1 Introduction

Contemporary transport systems have brought great convenience to the public, but they also bring about many problems, such as road congestion, air pollution, and traffic accidents. To solve these problems, a transition toward autonomous mobility is underway (Hopkins & Schwanen, 2018). Connected and autonomous vehicles (CAVs), resulting from the synergy between artificial intelligence, robotics, automotive designs, and information technologies, have the potential to be the most robust intervention in the history of road mobility by enabling the cars to interact with their surrounding environment and make decisions by themselves (Nikitas et al., 2017). Not accounting for any changes in demand, the application of CAVs could significantly reduce the frequency of traffic accidents and traffic congestion, as autonomous vehicles will make fewer errors than human drivers and increase the effective road capacity. Parking patterns can also be changed with the arrival of CAVs, as they can self-park in less expensive parking areas (Clements & Kockelman, 2017), although that would probably require additional driving distances. In addition, if CAVs become electric, that could mitigate the air pollution caused by automobile exhaust emissions. In the impending automated mobility era, CAVs will dramatically change the landscape of cities, the regulatory environment for transportation, and the operating modes of society.

Therefore, to successfully promote the use of CAVs, a number of different fields, including transportation infrastructure, policy and legislation, technology and innovation, and consumer acceptance, have to be well prepared. In this paper, we focus on infrastructure. On the one hand, CAVs need to be integrated into the existing transport network by improving the current infrastructure. On the other hand, new infrastructure must be designed and deployed to satisfy the unique requirements of CAVs. Thus, a framework including the fundamental CAV-related infrastructure design principles is necessary to guide urban planning in the next decades. In this study, we develop such a framework integrating different aspects sparsely addressed in the existing literature. We compile and combine a number of principles covering different aspects of CAV infrastructure, to provide unified guidelines for the planning, design, deployment, and evaluation of such infrastructure. In addition, we illustrate how to apply these principles to a specific city by conducting a case study in Oxford, UK. We hope the proposed design scheme can provide some references for the development of future automated transportation systems in Oxford and other cities around the world.

The rest of the paper is organized as follows. Section 2 presents a literature review of different infrastructure related studies within the context of CAVs. Section 3 introduces a new framework to plan, design, deploy, and evaluate infrastructure-related investments, based on the fundamental design principles associated with CAV-infrastructure. Section 4 illustrates such framework using the city of Oxford as a case study. Section 5 discusses the main insights from this application. Section 6 offers some concluding remarks and discusses the implications of the study.

2 Literature review

Infrastructure related to the operation of CAVs includes both physical and virtual aspects (Liu et al., 2019). Below, we discuss each of them separately. We then discuss some additional literature related to the design of such infrastructure.

2.1 Physical infrastructure for CAVs

Previous studies have classified CAV-related physical infrastructure into five subcategories, namely: traffic signs and road markings, road network, parking facilities, safe harbors, and charging facilities, as listed in Table 1. This article delves into the primary advantages and disadvantages affiliated with each of these categories concerning CAVs. The ensuing discussion takes into account the limitations of existing configurations, while also weighing the advantages and potential concerns posed by prospective restructured patterns.

Table 1 Examples of previous studies on CAV-related infrastructure
Full size table

2.1.1 Traffic signs and road markings

Traffic signs and road markings are foundational components for upholding the efficient and secure flow of traffic (Babić et al., 2022; Babić & Brijs, 2021; Burghardt et al., 2020). These elements offer a direct and cost-effective mechanism for ensuring the systematic operation of road networks. They serve as indispensable instruments for conveying vital road-related information to both drivers and pedestrians, facilitating unambiguous communication about road conditions and the surrounding environment. Notably, prominently visible traffic signs and road markings play a critical role in guiding CAVs during navigation, positioning, and parking activities. It is essential that the visibility of these elements meets the criteria of both CAVs and human drivers, accounting for daytime contrast and nighttime retro-reflectivity (Burghardt et al., 2021).

Presently, road markings and signs are primarily designed with human drivers in mind. However, scenarios that may pose minimal challenges for human drivers could potentially befuddle CAVs. For instance, the presence of a combination of old and newly painted stripes might not be as clearly interpreted by CAVs. To address this, efforts are underway to devise road markings and signs better suited for CAVs, encompassing considerations such as striping, visibility, and material composition (Singh & Islam, 2020; Burghardt et al., 2021; Moreno-Navarro et al., 2019). To ensure seamless interpretation by CAVs, a set of comprehensive standards should be established and widely implemented for the design and deployment of optimal road markings and signs. Additionally, stakeholders must consistently assess and maintain these road markings and signs to guarantee their functionality, particularly concerning CAVs (Xu et al., 2021).

2.1.2 Road network

There are two main ways to incorporate CAVs into the traffic road network. The first is an independent right-of-way model, which means setting up dedicated lanes for CAVs. The second is a mixed right-of-way model, in which CAVs are mixed with conventional vehicles on original roads. With the independent right-of-way model, CAVs can operate more safely and efficiently, minimizing the interactions with human-driven vehicles. The width of the lane can be reduced compared to existing lanes, as CAVs can track and stick to their road lanes more effectively than humans. In fact, CAVs’ wheels are hardly expected to wander laterally unless malfunctioning (Zhou et al., 2019). In addition, CAVs can increase the capacity of roads due to shorter headways and faster reaction times (Lu et al., 2020). As penetration rate increases, this could then lead to a reduction in the necessary number of lanes to move the same number of vehicles. The saved road space can then be allocated to other more sustainable transport modes, such as pedestrians or cyclists, encouraging more people to walk or bike (Mead et al., 2014; Menendez & Ambühl, 2022). The disadvantages of this independent right-of-way model include, taking up limited road resources especially when the low CAV penetration rates do not fill the dedicated lanes, potentially increasing congestion in the other lanes if some are taken away, and causing confusions to traffic users. In contrast, the mixed right-of-way model should not cause significant impact on current road users as the traffic network could be kept pretty much the same. However, it may hinder the benefits CAVs bring on traffic performance, such as increased road capacity and saved road space. Li et al. (2020) found that appropriate road-of-way reallocation strategies can further increase road capacity compared to mixed strategies, although this depends on the penetration rates and overall demand. For instance, Ye and Yamamoto (2018) demonstrated that the benefit of setting dedicated lanes for CAVs can only be obtained when the vehicle density is medium.

2.1.3 Parking facilities

As the penetration rate of CAVs increases, the demand for very centric and expensive parking areas can be reduced. CAVs can drive themselves to less expensive parking areas farther away (Millard-Ball, 2019). However, this could contribute to a higher vehicle-distance driven, with empty vehicles driving all over the network and potentially increasing congestion. (Zhao et al., 2021) presented a centralized parking dispatch approach to guide floating autonomous vehicles to find optimal parking lots. Those can be built and operated away from the city center by adjusting parking fees (Levin et al., 2020). Several studies have been conducted to explore the design and management of CAV parking lots (Liu, 2018; Nourinejad et al., 2018; Wang et al., 2021). The saved space in central areas could then be reused for residential or commercial purposes. In addition, banning on-street parking would leave more space for pedestrians and cyclists (Jakob & Menendez, 2019), which would again, encourage more sustainable mobility practices (Guan and Forsyth, 2020; Guan et al., 2020). Special zones can allow CAVs to pick up and drop off their passengers without interrupting traffic flow. Some on-street spaces can be converted into safe harbors.

2.1.4 Safe harbors

Safe harbor areas are essential for incorporating CAVs into the transportation system. When CAVs cannot operate safely in the surrounding environment or meet any malfunctioning, simply stopping on the road will interrupt the traffic flow, leading to a series of traffic accidents. It is important then for CAVs to have specific areas where they can wait for rescue when needed. Safe harbors have been listed as required infrastructure to accommodate CAVs by several researchers and organizations (Liu et al., 2019; NASEM, 2015). The design of such places is still under research. Khan et al. (2022) pointed out that CAVs should have priority in using safe harbors. In addition, these safe harbors must be documented and mapped so that CAVs can identify and navigate to these areas automatically (Othman, 2021).

2.1.5 Charging facilities

With the expectation that most CAVs (if not all) will be powered by electricity, charging facilities are one of the most crucial pieces of physical infrastructure needed (Chen et al., 2023). Charging stations could provide charge through a cable, or potentially the ability to swap batteries (Das et al., 2020; Sachan et al., 2020). A more ambitious idea to meet the charging demand of electric vehicles is to develop wireless charging. This could happen in a stationary or a dynamic manner (Soares & Wang, 2022; Chen et al., 2017). The stationary charging is achieved by vehicles stopping on a charging pad placed on a parking space. The dynamic charging is achieved by driving on a road with an embedded charging module. Dynamic charging could significantly solve driving range anxiety issues and save users’ time (Duarte et al., 2021; Tan et al., 2022). This technology has been prioritized worldwide, especially in the United Kingdom (UK), Germany and South Korea (Machura & Li, 2019).

2.2 Virtual infrastructure for CAVs

Similar to the categorization of physical infrastructure, past research has classified CAV-related virtual infrastructure into four subcategories: intelligent transport systems, digital networks, vehicular clouds, and traffic control, as also outlined in Table 1. The following section provides an analysis of the advantages and disadvantages associated with these virtual infrastructure components.

2.2.1 Intelligent transport systems

In terms of virtual infrastructure, the smooth operation of CAVs partly depends on the communications between vehicles, infrastructure, and people. Intelligent Transport Systems (ITS) can help achieve these intercommunications. ITS is the synergy of sensors, analyzers, controllers and communication systems, which can improve safety and efficiency of all components in the transport system, including vehicles, pedestrians, and infrastructure (John et al., 2019; El Hamdani et al., 2020). The applications of ITS include, but are not limited to, responding promptly to traffic incidents, guiding vehicles to empty parking spaces, and adjusting speed limits and signal timings based on real-time conditions (Muthuramalingam et al., 2019). ITS can be further improved by leveraging the Internet of Things (IoT), wherein a large number of physical devices are connected to the Internet through the aid of new sensors, computing, and networking technologies (Bello & Zeadally, 2019). The Internet of Everything (IoE) is the advanced development of IoT, connecting heterogeneous things, process, data, and people (Miraz et al., 2015).

2.2.2 Digital networks

The massive number of connected devices and huge amount of data within the ITS require the support of stable and high-speed digital networks. 5G (the fifth generation of mobile technology) will play an important role in guaranteeing the fast interconnections and security within ITS (Storck & Duarte-Figueiredo, 2020). Compared to 4G, 5G will support 10–100 times higher data rates and number of connected devices (Gohar & Nencioni, 2021). It will ensure continuity, lower latency, and ubiquity of the network (Guevara & Auat Cheein, 2020). 5G is under development in many countries around the world, with China, South Korea, the United States (US), some countries in Europe and many others vying for the leadership (Oughton & Russell, 2020). In addition, a more advanced wireless communication, the six-generation system (6G), is expected to be implemented around 2030 (Chowdhury et al., 2020; Oinas-Kukkonen et al., 2021). Ubiquitous Wi-Fi is also useful in the operation of ITS (Adegoke et al., 2019). Dedicated Short-Range Communications (DSRC) is a Wi-Fi derivative technology used to meet specialized requirements for secure, low latency, and communications. For instance, DSRC with 75 MHz of wireless spectrum (5.850–5.925 GHz) is allocated to support ITS in the US (USDOT, 2015).

2.2.3 Vehicular clouds

As an emerging technology, vehicular cloud computing stands out as a promising solution for sustaining and advancing ITS. This innovative approach involves the pooling of vehicles’ resources, encompassing internet connectivity, storage, and computational capabilities, to create a shared cloud environment (Ahmad et al., 2017; Lin et al., 2018). Such a vehicular cloud possesses the potential to address various aspects of the traffic ecosystem. Initial applications encompass the management of parking facilities, with possibilities of converting parking lots, airports, and shopping malls into functional data centers. This technology can be further developed to optimize traffic signals, implement dynamic traffic management strategies, alleviate recurring congestion issues, and facilitate effective evacuation procedures (Whaiduzzaman et al., 2014; Boukerche & De Grande, 2018; Kang et al., 2015).

However, the implementation of vehicular clouds presents a challenge in the form of the dynamic nature of the cloud’s composition. Vehicles within a micro vehicular cloud continuously join and leave the network, posing a requirement for an adept handling of resource fluctuations (Coutinho & Boukerche, 2019). A mature vehicular cloud must effectively manage these shifts in resource availability without disrupting ongoing tasks and operations.

2.2.4 Traffic control

Traffic management is essential to maintain smooth and efficient traffic flow, especially over crossings and junctions. Intersections under the control of signals will be easier to deal with for CAVs than those without. Traffic signals can also assure the safety of pedestrians when they cross intersections. At the initial stages of CAVs penetration, pedestrians’ sense of comfort may be challenged out of trust issues towards the new technology, especially when they cross unsignalized intersections, as they won’t be able to communicate as they do with human drivers through eye contact, multiple facial expressions, and hand gestures (Deb et al., 2018). Thus, traffic signals should initially be established over intersections with no existing signal, so that CAVs can safely interact with pedestrians (Yin et al., 2021). In addition, signal timings should be properly set to avoid unnecessary delays and energy consumption. Information from CAVs could be used to develop more complex traffic management strategies that optimize traffic performance both at the intersection level and throughout the network (Yang et al., 2018). The introduction of CAVs would also enable signal timings and phase plans to adapt to current traffic conditions, thereby increasing the efficiency of intersections, and/or providing additional priority to more sustainable modes such as public transport (Yang et al., 2019a; Yang et al., 2019b). CAVs can also change their speed and acceleration rates according to signal timing to avoid waiting at intersections (Yang et al., 2016; Guo et al., 2019; Yu et al., 2019; Niroumand et al., 2020; Liang et al., 2019; Qi et al., 2020). A more ambitious idea is to develop signal-head-free intersections, where CAVs cross the intersections smoothly by cooperating with other CAVs and the infrastructure (Mirheli et al., 2018). However, this approach can only be achieved in a fully CAV environment, and can still be problematic for pedestrians.

While we initially present physical and virtual infrastructures as distinct components, it’s essential to underscore that these infrastructures are intricately interconnected, collaborating to ensure the seamless, secure, and efficient functioning of urban transportation systems. The operational efficacy of physical infrastructure for CAVs hinges on the presence of corresponding virtual infrastructure. To illustrate, CAVs leverage Intelligent Transportation Systems (ITS) to identify suitable parking areas and charging stations. Simultaneously, data regarding the status of physical infrastructure can be fed into ITS and traffic control systems to optimize CAV operations within the urban traffic ecosystem. This holistic network operates through the transfer, storage, and analysis of substantial data volumes encompassing infrastructures and road users. However, it’s important to acknowledge that this integrated system, reliant on extensive data and digital networks, is susceptible to privacy breaches and cybersecurity threats, as highlighted by Nikitas et al. (2022). Cyberattacks can potentially compromise not only the virtual infrastructure but also have cascading effects on the utilization of physical infrastructure.

2.3 Designing CAV-related urban infrastructure

The assessment of existing and newly developed urban infrastructure for compatibility with CAVs has been widely conducted through real-world trials across various global locations (Dowling & McGuirk, 2022). A prime example is Beijing, China, where approximately 278 roads spanning 6 districts with a total length of 1,028 km have been designated for CAV testing, encompassing urban, rural, and highway settings (http://www.mzone.site/index.php/indexen/index.html). These trials encompass various infrastructure components such as traffic signs and road markings, mixed right-of-way models, Vehicle-to-Everything (V2X) systems, digital networks, and traffic control. Moreover, specific simulated environments have been established for CAV testing, including locations like Chang’an University in China, Mcity in Michigan, and GoMentum in California (https://gomentumstation.net/av-testing-services/), which also evaluate infrastructure aspects like traffic signs, road markings, mixed right-of-way models, Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and V2X systems, digital networks, and traffic control (Yang et al., 2021).

Large-scale collaborative projects have also emerged as avenues for infrastructure enhancement, involving stakeholders from various sectors. In the UK, numerous projects funded by the government have engaged over 200 participants from automotive manufacturers, research institutions, universities, and tech enterprises since 2014. Initiatives like GATEway (https://trl.co.uk/projects/gateway-project/), VENTURER (https://www.venturer-cars.com/), and FLOURISH (http://www.flourishmobility.com/about-flourish) prioritize the interaction between CAVs and roadside infrastructure. These endeavors primarily aim to determine the integration potential of CAVs within the existing transport network, hence focusing on essential infrastructure components such as traffic signs, road markings, V2X systems, digital networks, and traffic control. Notably, certain vital elements like safe harbors and advanced structures like vehicular clouds remain largely unexplored in these trials. Nevertheless, the data and driving patterns garnered from these trials hold potential for optimizing present road infrastructure and formulating more sophisticated infrastructure solutions. Generally, the assessment of optimized or novel infrastructure is carried out through meticulous experiment design (Laporte et al., 2019; Levy & Haddad, 2022; Labi et al., 2023).

While the quantity of studies related to CAV-supporting infrastructure design has surged, a limited number of them delve into the integration and deployment of these infrastructure components within specific cities. For instance, (Manivasakan et al., 2021) presented a conceptual road network design for Melbourne, Australia, considering road facilities, signage and markings, parking and service stations, operational management, communication networks, and maintenance considerations. The array of infrastructure required by CAVs exhibits diversity, each carrying multiple variables that demand attention. Without a comprehensive design framework, urban planning efforts could lack organization, consume time inefficiently, and potentially overlook critical details (Guan et al., 2023). Consequently, the establishment of universally applicable CAV-related infrastructure design principles holds paramount importance when preparing cities for the integration of CAVs. Nonetheless, it’s crucial to adapt these principles to match the distinctive characteristics of individual cities. Our paper aims to furnish a CAV-related infrastructure design framework and subsequently apply it to the context of Oxford.

3 Design principles for CAV-related urban infrastructure

The foundational design principles for CAV-related infrastructure in future scenarios draw upon widely accepted standards from existing literature (Barnes & Swan, 2018; Dowling & McGuirk, 2022; Mitteregger et al., 2021; Liu et al., 2019). We have organized these principles into physical and virtual infrastructure for two distinct phases (Jia et al., 2021), as outlined in Table 2. We have maintained a consistent sub-category structure with the previous section. For instance, the physical infrastructure is divided into five key principles: traffic signs and road markings, road network, parking facilities, safe harbors, and charging facilities. To provide further clarity, each subcategory encompasses crucial principles as identified in existing literature. As an example, the principle related to traffic signs and road markings (P1) encompasses activities such as updating traffic signs (Babić & Brijs, 2021; Babić et al., 2022), maintaining road markings (Burghardt et al., 2020; Xu et al., 2021), and establishing unified standards and signals for various users (Singh & Islam, 2020; Burghardt et al., 2021; Moreno-Navarro et al., 2019).

Table 2 Design principles for CAV-related infrastructure in future scenarios
Full size table

We have partitioned future scenarios into two distinct phases due to variations in the requirements for CAV-related infrastructure, which are contingent on CAV automation levels and CAV penetration rates (Jia et al., 2021; Mitteregger et al., 2021). Automation levels span from 0 to 5 (SAE International, 2019), with level 5 denoting fully autonomous vehicles. High/full driving automation necessitates more stringent requirements for certain CAV-related infrastructure while enabling smarter utilization of others compared to partial driving automation. Additionally, the CAV penetration rate represents a critical factor influencing the design of CAV-related infrastructure. The allocation of resources for CAVs must accommodate the current volume of CAVs and anticipate the projected short-term growth. Nevertheless, allocating excessive resources to CAVs can lead to resource wastage and inconvenience for traditional vehicles. Consequently, resource allocation for transportation relies heavily on both the current and accurately projected CAV penetration rates.

The first phase characterizes a traffic composition scenario where partial driving automation predominates among CAV types, and CAV penetration rates remain low. The second phase reflects a scenario in which high/full driving automation dominates the CAV market. However, it’s important to note that achieving the second phase doesn’t necessarily imply the complete replacement of human-driven vehicles or lower automation CAVs, as this process may be protracted (unless regulations expedite it). In this phase, CAV penetration rates are high, exceeding 50%. It’s worth highlighting that there is no strict demarcation line between these two phases, and effective urban planning can facilitate a seamless transition from phase 1 to phase 2.

3.1 Design principles for the first phase

During the initial phase, a focused set of design principles is proposed to facilitate the integration of CAVs while considering both their operational needs and the existing urban environment.

To ensure effective communication with both CAVs and human drivers, the renovation of current traffic signs and road markings is imperative. Existing designs, tailored for human perception, should be reevaluated to address potential confusion for CAVs. High visibility, essential for CAV recognition, demands frequent maintenance. New signs and markings introduced in this phase should also be conceived with a forward-looking perspective to remain pertinent in the subsequent phase, aiding sustainability. Given the anticipated low penetration rate of CAVs, preserving the current road networks is recommended for the first phase. Dedicated lanes for CAVs could lead to resource inefficiency and congestion in other lanes.

Basic automation of CAVs characterizes this phase, with drivers prepared to assume complex driving tasks. Integration of CAV parking into existing lots and the addition of EV charging stations are advised. Safe harbors become crucial even with driver presence in CAVs. These serve as waiting zones for driver intervention or as locations for rescue during breakdowns. Charging facilities, additional batteries, and charging points should be strategically situated at bus stops, pick-up & drop-off zones, and safe harbors.

Leveraging advancing technology and digital networks, ITS can be bolstered to enable seamless communication between vehicles and infrastructure. The maturation of the 4G data platform and widespread availability of 5G, along with the allocation of DSRC, will contribute to enhanced connectivity. This robust network is pivotal to ensuring safe, efficient urban transportation. Vehicular resources can be harnessed for parking and traffic management, or even utilized as computing centers. Given the significance of traffic management for CAVs, traffic signals are recommended at all junctions and crossings. These facilitate the smooth functioning of CAVs and uphold pedestrian safety.

3.2 Design principles for the second phase

In the second phase, CAVs are anticipated to possess a higher degree of automation, and the prospect of fully driverless vehicles becomes feasible. This phase necessitates more stringent road sign requirements, prompting the development of standardized and systematic road signage. Particularly crucial are signs that alert CAVs to potential hazards ahead (such as ongoing road maintenance), although such information can also be conveyed via V2X communication channels.

The determination of right-of-way configurations hinges upon the penetration rate of CAVs and vehicle density. In instances of high CAV penetration and density, the establishment of dedicated lanes for CAVs becomes a strategic choice. This facilitates a reduction in lane width and count in these dedicated lanes, subsequently creating more space for pedestrians and cyclists. Additionally, repurposing on-street parking zones for alternative uses can enhance comfort and safety for pedestrians and cyclists, aligning with a human-centric urban design philosophy (Riggs et al., 2020; Srinivasan et al., 2020).

The parking paradigm shifts in the second phase, requiring the creation of CAV-only parking lots, either in new areas or integrated into existing parking zones. The capacity of these lots can be maximized through closer parking enabled by CAVs’ self-parking capabilities. Given the absence of drivers, safe harbors assume enhanced importance to avert severe traffic incidents. Incorporating automatic alarms for emergency centers in safe harbors becomes necessary, given the potential lack of a driver to trigger emergency responses.

With wireless charging technology maturing, the charging landscape is expected to improve significantly. Charging pads in parking areas could save space compared to traditional charging piles. The feasibility of constructing electrified roads should be explored where environmental conditions allow. The second phase anticipates the utilization of further developed ITS, facilitating interconnectedness among people, processes, data, and components within the urban transport framework. A predominance of 5G and 5G+ digital networks is projected, delivering robust support for the traffic system. The expansion of vehicular cloud concepts offers potential for dynamic and personalized traffic management enhancements.

Advanced traffic signal timing optimization based on real-time conditions becomes achievable in this phase. This optimization proves pivotal in alleviating frequent traffic congestion, curtailing unnecessary delays, and mitigating energy consumption.

The transition from the first to the second phase hinges on CAV development levels and penetration rates. To maximize cost-effectiveness and environmental efficiency, adjusting existing infrastructure from the first phase is recommended, given the lower CAV development level and penetration rate. In cases where new infrastructure is necessary (due to replacement or urban growth), these designs must account for the anticipated future CAV development and penetration. Ideally, new infrastructure should align with high-level CAV requirements or at least allow for future renovation and upgrading, considering the long-term durability of traffic infrastructure. Forethought should also be given to new parking and charging facilities, factoring in future CAV distribution. This approach leverages first-phase investments as a foundation, avoiding the abandonment of functional infrastructure that doesn’t meet high-level CAV demands.

4 Case study

In order to illustrate the design principles discussed in the previous section, here we use a case study based on the city of Oxford. Oxford is an attractive and cosmopolitan city in the UK (Oxfordshire City Council, 2016). Its increasing population is adding pressure to the city’s existing transport system. Cars and buses have to compete for limited space with pedestrians and cyclists. To alleviate the current situation, the Oxford City Council has been working to explore new models for transport system development. CAVs, especially shared CAVs, combined with walking and cycling, will be a much more suitable and sustainable transport mode for Oxford in the future. To prepare the city for this, we are proposing new CAV-related infrastructure on four typical streets located in commercial, historical, and residential areas.

4.1 Oxford: an overview of the city

Situated in Southeast UK, Oxford is one of the five local authority districts within Oxfordshire county (Fig. 1). While encompassing a densely developed urban region, approximately 52% of the city consists of open spaces, though our study focuses primarily on the urban area. Renowned for its prestigious university heritage, Oxford boasts a variety of attractions beyond academia. In 2016, tourism contributed €832 million to the local economy, underscoring the significance of safeguarding historical structures and catering to the transportation needs of tourists.

Fig. 1
figure 1

The study area of the city of Oxford, Oxfordshire, UK. The left image shows the administrative area of Oxfordshire and its location in the UK. The five administrative districts are Oxford City, Cherwell, South Oxfordshire, Vale of White Horse, and West Oxfordshire. The right image shows the city of Oxford. Source: Esri Map

Full size image

Concerning the current state of traffic infrastructure, a plethora of damaged road signs and unclear road markings have been reported to the county council. Oxford has also established over 100 charging points, positioning it favorably for the transition to electric vehicles. While comprehensive 4G and 4G+ digital coverage extends across most areas, a stable and robust 4G network isn’t uniform throughout the road network.

Guided by the Oxford City Council Local Plan, infrastructure enhancements are envisioned to elevate the allure of walking and cycling, paralleled by controlled parking zone expansion. These zones permit parking solely within designated bays while restricting roadside parking. Concurrently, the government endeavors to double the capacity of Park & Ride stations, fostering public transport and steering toward a zero-emission zone. Emission-based charges would be levied on vehicles within these zones, with zero-emission vehicles exempt from fees. Substantial government investments target congestion alleviation, road safety enhancement, and improved road infrastructure.

4.2 Fundamental planning guidelines for Oxford

Taking into account the existing CAV technology and Oxford’s distinct characteristics, we present essential planning guidelines for CAV-related infrastructure in the city.

Historic-friendly design: Oxford’s rich historical heritage, particularly concentrated in the city center, necessitates a design approach that harmonizes with its cultural legacy. Preserving the city’s original appearance is paramount, with minor and discreet adaptations to existing infrastructure. Any new CAV-related additions must seamlessly blend in, ensuring they do not compromise the ornamental essence of the historical cityscape. Even outside conservation areas, integration within the existing infrastructure system remains the preferred and resource-efficient approach (Löfgren et al., 2018).

Promotion of environmentally friendly travel: Transport holds a significant share of negative environmental impacts (Thøgersen, 2018). The emissions from fossil fuel-powered vehicles, including carbon dioxide and nitrogen oxides, pose severe consequences for air quality, climate change, health, and historical building preservation. Encouraging sustainable, low-carbon travel modes such as walking, cycling, shared and public transport, along with curbing heavily polluting vehicles, is crucial to mitigate these impacts (Agarwal et al., 2020). The expansion of controlled parking zones aligns with this objective, allowing pedestrians and cyclists to access increased road space while reducing the preference for private vehicles.

Consideration of CAVs in planning: While controlled parking zones may discourage private vehicle usage, CAVs remain unaffected due to their autonomy in parking (Chee & Fernandez, 2013). The inherent capability of CAVs to self-navigate after passenger drop-off ensures their mobility is unhindered by parking limitations.

Trials for separated lanes: Dedicated infrastructure for pedestrians, cyclists, and vehicles has been shown to significantly reduce traffic accidents (Marshall & Ferenchak, 2019). Conducting trials for segregated lanes on select streets can provide valuable insights into the feasibility and efficacy of such a setup. These trials will help determine whether separated lanes can be practically implemented to enhance safety and convenience for different road users.

These guidelines, grounded in Oxford’s historical essence and urban objectives, inform the forthcoming integration of CAV-related infrastructure. They underline the importance of preserving heritage, prioritizing sustainable transport, and considering the unique attributes of CAVs in city planning.

4.3 Selection of typical streets

Urban streets can be classified by traffic features (e.g., the function (link or access) and traffic volume of the street), urban characteristics (e.g., surrounding land usage), and road characteristics (e.g., streets’ cross section and surface materials) (Soteropoulos et al., 2020). In this study, we differentiated streets according to the land use function along the street. Such function has been identified using POI (Point of Interest) data (details shown in Supplement). POI data include all geographic entities that can be abstracted as points, such as restaurants, banks, hotels, and universities (Hu & Han, 2019).

The original POI types in Oxford included a wide range of categories, including accommodation, automotive, business, education, food, health, public service, religious, settlement, shop, sports, tourism, transport, and other land use. To streamline our analysis, we reclassified these POI data into three primary categories: economy entities, non-economy entities, and residential communities. Economy entities comprise accommodation, business, food, shop, and sports. Non-economy entities encompass automotive, education, health, public service, and religious establishments. Residential communities pertain to the settlement-type POI data. For the purposes of our study, we did not consider tourism, transport, and other land use-type POI data. The distribution of the reclassified POI data is illustrated in Fig. 2.

Fig. 2
figure 2

The distribution of POI data in Oxford. The red dots represent economy entities. The green dots represent non-economy entities. The blue dots represent residential communities

Full size image

Based on the surrounding land use function, we selected four representative streets in Oxford: (1) High St: This street features a mix of both economy and non-economy entities along its stretch; (2) Merton St: Here, non-economy entities dominate the street’s surroundings; (3) Thames St: Near this street, economy entities are prevalent; (4) Canal St: Situated within a residential community, Canal St exhibits a different set of characteristics and CAV-related infrastructure requirements.

These four streets possess unique features and infrastructure needs, which in turn lead to distinct design principles and outcomes. However, they all share common overarching requirements and adhere to the fundamental planning guidelines discussed in the preceding section. We provide further elaboration on each of these streets below.

4.3.1 High St

High St is a main street across the city center, with many economy and non-economy entities on both sides. The street also has many historical buildings, including some from the University of Oxford. There are many bus lines and a large flow of pedestrians and cyclists. On-street parking is allowed. The current conditions of this street are shown in Fig. 3a-c. Figure 3a is somewhere in the middle of this street. Figure 3b and c show the two ends of this street, a crossing and a Y-junction, respectively. We find that the current lane markings and traffic signs are not clear. It is especially difficult to identify the signs due to variable natural light. In addition, there are no sufficient traffic signals over the crossing at the ends of this street.

Fig. 3
figure 3

(a-c) Current status of High Street. (a1-c1) Design proposals of the first phase, and (a2-c2) for the second phase. (a11-c11) Corresponding axonometries of (a1-c1). (a22-c22) Corresponding axonometries of (a2-c2). The street views were obtained from Google Maps

Full size image

4.3.2 Merton St

Merton Street is a historic and picturesque cobbled street adjacent to High Street, with low traffic and no bus line. There are a lot of colleges around this street. On-street parking is allowed, and the street also offers access to a small parking lot. The current conditions of this street are shown in Fig. 4a-c. Figure 4a is somewhere in the middle of this street. Figure 4b and c are the two ends of this street, a T-junction and a corner, respectively. Road boundaries and public parking spaces on the roadside are not always clear.

Fig. 4
figure 4

(a-c) Current status of Merton Street. (a1-c1) Design proposals of the first phase, and (a2-c2) for the second phase. (a11-c11) Corresponding axonometries of (a1-c1). (a22-c22) Corresponding axonometries of (a2-c2). The street views were obtained from Google Maps

Full size image

4.3.3 Thames St

Thames St is located near Westgate, a great shopping center in Oxford. Through field study, we found that it has had separated lanes already, but the lane markings are not clear enough. This street has access to a large parking lot; however, the parking spaces are not used efficiently. The current conditions of this street are shown in Fig. 5a-c. Figure 5a is somewhere in the middle of this street. Figure 5b and c are the two ends of this street.

Fig. 5
figure 5

(a-c) Current status of Thames Street. (a1-c1) Design proposals of the first phase, and (a2-c2) for the second phase. (a11-c11) Corresponding axonometries of (a1-c1). (a22-c22) Corresponding axonometries of (a2-c2). The street views were obtained from Google Maps

Full size image

4.3.4 Canal St

Canal St is located in a residential area and it is away from the city center. Jericho Community Center is located on this Street. It has one separated lane for parking, but there are no traffic signals nor buses along this street. Some Park & Ride areas are near this street. The current conditions of this street are shown in Fig. 6a-c. Figure 6a is somewhere in the middle of this street. Figure 6b is a crossing located on this street. Figure 6c is one of the two ends of this street, a complex intersection.

Fig. 6
figure 6

(a-c) Current status of Canal Street. (a1-c1) Design proposals of the first phase, and (a2-c2) for the second phase. (a11-c11) Corresponding axonometries of (a1-c1). (a22-c22) Corresponding axonometries of (a2-c2). The street views were obtained from Google Maps

Full size image

4.4 Deployment of CAV-related urban infrastructure on selected streets

In this section, we present our proposal for the deployment of CAV-related urban infrastructure across the four chosen streets during each of the two phases.

4.4.1 Phase 1

In alignment with Principles P1-P5 (Physical Infrastructure), as detailed in Table 2, our phase 1 deployment entails:

(1) Enhancement of signs and markings: Improving visibility for CAVs through lane restriping, upgraded traffic signs, and regular maintenance. Adhering to city council standards for sign and marking design is imperative. Material selection should consider both current CAV penetration rates and anticipated future traffic flows. (2) Retention of road network: Given the relatively modest CAV penetration rate, maintaining the current road network is recommended. Establishing dedicated CAV lanes could lead to resource waste and congestion in other lanes. (3) Sustainable parking provision: Upholding existing on-street parking provisions while adhering to city council regulations for allowable parking areas. (4) Safe harbors and pick-up/drop-off Locations: Designating spaces for safe harbors and convenient pick-up/drop-off zones requires meticulous analysis of factors such as traffic flow, land use, and public preferences. The locations must be strategically chosen based on thorough research. (5) Charging infrastructure: Installing charging piles and preparing alternative batteries in key areas such as bus stops (High St and Thames St), parking zones (Merton St and Canal St), safe harbors, and pick-up/drop-off locations. The historical aesthetics of the environment, particularly on streets like Merton St, must be considered in the design of charging facilities.

Complying with Principles V1-V2 (Virtual Infrastructure), all infrastructure should possess internet connectivity. Principle V4 advocates for the addition of more traffic signals at crossings, enhancing CAV introduction, and ensuring overall road user safety and comfort. Precise signal phase and timing configurations necessitate careful consideration, informed by thorough research into street-specific traffic conditions.

The schematics for phase 1 deployment are illustrated in Fig. 3 (a1-c1) for High St, Fig. 4 (a1-c1) for Merton St, Fig. 5 (a1-c1) for Thames St, and Fig. 6 (a1-c1) for Canal St. Sub-figures (a1-c1) correspond to the same sections as sub-figures (a-c) for each respective street, while sub-figures (a11-c11) provide the corresponding axonometric views of sub-figures (a1-c1).

4.4.2 Phase 2

In the second phase, we align with Principles P1-P5 to present our deployment strategy:

(1) Standardized lanes and signs: Implement unified regional or country standards for CAV-adapted traffic signs and markings, and ensure their regular maintenance. We anticipate the availability of standardized guidelines for CAV-suitable signage and markings. (2) Access restriction and dedicated lanes: Retain the current road network, potentially reserving certain streets exclusively for CAV access. Canal St, given its surroundings and road material, is a strong candidate for dedicated CAV lanes. However, any implementation of separated lanes should involve public engagement, informing them of the pros and cons. (3) Parking reallocation: Eliminate on-street parking (High St, Thames St, and Canal St) to allocate space for pedestrians and cyclists. Merton St can retain on-street parking, with more designated as disabled spaces. In Canal St’s parking lot, spaces dedicated to CAVs should be designated. (4) Continued safe harbors and pick-up/drop-off areas: Maintain the safe harbors and pick-up & drop-off points established in Phase 1, enhancing safe harbors with automatic alarms for situations without a driver present. (5) Charging infrastructure enhancement: Introduce charging pads over bus stops (High St and Thames St), off-street parking (Merton St and Canal St), on-street parking (Merton St), safe harbors, and pick-up & drop-off zones. Canal St remains a strong candidate for dynamic charging through electrification due to its attributes.

Incorporating Principles V1-V4 (Virtual Infrastructure), the advanced ITS technologies, digital networks, and vehicular clouds facilitate dynamic traffic management. This enables adaptable signal phase and timing adjustments in real-time to minimize delays and energy.

The schematics for the second phase are shown in Fig. 3 (a2-c2) for High St, Fig. 4 (a2-c2) for Merton St, Fig. 5 (a2-c2) for Thames St, and Fig. 6 (a2-c2) for Canal St. In all cases, sub-figures (a2-c2), correspond to the same passages shown in sub-figures (a-c) for that specific street; and sub-figures (a22-c22) show the corresponding axonometries of sub-figures (a2-c2).

We have presented an extensive framework of design principles for CAV-related infrastructure, meticulously addressing a multitude of essential factors in their deployment. This comprehensive framework offers invaluable guidance to city planners as they prepare their urban areas for the integration of CAVs in an efficient and systematic manner. However, the practical implementation of these principles in the real world poses notable challenges. As city planners embark on creating new traffic infrastructures, they must account for an intricate web of factors. These factors encompass present traffic patterns, road conditions, environmental contexts, public preferences, and extend to the future’s evolving technology, economic landscape, and societal desires. Achieving sustainable development requires a harmonious balance of all these aspects.

Predicting the future is a complex and risky endeavor. For instance, in anticipation of high CAV penetration rates, planners might opt for new out-of-city-center parking lots optimized for CAVs. However, if the actual penetration rate falls below expectations, this strategy could result in resource wastage and traditional vehicles struggling to secure parking spots. The inherent uncertainty makes the endeavor of future-oriented planning all the more challenging. In essence, traffic planning in the context of CAVs is a multifaceted undertaking. It demands the harmonious fusion of current realities, future aspirations, technological advancements, economic dynamics, and public sentiments. Successful navigation of these complexities requires strategic foresight, adaptable planning, and a willingness to recalibrate approaches as conditions evolve. By embracing this complexity, cities can pave the way toward a sustainable and seamlessly integrated future for connected and autonomous vehicles.

5 Discussion

This paper introduces a comprehensive framework of CAV-related infrastructure design principles, addressing a multitude of often-overlooked aspects that are crucial for effective planning, deployment, and evaluation. This framework offers unified guidelines for cities as they prepare to integrate an increasing number of CAVs into their urban landscapes. The case study of Oxford exemplifies the application of these principles in a real-world context, showcasing their practical relevance.

Urban planning is inherently forward-looking, requiring a balanced consideration of both immediate needs and long-term goals. Yet, the planning and development across different time periods are interwoven. If current infrastructure aligns with future development needs, retrofitting demands might remain minimal and cost-effective. However, the alternative could entail elevated costs as existing facilities are torn down to make way for new ones, particularly in capital-intensive sectors like transportation. Hence, it’s imperative to design and construct facilities with a future-oriented perspective that accommodates not only current conditions and needs, but also those anticipated. While predicting the future involves numerous variables, extrapolating from present development trends can yield informative insights. According to the CAV upgrade plan (Liu et al., 2019), levels 1–2 CAVs are projected to mature, and levels 3–4 CAVs to attain efficiency by the 2030s. By the 2050s, levels 3–4 CAVs are anticipated to be mature, with level 5 CAVs achieving efficiency. Given the lifecycle of transportation projects and infrastructure, planning for these scenarios should already be underway. However, it’s important to note that infrastructure requisites for various CAV levels differ. Higher-level CAVs necessitate greater V2X requirements. Consequently, the advancement of CAV-related technologies, including mobile networks, must be integral to planning. In this study, we address two distinct time periods, each coinciding with a specific CAV development stage. We thereby furnish the corresponding design framework to equip cities for the evolving landscape of CAV technology.

When applying the design framework to a specific city, a thorough investigation into the city’s current conditions becomes paramount. Cities worldwide vary significantly in terms of scale, topology, infrastructure, and needs. The configuration of CAV-related infrastructure must be tailored to the unique context of the study area. We propose the following procedure for assessing the existing city conditions:

  1. (1)

    Identify the main function of the city: Cities can serve as manufacturing hubs, tourist destinations, financial centers, or a combination of roles. Different city functions entail distinct transportation characteristics and requirements. The city’s primary function can often be deduced from its economic composition.

  2. (2)

    Recognize current land use patterns and building conditions: Special zones like conservation areas and historical sites deserve particular attention. Generally, making significant changes to these areas is challenging.

  3. (3)

    Understand the local topography: The terrain can limit the feasibility of constructing tunnels, rail lines, or other transportation infrastructure.

  4. (4)

    Understand the existing transportation network topology and urban design: Modifying road and rail networks is possible but usually comes with high costs and constraints due to the city’s established design.

  5. (5)

    Study the travel patterns and preferences of both local residents and potentially city visitors: Residents (and/or tourists) might prefer one transport mode over the other. We should then design the appropriate strategies for either supporting those preferences (e.g., providing additional biking infrastructure for people to commute by bike), or promote the use of more sustainable transport modes (e.g., provide better public transport services to induce a mode shift towards public transport).

  6. (6)

    Take into account the future development plan designated by the local government: New designs and/or modifications should be integrated into the overall plan and vision of the city.

Examining the attributes of diverse streets within the city constitutes another pivotal starting point. Given the multitude of streets in any urban area, each with its distinctive traits, devising singular design guidelines for each street is an impractical endeavor. A more pragmatic approach involves categorizing streets into broader classifications, enabling the formulation of general design guidelines. Typically, streets can be grouped into categories such as pedestrian-only streets, commercial-shared streets, residential-shared streets, central one-/two-way streets, and more (Global Designing Cities Initiative, 2016). Street typologies often correlate with the surrounding land use patterns, as streets are required to fulfill the social, environmental, and economic demands of their immediate vicinity. Long-standing POI data has proven useful in discerning the roles of individual entities or buildings. By leveraging such data to understand the functions of structures along each street, we can potentially identify representative streets that serve as foundational platforms for shaping our design guidelines.

6 Conclusion

The shift towards autonomous mobility is a pivotal response to the challenges posed by contemporary urban and transport systems. The integration of CAVs has the potential to significantly mitigate traffic accidents, congestion, air pollution, and energy consumption, especially when coupled with electric vehicles and shared-mobility concepts. Establishing CAV-related infrastructure forms the cornerstone of successfully introducing CAVs into urban environments. In this study, we have developed a comprehensive framework of design principles for CAV-related infrastructure encompassing both physical and virtual aspects. Moreover, these principles are designed to adapt and evolve in tandem with the progression of CAV technology. By leveraging a case study centered around Oxford, we have illustrated the practical application and evolution of our proposed framework. This case study involves evaluating the city’s existing conditions, selecting representative streets, and crafting CAV-related infrastructure designs that align with both street characteristics and the evolving CAV landscape.

Drawing from the current urban conditions and the future transport system envisioned by local governments, we have designed layout schemes for CAV-related infrastructure across four distinct street types in Oxford. For commercial areas, our recommendations encompass integrating CAV parking facilities within existing lots and introducing dedicated pick-up and drop-off points. In residential zones, we suggest reducing on-street parking and increasing charging infrastructure, repurposing freed space for alternative activities or sustainable transport modes like pedestrians and bicyclists. Narrow historical streets, by contrast, necessitate minimal alterations, strictly in adherence to the street and city aesthetics, especially respecting historical buildings and the overall built environment.

While our case study is rooted in the context of the UK, the design principles discussed herein hold universal applicability. However, it is essential to approach their implementation with a measure of caution, as the distinct characteristics and requirements of individual cities may warrant modifications to the proposed framework. As nations worldwide strive to integrate CAVs into their current transportation systems, we posit that the CAV-related infrastructure design framework outlined in this study stands poised to serve as a vital reference.