Industrial development will actively promote my country's economic growth.
A vehicle-road-cloud integration supplier told reporters that the current cost of roadside equipment in the industry is about 1 million to 2 million yuan per kilometer, and it is expected that with the increase in scale, the cost will still have room for reduction. If the cost of roadside equipment is reduced to 500,000 yuan per kilometer, the cost of national highway reconstruction will also require 2.7 trillion yuan, which is comparable to the scale of infrastructure investment in high-speed rail.
Driven by huge market expectations and the optimism brought about by frequent government policies, starting from July this year, "vehicle-road-cloud integration" and Robotaxi have become the hottest trends, and related concept stocks have continued to rise.
"From the overall perspective of the development of the entire industry, vehicle-road-cloud integration is an important trend in the future development of intelligent transportation systems and autonomous driving technologies. Its core lies in achieving a significant improvement in the efficiency, safety and intelligence of the transportation system by deeply integrating the capabilities of vehicles, road infrastructure and cloud data centers." Wu Dongsheng, senior vice president of Gosuncn Technology Group and general manager of Gosuncn Connected Technology Co., Ltd., told reporters.
Many problems to be solved
Although the vision is beautiful, the linkage between vehicles, roads and clouds still faces many technical difficulties that have not yet been overcome.
"There is still a lot of technical content that we need to continue exploring and implementing, including networks, computing power, algorithms, standards, etc." Wu Dongsheng said.
Around 2018, there were many autonomous driving startups in China that used vehicle-road collaboration technology as their corporate feature; before 2020, Ford, General Motors and other automakers began to introduce products with C-V2X (cellular network-based vehicle-to-everything system) functions in China, which could realize partial blind spot prediction, vehicle speed guidance and other functions on some pilot sections.
However, at that time, 5G networks were not yet popular, and 4G networks had large delays, making it difficult for road test units to provide effective support for vehicle intelligent driving and other systems. Today, 5G networks are relatively popular, but improvements in latency and reliability are still far from the industry's expectations.
"Compared with the 4G era, 5G has made significant improvements in latency, number of accesses, etc., but there is still a certain gap from the 'ideal'. It can be said that there is only a layer of 'window paper' missing." said Lu Bin, vice president of Mushroom Auto Union.
Wu Dongsheng believes that compared with single-vehicle intelligence, vehicle-road-cloud integration requires a powerful network with high reliability, low latency and wide connectivity to achieve the networking of vehicles, roadside infrastructure and the cloud; the functional application scenarios that vehicle-road-cloud needs to provide need to cover three different levels of networked functions: collaborative early warning, collaborative assisted driving and collaborative autonomous driving. Different functions will have different requirements for the network.
In this context, this multi-mode network form is needed to support the development of vehicle-road-cloud integration. With C-V2X as the technical foundation, and then superimposed with some operators or other dedicated networks (such as LTE-V2X, NR-V2X, 4G, 5G, 6G, satellite, ETC 2.0 and other communication technologies), a reliable multi-mode network integrating vehicle, road and cloud is built to better support the different requirements of collaborative early warning, collaborative assisted driving and collaborative autonomous driving, and realize seamless switching and efficient collaboration between different communication technologies.
However, at present, the construction of this network is still imperfect and needs to overcome multiple difficulties such as technical compatibility, network optimization and system integration.
"In addition to latency, algorithms are also a difficulty. Although the perception unit of the road test can provide the vehicle with blind spot information, there is currently no standard or formula. Should the road test unit process the data and send the results directly to the vehicle, or should it directly transmit the perception data to the vehicle and allow the vehicle to process the data autonomously and make decisions?" A management of an autonomous driving company told reporters that based on the current tests in the pilot areas, the data support provided by the road test unit to L4 Robotaxi is still relatively limited.
From a technical point of view, if the road test unit directly transmits video, picture and other data to the vehicle, the huge amount of data will bring certain difficulties to the transmission speed, and each company's intelligent driving technology solutions are not consistent, there will be problems with how the vehicle processes the data after receiving it; if the road test unit has its own computing power and algorithms, and processes the data and transmits it to the vehicle, then this puts higher requirements on the computing power and algorithm of the road test unit.
Wu Dongsheng told reporters that both single-vehicle intelligence and vehicle-road-cloud integration require the support of computing power. It is necessary to better explore the reasonable deployment of computing power in the end-edge cloud, as well as the distribution and coordination of this computing power network. On the one hand, it can meet the specific computing power requirements of the vehicle, transportation, and city, and on the other hand, avoid the waste of computing power resources, so as to better support the development of the overall vehicle-road-cloud integration industry.
In addition, vehicle-road-cloud integration emphasizes the coordinated development of various links such as vehicles, roads, clouds, networks, maps, and safety, covering different levels of networked applications such as collaborative warning, collaborative driving assistance, and collaborative autonomous driving. This requires reliable algorithms and protocol stacks with comprehensive functions and reliable performance to provide reliable business support to car companies, smart transportation users, and traffic managers. Wu Dongsheng believes that the industry needs to continue to improve its algorithm capabilities.
Intelligent transportation systems involve the collection and transmission of a large amount of sensitive data. How to ensure the security and privacy of data has become an urgent problem to be solved. Once the data is leaked or illegally used, it will have a serious impact on personal privacy and public safety. This requires the establishment of a sound data security management system to ensure the security and compliance of data during collection, transmission and use.
Commercialization model needs to be established
The vehicle-road-cloud integration project also needs to face the problem of software operating system, which requires a huge data model to support it. Taking the Beijing vehicle-road-cloud integration pilot as an example, there are 10,000 intersections in Beijing, and 7,350 intersections will be paved this time, and the number of vehicles served will reach more than 4 million, which requires a huge software platform.
In addition to technical constraints that still need to be overcome, vehicle-road-cloud integration also has problems such as inconsistent standards in different regions and unclear commercial prospects.
"Although the domestic industry standards for intelligent connected vehicles are at the forefront of the world, they still need to be further improved, especially around the integration of vehicles, roads and clouds, such as safety standards, testing and verification standards, etc. Various places have promoted local group standards and industry standards, but how to achieve unification, interconnection and mutual recognition among our local standards is also very important." Wu Dongsheng said.
The transportation sector has a strong localized management, and there are certain barriers to achieving full sharing of data between regions. The data obtained from infrastructure construction is not of sufficient quality or accuracy, and may not be available for use by smart cars. In the past, the construction standards of pilot projects were not unified, and the capabilities provided to cars in different regions were also different.
Dai Dunfeng, public relations director of Pony.ai China, told reporters that in order to achieve the best results in actual applications, the vehicle-road-cloud system also needs to be connected to some dedicated traffic networks to obtain real-time and accurate traffic information, so as to help achieve optimal traffic efficiency.
In terms of business model, many industry insiders pointed out that although the vision of vehicle-road-cloud is excellent, the profit model of the current vehicle-road-cloud integrated project is still unclear. How to form a sustainable profit model and achieve long-term operation and development of the project is an issue that the industry needs to explore in depth.
"At present, there is no consensus on whether the construction of vehicle-road-cloud networks should be classified as infrastructure for social public services or facilities with commercial operation attributes," Han Guohua told reporters.
He said that the industry needs to find more users and more scenarios to make investment and returns more cost-effective, which is the direction of future efforts. "After scale-up, costs will be reduced, so these two curves will eventually intersect and will be more in line with future driving."
The reporter learned that the funding sources for vehicle-road-cloud integration are mainly from three sources: national financial support, local financial support, and self-raised funds from enterprises. However, the industry is still exploring the specific profit model.
"The vehicle-road-cloud integration project is actually a new infrastructure project invested by the government, which will eventually provide services for autonomous driving. From a business model perspective, the early infrastructure costs such as hardware equipment are high, and are shared by the government, gradually shared to the B-end, and finally to the C-end users. When every user uses the product, the cost will be greatly reduced," said Lu Bin, vice president of Mushroom Auto Union.
Taking the Beijing vehicle-road-cloud integration project as an example, the project's construction funds come from "government investment + state-owned enterprise self-raised funds", of which the government investment is 70% and the state-owned enterprise self-raised 30%. The industry believes that government funding can leverage more social capital to participate.
Lu Bin believes that the integration of vehicle, road and cloud has moved from the past technical demonstration to the stage of large-scale construction, and the time for business model acceptance will be 2026. From the perspective of the business model, the data collected through the vehicle, road and cloud integration project will be sold to car companies, allowing them to pay for the data services, and the car companies will eventually sell them to consumers.
However, subscription-based or paid intelligent driving assistance systems are not yet popular. Currently, only a few companies such as Tesla, NIO, and Huawei's Hongmeng Intelligent Driving have launched subscription-based or paid intelligent driving assistance systems. Xpeng, Ideal, Leapmotor, etc. have equipped some of their products with advanced driving assistance systems as standard, and the costs are included in the price of the vehicle. However, the software and hardware costs of the driving assistance system are still the "bulk" of the total vehicle cost, and automakers need to spend extra money or increase the price of the vehicle to purchase vehicle-road collaboration data.
However, "McKinsey China Automotive Consumer Insights 2024" pointed out that under the influence of price wars, although Chinese consumers' acceptance of smart driving is increasing, their willingness to pay for smart driving has declined.
In the entire vehicle-road-cloud collaborative system, car companies are only one of the participants. In this system, there are also vehicle operators and individual car owners, road infrastructure builders, cloud control basic platform operators, cloud control related support platform service providers, cloud control application service providers, network operators, and high-precision map service providers. They are independent operating entities, and the profit models of different operating entities are also different.
The Guangzhou Industry and Information Technology Bureau told the First Financial reporter that the biggest difficulty encountered in promoting the construction of vehicle-road-cloud integration is how to create a closed business loop of vehicle-road-cloud integration. At present, Guangzhou's understanding of the profit model of vehicle-road-cloud integration construction is still limited to insurance services and providing data services to relevant traffic law enforcement departments. It is urgent to explore new application scenarios of vehicle-road-cloud integration with enterprises and all sectors of society. In addition, the vehicle-road-cloud integration system involves a large amount of personal and vehicle data. How to ensure data and network security and realize data sharing? How to escort the construction of vehicle-road-cloud integration in terms of legislation will also be a major challenge.
(This article comes from China Business Network)
