1. Introduction
Intelligent vehicles represent the strategic high ground in the future form and key technologies of automotive products. Autonomous driving, as a key feature of intelligent vehicles, has become an important factor in determining the future of the automobile industry. Intelligent vehicles can either support or fully take over the driving process, offering a more comfortable and convenient experience for drivers. They also enhance traffic efficiency [
1,
2,
3,
4], contribute to environmental sustainability by cutting down on energy use and carbon emissions [
5,
6,
7], and improve traffic safety [
8,
9]. The development of autonomous driving technology, though in its infancy, is progressing rapidly. The combined driving assistance function, which represents primary autonomous driving, is gradually becoming a standard configuration for new car models released by car manufacturers [
10]. Intermediate and advanced autonomous driving are still mainly in the testing and regional demonstration stages. Although the commercialization process in simple scenarios such as closed parks and trunk logistics is accelerating, there are still many bottlenecks. Addressing the technical hurdles inherent to a higher level of autonomous driving as well as the economics associated with large-scale application, intelligent connected vehicles (ICVs) are globally acknowledged as the pivotal path for future development, which is powered by the latest generation of information and communication technology [
11,
12]. More and more countries and companies are gradually turning to the vehicle-road collaborative intelligent route [
13].
Under a collaborative intelligent technical route, ICVs interact with roadside intelligent infrastructure to enhance their environmental perception and make optimized decisions and driving behaviors. Therefore, the intelligent infrastructure should match the functions and technical requirements of ICVs and complete the intelligent upgrade [
14]. Intelligent infrastructure can provide vehicles with a wider range of perception data, expand the vehicle’s perception range and capabilities, improve safety, and solve a series of “perception long-tail” problems such as perception occlusion and extreme weather [
15]. Intelligent infrastructure can also reduce the performance requirements for on-board sensing equipment, thereby reducing the on-board computing requirements related to perception. By providing global route planning and decision optimization for ICVs, collaborative intelligence can meet the challenges of multi-vehicle autonomous driving decision-making games in mixed traffic scenarios, improving traffic efficiency while reducing the computing power related to planning and decision. Meanwhile, the collaborative intelligence route means adding the cost of roadside intelligent infrastructure deployment, energy, and maintenance while significantly reducing the relevant cost on the vehicle side. How to combine roadside intelligent infrastructure with vehicle-side intelligent configurations to meet the needs of different levels of autonomous driving and how to deploy roadside intelligent infrastructure cost-effectively at different stages of technological development are issues that warrant the attention of governments and related enterprises. This study is based on this background.
Some related studies and literature are listed in
Table 1. Currently, most related research focuses on the analysis and optimization of the specific functions and technologies involved in the collaborative intelligence system [
16,
17,
18,
19,
20]. Chang conducted a detailed study on the functional distribution from the perspective of intelligent connected cloud control systems and verified it through a real vehicle test platform [
16]. Wang et al. outlined the application scenario requirements for high-level autonomous driving vehicles, clarifying the overall architecture, system functions, and performance requirements of the vehicle-road collaborative system [
17]. Sánchez et al. proposed an intelligent transportation plan using a collaborative driving strategy where each vehicle infers traffic conditions based on its own status and information shared by surrounding vehicles and makes decisions based on real-time inferred traffic levels using fuzzy logic methods [
18]. Zheng et al. established a two-stage optimization model for optimizing vehicle trajectories and signal timing at intersections under mixed traffic conditions to reduce vehicle delays and fleet carbon emissions [
19]. Other scholars have conducted research on the layout of certain hardware facilities involved in the upgrade of transportation infrastructure, such as sensors [
16,
21,
22,
23,
24], roadside units [
25,
26,
27,
28], communication infrastructure [
14,
29,
30], and edge computing platforms [
31,
32]. These studies generally establish mathematical optimization models for the layout of certain hardware facilities, aiming to find the most cost-effective, energy-efficient, or optimal performance layout plan. However, its combination with the functional requirements of ICVs is not strong. The China Highway Society has proposed a classification for vehicle-road collaborative autonomous driving systems, believing that the same level of collaborative systems can be achieved through different levels of ICVs and intelligent roads [
33], but it is more limited to qualitative analysis. Liu et al. designed the deployment methods and upgrade routes of intelligent infrastructures by sorting out the demands of ICVs [
14]. In summary, there is little research on the redistribution of vehicle and roadside intelligent capabilities corresponding to satisfying different levels of autonomous driving under the background of collaborative intelligence. In actual industrial practice, there is still a lack of theoretical basis for the prerequisite of function-cost effectiveness for the intelligent transformation of road infrastructure.
The purpose of this study is to clarify the substitution logic of intelligent infrastructure for on-board intelligent configuration, re-coordinate the distribution of vehicle-road intelligent capabilities, and analyze the combination and distribution of vehicle-road intelligent configurations for different levels of autonomous driving. The intelligent-related costs of the vehicle side and the roadside under different schemes are also quantitatively analyzed. Different road types have different scene characteristics and require different intelligent infrastructure schemes, combined with the various usage intensities. This study constructs a vehicle-road intelligent cost-sharing model to guarantee the deployment planning of intelligent infrastructure to be carried out under the premise of meeting function-cost effectiveness, avoiding the waste of road intelligent infrastructure as a social public resource, and achieving a higher level of autonomous driving with the lowest total cost. Beijing has 7.128 million motor vehicles [
35], ranking first among Chinese cities. Urban transportation’s safety, congestion, environmental, and energy issues are prominent [
36], and the demand for the intelligent transformation of transportation infrastructure is urgent. This study selects Beijing for case analysis, and the findings provide a reference for city managers when making decisions about deploying intelligent infrastructure.
This study is organized as follows: Firstly, it combs typical vehicle-side and roadside perception schemes. Thereafter, the quantification sub-model for perception capabilities and the matching sub-model for computing power are designed to identify combinations of vehicle-side schemes and roadside schemes that enable different levels of autonomous driving functions. Next, the cost-sharing model for vehicle-road intelligence is designed to calculate the per-mile cost corresponding to different vehicle intelligent schemes over the vehicle’s lifecycle as well as the annual average cost of intelligent infrastructure for different roadside schemes over their lifecycle. Subsequently, by integrating the average traffic flow data of different road types in Beijing, the study determines the function-cost effectiveness reflected at the vehicle-side after different levels of roadside intelligent transformation at various stages. Finally, implications for governments and related enterprises in developing vehicle-road collaborative intelligent systems are drawn.
4. Conclusions and Policy Suggestions
Contributions: This study, based on the architecture of the vehicle-road collaborative intelligent system, has constructed a model for the allocation of vehicle-road intelligent capabilities and a cost-sharing model. It has decomposed the intelligent capabilities of the vehicle side and roadside in terms of perception and computation, hardware and software, clarifying the mechanism by which vehicle-side intelligent capabilities are transferred to the roadside within the collaborative intelligent system. The study has built a quantification sub-model for perception capabilities and a matching sub-model for computing power, analyzing the allocation plans for vehicle-road intelligent capabilities to achieve different levels of autonomous driving functions. Different road types have different scenario characteristics. This study constructs a vehicle-road intelligent cost-sharing model and selects Beijing as a case study. Combining statistical data on the mileage scale and usage intensity of different highways and urban roads, as well as forecast data on vehicle ownership and the penetration rate of intelligent vehicles, it calculates the function-cost effectiveness of roadside intelligent schemes for different road types and analyzes the selection of roadside intelligent schemes and the corresponding deployment order for different road types in Beijing at different stages of development.
From the perspective of the allocation of vehicle-road intelligent capabilities, the research findings indicate that primary intelligent infrastructure primarily serves the intermediate autonomous driving of ICVs, while intermediate and advanced intelligent infrastructure can better cater to the advanced autonomous driving of ICVs, significantly reducing the cost of perception configurations and computing power on the vehicle side. In terms of the impact of roadside intelligent infrastructure deployment on the intelligent cost of vehicles, the primary intelligent infrastructure can reduce the lifecycle cost of the vehicle-side intelligent scheme for intermediate autonomous driving from ¥65,301 to ¥37,703. Advanced intelligent infrastructure can reduce the lifecycle cost of advanced autonomous driving from ¥126,938 to ¥42,180. In the future, with the deployment and utilization of urban cloud platforms, through collaborative planning and decision-making at the city road network level, the lifecycle cost for advanced autonomous driving on the vehicle side will further decrease to ¥40,687. The per-vehicle per-kilometer allocation cost will be reduced from ¥0.205/km to ¥0.197/km.
From the perspective of the function-cost effectiveness of intelligent infrastructure for the autonomous driving functions of ICVs, urban roads are generally superior to highways. The order of intelligent upgrading for urban roads is urban expressways, urban main roads, and urban secondary roads, while for highways, it is expressways, class-1 highways, class-2 highways, and class-3 highways. Among them, urban expressways, with the smallest mileage scale and the highest “function-cost effectiveness,” can be prioritized for intelligent upgrading on urban roads, directly selecting advanced intelligent infrastructure schemes to support the demonstration and implementation of advanced autonomous driving. In contrast, class-3 highways will only gradually become function-cost effective after deploying advanced intelligent infrastructure after 2035 and can be considered the last objects for intelligent infrastructure upgrading in the urban road network.
From the perspective of the external support service, the urban communication operator is also the focus of this study in terms of costs and expected returns. By integrating the performance and costs of communication infrastructure with mainstream deployment schemes, this study calculates the deployment costs, power consumption costs, and maintenance costs associated with communication infrastructure. The annualized cost for communication operators to achieve 5G network coverage within the road network of Beijing is ¥2.938 billion. Under the baseline scenario, the 5G communication service fees obtained by communication operators from users will cover the annualized costs of the related infrastructure by 2028. The annualized allocation cost per vehicle for communication infrastructure will decrease from ¥6467 in 2023 to ¥314 by 2050. It can be said that as the support role of low-latency, high-bandwidth communication networks in intermediated and advanced autonomous driving functions of ICVs becomes increasingly evident, intelligent transportation will become an important application scenario for 5G networks in the future, driving further expansion of the market size of the communication industry.
Policy suggestions and industry participant recommendations: Based on national top-level planning, the government should improve the standard system construction, accelerate the deployment of the C-V2X network environment, and promote the demonstration of vehicle-road collaborative applications, preparing for the standardization and large-scale deployment of intelligent infrastructure. Financial support from the government for the construction of intelligent infrastructure will share the costs of vehicle intelligence with vehicle manufacturers, addressing the industry’s profitability issues while reducing the costs of vehicle intelligence paid by users. To avoid the waste of intelligent infrastructure resources in the early stages, intelligent deployment and technological iterations can be prioritized on urban expressways and motorways. All industry participants in the collaborative intelligent system should leverage their existing business foundations to focus on the field of vehicle-road collaboration, build on their technological strengths, collaborate and divide tasks, actively participate in the construction of the intelligent connected vehicle industry ecosystem, and continuously accelerate the formation of a complete industrial system. This approach aims to implement higher-level autonomous driving solutions at a lower societal cost. Internet technology companies, with rich experience in big data and software algorithms, need to build perception networks, collaborative planning, and decision-making networks oriented towards the technological pathways of collaborative intelligence. Intelligent component enterprises should actively promote the landing of products with higher performance-cost ratios through technological innovation and actively develop roadside equipment adapted to different scenario characteristics, addressing the issue of low service life. With the support of intelligent road infrastructure, automotive companies should actively seek multi-party cooperation to achieve technological integration, accelerate the deployment of high-performance communication modules, and provide the terminal market with ICVs that are lower in cost and better in functional performance. Telecommunications companies, having accumulated rich experience in networking technology, should participate with the government in the construction of urban network environments, continuously expanding the application scenarios and business models of 5G-V2X.
This study has the following limitations. First, the “vehicle intelligence” scenario and the “collaborative intelligence” scenario are both based on the penetration rate forecasts of intelligent vehicles in the ‘Intelligent Connected Vehicle Technology Roadmap 2.0.’ without considering the potential increase in the penetration rate under the “collaborative intelligence” scenario. The deployment and use of intelligent infrastructure will reduce the cost of vehicle-side intelligence and significantly enhance the value of ICVs. However, the willingness of users to purchase intelligent vehicles is influenced by various factors, including the incremental costs of intelligence, the socio-economic situation, and the maturity of technology, which in turn affects the future market penetration rate. In our following research, we intend to quantify the correlation between the incremental costs associated with intelligence and the market penetration rate of intelligent vehicles. Second, in addition to achieving higher function-cost effectiveness in autonomous driving, collaborative intelligence schemes will also generate significant social and public benefits such as safety, efficiency, energy conservation, and emission reduction. Assessing and quantifying these benefits will represent a significant and valuable component of our forthcoming research endeavors.