Five questions about "vehicle-road-cloud integration": How to unlock the "large-scale commercialization" code for autonomous driving?

Is "Vehicle-Road-Cloud Integration" an Optional or Mandatory Task?

In the selection of autonomous driving technology paths, the debate between "single-vehicle intelligence" and "vehicle-road collaboration and vehicle-road-cloud integration" has never ceased.

On one hand, single-vehicle intelligence technology is widely applied in assisted driving scenarios and is gradually expanding to higher-level autonomous driving scenarios. The "intelligentization" competition among major automakers is intensifying. On the other hand, many regions have included vehicle-road-cloud integration and intelligent connected vehicles in their 2025 government work reports, and the construction of vehicle-road-cloud integration is accelerating. A revolution in the transportation system is advancing rapidly, and a blueprint for future intelligent transportation is gradually becoming clear.

Is there a competitive relationship between single-vehicle intelligence, vehicle-road collaboration, and vehicle-road-cloud integration? Where are the business opportunities for vehicle-road-cloud integration? What is the key strategy for achieving "economies of scale" in autonomous driving? "Vehicle-Road-Cloud 50" interviewed several industry insiders and attempted to sort out and answer these questions.

The Relationship among the Three: From Individual to Galaxy, from Terminal to Global

Q: "Single-vehicle intelligence" and "vehicle-road collaboration" have always been regarded as two major technological directions for achieving autonomous driving. Currently, single-vehicle intelligence is leading in the commercialization process, while vehicle-road collaboration faces many challenges. In this context, what is the future prospect of "vehicle-road-cloud integration"?

A: There is no competitive relationship between single-vehicle intelligence, vehicle-road collaboration, and vehicle-road-cloud integration.

Single-vehicle intelligence means that a vehicle independently completes environmental perception, decision-making and planning, and vehicle control by relying on its own sensors, computing units, and algorithms to achieve autonomous driving functions.

The concept of vehicle-road collaboration is an exploration from "single vehicles" to the "relationship between vehicles and traffic." In essence, it does not aim to compete with single-vehicle intelligence because vehicles are only one part of the transportation system, not the whole. From the perspective of the "transportation system," for a long time, many problems caused by human-centered self-organized transportation have become more prominent against the background of rapid urbanization. Vehicle-road collaboration technology attempts to promote the intelligentization and automation of the transportation system by introducing more external information and control.

The intelligentization trend in the automotive industry is driving the transformation of vehicles from simple means of transportation to mobile intelligent spaces. In this process, the mobile Internet, mobile Internet of Things, and mobile energy network, corresponding to the intelligent cockpit, autonomous driving, and power system of vehicles respectively, will become important fulcrums for promoting this transformation and upgrading.

The industry generally compares the intelligent transformation of automobiles to the leap of mobile phones from feature phones to intelligent terminals. The development of the smartphone industry mainly revolves around two main lines: one is the intelligent upgrade of the device itself, that is, "smartphone intelligentization"; the other is the in-depth collaboration between cloud services, communication networks, and terminal devices, that is, "cloud-pipe-terminal integration."

Therefore, we can make such a judgment: To achieve the goal of "autonomous driving," two conditions are required. One is "vehicle intelligentization," and the other is "vehicle-road-cloud integration." Single-vehicle intelligence and vehicle-road-cloud integration are not alternative solutions but two indispensable components for fully realizing autonomous driving.

From the perspective of "cloud-pipe-terminal," it can be seen that the focus of single-vehicle intelligence development lies in strengthening the capabilities of the "terminal." The key to vehicle-road collaboration is to activate the bridging role of the "pipe." The breakthrough of vehicle-road-cloud integration lies in connecting the "cloud-pipe-terminal" to achieve global interaction. To some extent, the development relationship among the three essentially represents a paradigm shift in intelligent transportation from "terminal autonomy" to "global collaboration."

Connected Supervision: Providing Support for the Adoption of Intelligent Assisted Driving

Q: At this stage, how can vehicle-road-cloud integration further promote the adoption of intelligent assisted driving through technological collaboration and systematic optimization?

A: Currently, the competition in the automotive industry continues to intensify, and the focus of competition is shifting from the "electrification" in the first half to the "intelligentization" in the second half. Automakers and autonomous driving companies are all investing in the research and development of autonomous driving technologies and are making every effort to promote the adoption of intelligent assisted driving functions in vehicles.

However, we also need to see that the so-called "high-level intelligent driving" functions installed in personal passenger cars, such as urban NOA, highway NOA, and automatic valet parking (AVP), are actually only L2+ assisted driving. The marketing strategy of automakers that blurs the technological boundaries, combined with users' cognitive biases about the system's capabilities, may induce a large number of "moral hazard" driving behaviors. (Note: "Moral Hazard" is an economic term referring to the phenomenon that individuals or organizations tend to take more risky actions after risk transfer, leading to an increase in the probability of risk.)

On one hand, driven by the motivation to seize the market, automakers may exaggerate technological capabilities or provide vague publicity. Under information asymmetry, users may overtrust the system, thereby increasing the risk of accidents. For example, Tesla's "Autopilot" (literally "autonomous driving") and "FSD" (Full Self-Driving) have been criticized for their overly vague names, which may mislead users into thinking that the vehicle already has full autonomous driving capabilities and ignore the fact that it still requires driver monitoring and intervention.

On the other hand, assisted driving assumes part of the human driving responsibility through technology. Users may relax their vigilance due to their reliance on assisted driving and reduce their active monitoring of road conditions (such as frequent use of mobile phones), resulting in accident risks beyond the system's responsibility boundary. When an accident occurs, users also tend to attribute the responsibility to technological defects rather than their own negligence. For example, in many traffic accidents involving Tesla's Autopilot, there have been disputes between system misjudgment and driver responsibility.

This inevitably leads to a problem: the technological benefits are concentrated in a few entities, but the possible risks and costs, such as increased safety hazards, rising insurance premiums, and increased accident handling costs, are shared by the whole society. In fact, new energy vehicles are now facing high insurance premiums, difficulties in insuring, and frequent insurance rejection phenomena, which are due to high accident rates and difficulties in risk assessment.

For various assisted driving vehicles where the driving responsibility still belongs to human drivers, it is necessary to supervise the real-time status of these vehicles' intelligent systems through networking, and at the same time monitor and remind the vehicle owners. Its function is similar to that of the air traffic control system monitoring the flight status of airplanes and issuing dispatch instructions to pilots.

Vehicle-road-cloud integration can strengthen vehicle networking services. It is like adding an "offshore balancer" between vehicle owners and automakers. It not only helps reduce the risk of accidents but also facilitates accident tracing and responsibility determination. It is beneficial for protecting the interests of users and provides regulatory support for the large-scale adoption of intelligent assisted driving and the subsequent development of the industry.

New Regulatory Paradigm: Meeting the Needs of Large-scale Commercialization of Autonomous Driving

Q: With the coordinated evolution of automotive intelligentization and networking, the global automotive industry is accelerating the leap from L2 assisted driving to L4 autonomous driving. In the process of large-scale commercialization of autonomous driving, "safety" is a threshold that must be crossed and is also a core issue in the formulation of regulatory policies. What role will vehicle-road-cloud integration play in this?

A: "Vehicle-road-cloud integration" provides a feasible solution for the regulatory paradigm of autonomous driving. "Autonomous driving within the regulatory framework" can be regarded as "autonomous driving in the environment of vehicle-road-cloud integration." Providing policy and regulatory support for the commercialization and large-scale development of autonomous driving is the starting point of vehicle-road-cloud integration.

Waymo's autonomous driving system is currently the only L4 system that has achieved large-scale commercial operation and is running well. Dmitri Dolgov, the co-CEO of Waymo, said in an interview with the media last October that the main obstacles to the promotion of autonomous driving technology do not lie in capital investment or operation but in safety and public trust.

When the steering wheel is handed over to AI, can traffic rules still be enforced in the "old way"? The answer is obviously no. From "regulating people" to "regulating data," from "trusting drivers" to "trusting systems" - currently, a revolution in the regulatory paradigm is in full swing. In fact, governments around the world are facing an urgent issue: How to use a new regulatory paradigm to pave the way for the large-scale commercialization of autonomous driving?

What characteristics should the new regulatory paradigm have? It should cover at least the following five aspects.

First, data transparency: Breaking the "black box" of AI decision-making.

Autonomous driving is a new type of driving form driven by data, and its decision-making subject is an AI intelligent agent that processes multi-source data in real-time. Different from traditional regulation that emphasizes the restraint of human drivers' driving behaviors, the new regulation must focus on the supervision of the AI intelligent agent's decision-making chain, and the core lies in "data regulation." The new regulation requires that the autonomous driving system must achieve data transparency: all data usage must be authorized in accordance with the law and be subject to full-process and transparent supervision.

Second, social credibility: AI drivers "obtain licenses to work."

How to gain the trust of the general public for AI drivers is a concern that the new regulation must address. Under the traditional driving regulatory system, regulatory authorities have established a social credibility mechanism for human drivers through driving qualification recognition, such as driving training and driver's license examinations. The new regulatory paradigm must also evaluate and certify the AI intelligent agents performing autonomous driving tasks and provide "credit endorsement" for autonomous driving to gain social trust by granting them driving qualifications (equivalent to "driver's licenses").

Third, risk controllability: The cloud is always ready to "take over."

With the continuous advancement of vehicle automation, autonomous driving will go through a process of risk accumulation and outbreak. How to take over and control the vehicle when an autonomous driving risk occurs in a single vehicle will inevitably be an important issue in the new regulatory system. It can be envisioned that the new regulatory system may take over the "driving right" of abnormal autonomous driving vehicles through a "centralized" mobile communication network, and the cloud will activate "remote driving" to drive the vehicle to a safe area.

Fourth, liability determination: From "driver takes responsibility" to "enterprise pays the bill."

Autonomous driving poses a new challenge in the legal aspect: the issue of accident liability determination and liability attribution. After the popularization of autonomous driving, the self-organized traffic mode mainly based on the decentralized decision-making of human drivers will transform into a heterogeneously organized traffic mode mainly controlled by AI algorithms, and the subject of driving responsibility will also shift from natural persons to legal persons (companies that control AI intelligent agents). Therefore, the new regulation needs to establish a coordinated system of "macro-control + AI algorithm control" to meet the needs brought about by this change.

Fifth, overall safety: Strictly controlling the "spillover" of risks.

During the operation of autonomous driving vehicles, the required communication, maps, positioning, and data have strong local attributes. As a result, safety risks spill over to other fields outside transportation, such as geographic information security and cross-border data flow. The spatial dependence and social embeddedness demonstrated by autonomous driving technology during its implementation and promotion process have a strong conflict with the traditional governance framework, making its regional and social attribute issues increasingly prominent. Therefore, it is urgent to design an innovative regulatory system at the national level.

"Vehicle-road-cloud integration" is a cyber-physical system (CPS) for the large-scale commercialization of autonomous driving and provides a feasible solution for the regulatory paradigm of autonomous driving.

The vehicle-road-cloud integration system integrates multi-source data from vehicles, roads, and the cloud. Through the cloud control basic platform, it can achieve comprehensive supervision of the operating status and compliance of autonomous driving vehicles. This includes not only monitoring compliance with traffic regulations but also evaluating adaptability to different road conditions, determining the liability for accidents of autonomous driving vehicles, restricting the compliance boundary of vehicle information collection, and controlling the overall impact of autonomous driving vehicles on society, regions, and national security.

In summary, vehicle-road-cloud integration not only focuses on solving safety problems in specific areas but also provides basic support for the wide deployment and regulation of intelligent connected vehicles.

Vehicle-Road-Cloud Integration: Giving Birth to a New Blue Ocean of "Driving Commodities"

Q: How can vehicle-road-cloud integration accelerate the "commercialization" process of autonomous driving from the perspective of "value-driven"?

A: AI driving brings about a transfer of driving responsibility. To distinguish between autonomous driving and assisted driving, we cannot only start from a technical perspective but from the subject of driving responsibility: if the driving responsibility lies with the vehicle owner or human driver, no matter what technology is used, it is assisted driving; when the responsibility is borne by the system, enterprise, or operation platform, it is autonomous driving, which can drive its real "commercialization."

The business model of real autonomous driving is "operation-oriented," and the object of operation is the "driving commodity." We can understand this business form as a new type of "Didi": for example, if I have a vehicle driven by an autonomous driving system, that is, the system provides me with driving services, then the driving responsibility belongs to the system owner (automaker or operator), and "driving" becomes a commodity. The "commercialization of driving" is the real change that autonomous driving brings to the automotive and transportation industries.

The generation of the "driving commodity" is achieved by inputting users' travel purposes and urban traffic information into the autonomous driving system and then realizing it through AI driving. It can be seen that this business form is essentially a "digital economy," a commodity produced through the comprehensive utilization of the entire urban spatial data resources.

These data resources are not only collected and provided by on-vehicle sensors but also include a lot of other information, such as traffic equipment information, traffic flow information, static space (such as parking lots) information, and climate information. Therefore, all kinds of facilities invested and built by cities and the various data produced can realize value through the "commercialization of driving."

From the perspective of local governments, this is the "digital finance," which regards the entire city as the "production workshop" of the digital economy. This "digital finance" is the huge economic effect that vehicle-road-cloud integration brings to local governments.

In the chain of driving commercialization, automakers, as "vehicle suppliers," pre-embed intelligent modules; network operators set up low-latency communication pipelines; government data platforms act as "raw material suppliers"; and financial institutions develop dynamic insurance products based on driving behavior data.

The autonomous driving operation platforms that connect all links in the chain, that is, value-added service operators (such as Robotaxi, Robobus, Robotruck, Robovan platforms), no longer dispatch drivers like Didi but directly dispatch AI to achieve real-time matching between transportation capacity supply and travel demand. As the core link in this ecosystem, value-added service operators have also become the key force driving the effective linkage and integrated development of the entire system.

Figure: The "Butterfly Model" of the Economies of Scale of "Vehicle-Road-Cloud Integration"

In general, the development of vehicle-road-cloud integration will not only integrate a super-large and cross-field upstream and downstream industrial chain, covering intelligent road investment and construction, application of mobile communication technology, cloud computing and big data processing, and application of artificial intelligence technology but also drive the effective collaboration among different market players to jointly explore the new blue ocean of "driving commodities."

Accelerating Vehicle-Road-Cloud Integration: Commodity Thinking and Value Realization

Q: Currently, the development of vehicle-road-cloud integration faces some confusion and doubts, which pose challenges to capital introduction and investors' patience. Where is the breakthrough point?

A: This question can be asked more intuitively: Vehicle-road-cloud integration itself has huge investment and development space. How to activate this space?

The answer is that the industrial development needs to shift from technical thinking and product thinking to value thinking and commodity thinking.

First, create enough space to attract and accommodate more players and capital to enter.

For example, to achieve this, the past vehicle networking networking model needs to be upgraded. Academician Wu Hequan of the Chinese Academy of Engineering mentioned at the "Annual Development Forum of Vehicle-Road-Cloud 50" last October that currently, a national unified vehicle networking investment and operation entity needs to be established. It is recommended that public