Home Journals Articles Updates About
CN
image
收藏切换
Vehicle-to-Everything Communication in Intelligent Connected Vehicles: A Survey and Taxonomy
收藏切换
PDF
Xinyu Zhang1, Junxian Li1, Jingyi Zhou1, Shiyan Zhang1, Jingyuan Wang1, Yi Yuan1, Jiale Liu1, Jun Li1
Automotive Innovation | 2025, 8(1) : 13 - 45
Less
收藏切换
Automotive Innovation | 2025, 8(1): 13-45
• Papers •
Vehicle-to-Everything Communication in Intelligent Connected Vehicles: A Survey and Taxonomy
Full
Xinyu Zhang1, Junxian Li1, Jingyi Zhou1, Shiyan Zhang1, Jingyuan Wang1, Yi Yuan1, Jiale Liu1, Jun Li1
Affiliations
  • 1 Tsinghua University State Key Laboratory of Automotive Safety and Energy, School of Vehicle and Mobility Beijing 100084 China
doi: 10.1007/s42154-024-00310-2
Outline
收藏切换

This paper conducts a thorough exploration of vehicletoeverything (V2X) communication in the realm of intelligent connected vehicles (ICVs). It initiates by tackling challenges across three pivotal phases of cooperative communication: precommunication, duringcommunication, and postcommunication. The discourse delves into a spectrum of concepts and strategies to surmount these challenges. Furthermore, it meticulously scrutinizes diverse communication scenarios and associated techniques, evaluating their significance and feasibility. Moreover, an indepth analysis of various datasets is undertaken, considering their distinctive attributes and suitability for diverse communication tasks. The paper critically examines and debates the platforms and frameworks used in the experiments, providing valuable insights into their performance. Following a comprehensive review of existing methods and datasets, the paper identifies potential research directions and challenges that warrant further exploration in the realm of V2X communication for intelligent connected vehicles. This comprehensive examination contributes to a deeper understanding of the subject, paving the way for future advancements in this dynamic field.

Vehicle-to-everything communication  /  Intelligent connected vehicles  /  Collaborative perception  /  Intelligent transportation system
Xinyu Zhang, Junxian Li, Jingyi Zhou, Shiyan Zhang, Jingyuan Wang, Yi Yuan, Jiale Liu, Jun Li. Vehicle-to-Everything Communication in Intelligent Connected Vehicles: A Survey and Taxonomy[J]. Automotive Innovation, 2025 , 8 (1) : 13 -45 . DOI: 10.1007/s42154-024-00310-2
Full Text
Less
1 Introduction
Less
Vehicle-to-Everything (V2X) communication provide comprehensive sensing capabilities for intelligent connected vehicles while improving transportation efficiency and safety. Moreover, V2X serves as the foundation for enabling higher levels of autonomous driving technology. In particular, V2X communication refers to the exchange of information between vehicles and their surroundings, including other vehicles, pedestrians, roadside infrastructure and networks.
A considerable number of researchers are devoted to the field of V2X communication in intelligent connected vehicles (ICVs), having proposed a number of useful methods. Several review papers have discussed parts of the techniques. For instance, Tan et al.[1] underscored the need for reliable sensing and communication technologies, also emphasizing human factors such as user comfort and reliability for widespread ICV adoption. Sarker et al.[2] extended this discussion, exploring multi-layer ICV controllers that integrate both human factors and connectivity. However, Matin and Dia [3] cautioned that the literature remains fragmented and identified four research pillars: safety, efficiency, communication, and technology. Alongside these, Taslimasa et al.[4] discussed the network security challenges in ICVs, introducing novel intrusion detection systems. Adding future perspectives, Duan et al. explored the potential of 5G-V2X communication architectures [5], identifying both challenges and future research directions.
Intelligent Connected Vehicles are an important part of modern transport systems, and they rely on advanced communication technologies to enable efficient, safe and intelligent transport solutions. The impact of communication technology on ICVs is mainly reflected in the following aspects:(1) Safety enhancement: Communication technology enables ICVs to realise V2X communication, including vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-pedestrian (V2P), etc., which greatly enhances the vehicle’s sensing ability and allows it to obtain timely information about the traffic environment and warns of potential dangers, thus significantly enhancing road safety.(2) Traffic efficiency improvement: Through effective communication, ICVs can receive real-time traffic information, such as traffic congestion, accidents, changes in road conditions, etc., so that vehicles can adjust their driving routes or speeds according to real-time data, optimise traffic flow, reduce congestion, and improve overall traffic efficiency.(3) Supporting the development of autonomous driving technology: Communication technology is the key to realising advanced autonomous driving functions. Through high-speed communication with other vehicles, traffic infrastructure, cloud servers, etc., vehicles can perform complex data exchange and processing to achieve highly automated driving.(4) Facilitating remote vehicle control and services: Communication technologies allow vehicles to exchange data with remote servers or control centres, supporting services such as remote monitoring, diagnosis, maintenance, and software updates. This, in turn, enhances the vehicle maintenance efficiency and service quality.(5) Data-driven decision-making and services: ICVs are able to collect a large amount of data and transmit it to manufacturers, service providers, or traffic management centres through communication technologies, supporting data-based decision-making, developing person-alised services and enhancing user experience.(6) Energy management and optimisation: In electric ICVs, communication technologies can also enable vehicle-to-grid (V2G) interaction, optimise energy use, support smart charging, and even allow vehicles to feed power back to the grid when needed.
However, while communication technologies bring various benefits to ICVs, there are also challenges, such as communication security, data privacy protection, and increased network dependency.
The motivation of this review is to explore V2X communication within ICVs, addressing challenges across pre-, during-, and post-communication phases. It aims to surmount these challenges through a comprehensive analysis of communication scenarios, techniques, and datasets. The significance lies in filling a crucial research gap by providing a review on methods for solving communication problems, analyzing application scenarios, and evaluating the significance and problems of various datasets and experiment platforms. This work contributes to a deeper understanding of V2X communication, paving the way for future advancements in the dynamic field of ICVs.
As shown in Fig. 1, this paper reviews the literature on V2X communication, delves into its problems and challenges, explore different communication scenarios, and evaluates the datasets and experimental platforms. The main contributions of this work can be summarized as follows:
(1) The challenges faced in different stages of communication are discussed and a comprehensive literature review on the methods to solve the problems is provided, filling a crucial research gap.
(2) The analysis of various communication scenarios is presented with evaluation on their significance, perception and communication effects and problems, along with corresponding solutions.
(3) The datasets pertinent to cooperative communication research are summarized, highlighting their characteristics and suitability for various tasks. Additionally, the platforms and frameworks employed in model testing and evaluation are outlined for the experiments.
The rest of this paper is arranged as follows: In Sects. 1 and 2, the groundwork is laid, presenting the context, challenges, and historical evolution of collaborative communication. In a parallel layout, Sects. 3 and 4 illustrate the depth of the issues and their real-world applications, from the minutiae of sensor accuracy to broad smart city applications. Sections 5 and 6 further underscore the depth of the paper, categorizing data by perspective and elaborating on associated tasks. Section 7 provides key conclusions and potential enhancements.
The abbreviations mentioned in this paper and their annotations are provided in Table 1. All references in this paper have been rigorously evaluated and assessed in a comprehensive quantitative and qualitative manner, and Table 2 summarises the papers that meet the research criteria for 2019 and before up to 2023. Among them, the selected references are categorised by percentage as IEEE Xplore 75.5%, arXiv 7.2%, Elsevier 2.9%, Springer 3.6% and others 10.8%.
2 Background
Less
2.1 V2X Communication Classification
V2X communication enables cooperative perception in intelligent transportation systems (ITS), allowing vehicles to exchange messages with other agents such as the infrastructures and pedestrians [6]. As shown in Fig. 2,V2X communication framework has four layers. The cloud layer serves as the central data hub, while the edge computing layer focuses on regional data processing with local gateway and multi-access edge computing (MEC) host. The infrastructure layer handles collaborative communications, and the client layer brings forth ICV applications. The following are introductions to different types of collaborative communications. Vehicle-to-vehicle (V2V) communication allows the exchange of real-time information between several vehicles, enhancing safety and efficiency on the road [7,8]. Vehicle-to-infrastructure (V2I) communication [9] helps to optimize traffic flow, improve safety and enhance overall transportation efficiency. Vehicle-to-roadside (V2R) communication enables communication with roadside sensors, cameras, and other infrastructure elements, helping drivers to make informed decisions and enhancing overall road safety and convenience [10]. Vehicle-to-home(V2H)communication builds the interaction between vehicles and connected houses, enabling energy management and power transfer [11]. Vehicle-to-building (V2B) communication enables vehicles to share data and perform various tasks, including electric vehicle charging and energy management [12]. Additionally, vehicle-to-pedestrian (V2P) communication involves the exchange of information with pedestrians and other vulnerable road users [13], which aims to help prevent accidents and improve overall pedestrian safety on the road. Vehicle-to-server (V2S) communication includes the interaction with central servers or cloud-based services, allowing the access and exchange of data with remote servers or services for various purposes [14]. Vehicle-to-grid (V2G) communication achieves the exchange of information and power between electric vehicles (EVs) and power grids [15], permitting EVs to draw power from the grid, and send surplus power back to the grid, effectively turning EVs into mobile energy storage systems. Vehicle-to-device (V2D) communication [16] enables seamless connection and integration between ICV systems and personal devices for remote control and customized functions.
The classification of V2X is of great significance to its application and development, which is embodied in the following aspects:
(1) Promote traffic safety: V2X can achieve real-time communication and information sharing between vehicles, provide rich traffic status information, and enhance driver’s awareness and responsiveness. Through V2X classification, different types of traffic participants can obtain the required information according to their own needs, thus forming a more efficient and safer transportation system [17].
(2) Improve traffic efficiency: V2X classification can improve the overall efficiency of traffic systems by identifying appropriate communication technologies and protocols based on different application scenarios and needs. For example, for urban traffic management departments, V2X technology can achieve signal optimization, congestion warning and other functions, reducing traffic congestion and travel time [18].
(3) Promote intelligent transportation development: The V2X classification can provide the guidance and specifications for constructing intelligent transportation systems, enabling various devices and systems to interconnect on the same platform. By connecting vehicles, roads and traffic management facilities, the functions of intelligent navigation, autonomous driving, and intelligent traffic management can be realized [19].
2.2 Framework for Comparative Studies Related to V2X
For a comprehensive understanding of V2X technology, Table 3 compares the advantages, limitations, values, and challenges of V2X architectures which can help researchers and practitioners to fully understand the different aspects of V2X technology, including the various communication modes such as V2V, V2I, and V2P. This in-depth understanding is the basis for evaluating the potential of the technology and guiding future research and practice. Comparing the architectures of different V2X technologies can provide a scientific basis for policymakers to help them formulate reasonable policies and standards for the healthy development and application of the technology. For investors and decision-makers, understanding the comparative advantages and limitations of various V2X technologies can help them make smarter investments and decisions, optimise resource allocation and improve investment efficiency. Comparative research on V2X-related architectures not only reveals the strengths and weaknesses of various technologies, but also identifies gaps in current research and practice, points out the direction and focus of future research, and promotes progress in the field.
In summary, an in-depth comparative analysis of V2X-related architectures can stimulate new research ideas and technological innovations, provide new perspectives and methods for solving practical problems, and accelerate the application and marketisation of V2X technology. In addition, the construction of V2X-related comparative research architectures is of great significance to promoting scientific research, technological innovation, policy formulation and market application in this field.
2.3 Wireless Communication Technologies in ICV
In Intelligent Connected Vehicles (ICVs), various wireless technologies play crucial roles. Dedicated Short Range Communications (DSRC) offers low latency and reliable communication for vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) interactions. Long-Term Evolution for Vehicles (LTE-V) extends the LTE mobile network to support vehicle communication, enabling high-speed data transmission. $4\mathrm{G}$ and $5\mathrm{G}$ technologies provide broadband connection for ICVs, enabling streaming services and high-resolution map downloads. Wi-Fi is common for in-vehicle hotspots, allowing passengers to connect devices. Bluetooth enables hands-free calling and audio streaming. Visible Light Communications (VLC) uses LED headlights to transmit data, offering a new dimension in V2V communication. Their features are shown in Table 4.
Using these technologies in ICVs is crucial for enhancing safety, efficiency and passenger comfort. They enable features like collision avoidance, traffic flow optimization, and in-vehicle connectivity. Despite their limitations, advancements in technology and infrastructure will continue to improve their performance and expand their applications in the ICV ecosystem.
2.4 Sensor Technologies in ICV
In ICVs, sensors are important components that provide sensory information to the vehicle or facility. Ultra Wide Band (UWB)[20] is used for secure and low-power distance measurement and communication by leveraging ultra-wideband signals, enabling accurate positioning and navigation. mmWave [21] provides high-speed data transmission for V2V and V2I communication. Bluetooth Low Energy (BLE) is commonly used for low-power, short-range communication between vehicles and mobile devices. Cameras [22] are employed for visual perception and environmental monitoring in ICV, enabling functions such as object detection, traffic congestion analysis, and pedestrian recognition. Global Positioning System (GPS) is utilized for accurate vehicle positioning and navigation, while LiDAR, by using laser beams, provides high-precision three-dimensional mapping and object detection capabilities for autonomous driving and advanced driver assistance systems.
The application of these technologies in ICVs is essential for enhancing traffic efficiency, safety, and convenience. With the increasing development of autonomous driving and advanced driver assistance systems, the study of UWB, $\mathrm{{mmWave}}$, BLE, Cameras, GPS, and LiDAR will continue to play a crucial role in enabling safe and efficient transportation systems in the future.
2.5 Validation Related to V2X Technologies
The significance of using methods such as formal methods, simulations, and testing for V2X technology validation lies in their ability to ensure the reliability, safety, and effectiveness of vehicular communication systems. Formal methods provide a rigorous mathematical approach to verify system properties, simulations allow for the evaluation of system performance under various conditions without the need for physical prototypes, and testing ensures that the technology functions correctly in real-world scenarios. Together, these methods comprehensively assess the readiness and safety of V2X technologies prior to deployment, addressing potential issues and optimizing performance [23].
(1) V2V Field Testing: Xu Yaqi et al. detailed the field testing of V2V communication performance at an obstructed intersection. It compares V2V with a message forwarding mechanism V2I2V (Vehicle-to-Infrastructure-to-Vehicle) and concludes that V2I2V offers superior performance in obstructed intersections due to lower packet error rates and communication delay compliance with 3GPP standards. This study emphasizes the influence of obstacles on V2X communication performance and suggests the V2I2V method as a viable solution to improve safety warnings at intersections [24].
(2) V2X Security Validation: He Kexun et al. discussed various methods used for validating V2X technologies. They outline simulation scenario tests, communication verification tests, and security communication tests. These methods involve simulating vehicle interactions using hardware-in-the-loop to construct complex scenarios, including attacks like message spoofing and eavesdropping. It also mentions the construction of a test system to evaluate the digital certificate security, privacy leakage, and the efficiency of signature verification and certificate revocation in V2X communications [25].
(3) C-V2X Test System: Zeng Lingqiu et al. described the development of a Cellular-Vehicle-to-Everything (C-V2X) automated test system designed for both field and in-chamber testing. It focuses on standardizing test interfaces and improving the automation level of testing systems through real-time data collection and analysis. This includes the use of the MQTT protocol for efficient data exchange and the deployment of a concurrency test to verify system validity. The document highlights the importance of integrating communication performance tests with runtime application tests, employing tools like traffic simulators and scenario injection modules for comprehensive validation [26].
(4) V2X Testing Platform: Senkus Burak et al. discussed the development of a human-in-the-loop testing platform for V2X applications. This platform enables multiple users to control virtual vehicles and test V2X applications in real time, aiming to offer a balance between virtual and real-world testing environments. It incorporates various technologies and methodologies, such as simulation, Hardware-in-the-loop (HiL), and Vehicle-in-the-loop (ViL), to facilitate comprehensive testing scenarios, including collision risk warnings and protocol stack update time assessments [27].
(5) C-V2X In-Chamber Testing: Chen Dongmei et al. introduced a comprehensive in-chamber test scheme for C-V2X applications, utilizing Digital Twin technology to simulate real-world scenarios. This approach includes GNSS simulation for accurate vehicle positioning, scenario reproduction based on real-world data collection, and cross-validation methods to ensure test result validity by comparing in-chamber results with on-road data [28].
2.6 Challenges in ICV Related to Communication
2.6.1 ICV Security and Safety
(1) Information Security: Car-connected devices in ICVs can gather extensive information regarding the condition of vehicles and the privacy of drivers. If this data is accessed by hackers or malicious third parties, it can result in the violation of personal privacy and its potential misuse for illegal activities. Additionally, car-connected devices often communicate with other devices or cloud services through wireless networks. Hackers can exploit these networks to infiltrate the vehicle’s network system, thereby stealing data or manipulating instructions that control the vehicle, ultimately leading to unusual vehicle movement and compromising security [29].
(2) Personal Safety: This also brings about certain difficulties in ensuring the safety of individuals during accidents [17]. One major challenge involves establishing a seamless and dependable network connection among vehicles. The challenge lies in the accuracy and reliability of the shared information between vehicles. This poses a significant risk, as inaccurate information can result in wrong decision-making by the vehicles, potentially leading to accidents or exacerbating the collisions.
2.6.2 Perception Issues
(1) Heterogeneous Information: In ICV scenarios, vehicles and infrastructure devices may have different sensing capabilities, such as different types and qualities of sensors. This heterogeneity creates challenges in achieving a unified and accurate perception of the surrounding environment [30].
(2) Sensor limitation: Moreover, these sensors have limitations, such as limited range, susceptibility to weather conditions, occlusion issues, and high computational requirements, which can affect the accuracy and reliability of perception systems [31].
(3) Localization: Additionally, accurate localization of vehicles and the creation of high-fidelity maps are essential for effective perception in ICVs. However, achieving accurate and up-to-date localization and mapping in real-time can be challenging due to factors such as GPS inaccuracies, dynamic environments, and rapidly changing conditions [32].
(4) Environment: Factors such as weather conditions, lighting variations, and complex traffic scenarios can affect the perception of the surrounding environment, and addressing these challenges is crucial for reliable operation [33].
2.6.3 Data Transmission Issues
(1) Limited Bandwidth: Moreover, the available bandwidth for communication between vehicles and infrastructure is often limited. Sharing this limited resource among numerous vehicles can lead to reduced data transfer rates and slower communication. Furthermore, network congestion and limited bandwidth can cause latency in transmitting data, impacting the real-time nature of ICV communication. In the context of collaborative perception, latency issues can be caused by the communication delay between multiple agents. This delay can cause potential performance degradation and high risks in safety-critical applications [34].
(2) Heterogeneous Communication Technologies: IoV systems typically consist of a diverse network comprising both infrastructure and autonomous vehicles (AVs). The differences in configuration between the infrastructure and vehicle sensors, including their types, noise levels, installation heights, and even attributes and modes, pose significant challenges in designing a V2X perception system [30].
2.6.4 Control Issues
(1) Motion Control: For motion control in the ICV, communication technologies utilized to help enhance the sensing range may bring various problems due to switching information flow topologies and the time delay. Moreover, in terms of heterogeneous dynamics, the assumption of homogeneity among vehicles is not realistic due to the wide range of models, types, and powertrain component variations. Meanwhile, the internal stability of a Cooperative Adaptive Cruise Control (CACC) system does not automatically guarantee string stability. If error signals are amplified upstream despite the closed-loop system’s internal stability, it can lead to collisions among consecutive vehicles. The nonlinear and heterogeneous nature of vehicle dynamics and the uncertain reliability of wireless communication links require careful design of CAV controllers to mitigate the impact of feedback channels and ensure string stability in CACC strings [35].
(2) Traffic Control: In the traffic control of ICVs, a range of intricate issues arise. For event prediction and path tracking, steering delay, path tracking error, and uncertainty are the pivotal factors affecting vehicle’s driving accuracy and safety. Path planning also faces challenges
3 Issues Analysis
Less
positioning accuracy. The phenomenon of occlusion can also
By conducting a thorough issues analysis, the complexities and obstacles inherent to ICV can be identified and addressed, ensuring a robust framework for implementation. As shown in Fig.3, ICV communication is susceptible to a myriad of challenges across various stages of communication. Therefore, the analysis will be done by three distinct stages, with a summary of literature for every communication stage shown in Tables 3,4,5 and a timeline in Fig. 4.
3.1 Pre-communication
The pre-communication stage sets the foundational parameters for the entirety of the communication process ( Table 6 ). Challenges such as noise interference and perception errors have the potential to jeopardize the security and reliability of ICV, as any discrepancies can compound and magnify as communication progresses.
3.1.1 Reduced Accuracy of Sensors
in adapting to diverse scenarios, dynamic urban net-works with congestion, and natural environments. The The vulnerabilities of sensors can cause noise interference timely dissemination of information is crucial for traffic to the communication. Firstly, environmental conditions can control, yet issues such as ensuring real-time transmis-introduce variables that significantly influence the function-sion, mitigating the impact of signal interruptions, and ality and accuracy of various sensors. For instance, high effective beam control present significant challenges. temperatures and humidity levels can adversely affect the Furthermore, synchronization problems, including precision of displacement sensors. External factors like inter-symbol interference, limited communication electromagnetic fields and lightning can introduce noise resources, distortion, and signal shielding in areas like interference. To mitigate the influence of adverse weather bridges and tunnels, can affect the accuracy and reli-on sensors, a domain-adaptive detection framework for fog ability of synchronization. These interconnected issues was introduced by Li et al.[33] and a calibration-free BEV need to be comprehensively addressed in the ICV traffic network was presented by Fan et al.[39]. Secondly, constant control systems to ensure safe, efficient, and reliable movement introduces variables that can significantly distort transportation systems [18,36-38]. sensor readings. For instance, as distance increases, there is a reduction in the discernible depth difference between a moving vehicle and the ground, which significantly reduces lead to spatial data misalignment. These factors hamper the precision and efficiency of object detection. Focusing on these challenges, BEVHeight [31], CoaliGN framework [40],CO3 [41], the unified bandwidth-efficient and multi-resolution based collaborative perception framework (UMC)[42] and VIMI framework [43] are proposed to improve the accuracy of 3D object detection.
3.1.2 Fusing and Collaborative Error
Incorrectly fused data can lead to misinterpretations and erroneous actions, with potential ramifications for safety and efficiency. Firstly, sensor accuracy problems can disrupt the synthesis of data, as seen with the VIMI framework [43], where the multi-scale cross attention module (AM) addresses asynchronous calibration noise and optimizes bandwidth utilization, leading to a significant accuracy boost. Additionally, the latency prediction framework underlines the importance of accurate data processing in real-time ICV operations [38]. Complications arising from sensor data inconsistencies are noticeable [39][46]. Techniques like Double-M quantification [44] offer remedies. The multi-task neural network system [45] underscores the complexities in vehicle collaborative actions and the need for specialized solutions.
3.1.3 Missing Information for Image Extraction
Sensor limitations in terms of dimension, scene occlusion, and resolution can result in incomplete image data. This could compromise the ability to accurately assess the environment, leading to functional constraints or safety risks. The issue of missing image data can be categorized based on the type of sensor involved, such as camera image gaps or radar image omissions. Estimating the pose of an object from a monocular RGB image without depth information, it can be difficult to accurately acquire the true size and spatial relationship of the object. To complete the image information, object-level deep reconstruction network (Old-Net) for category-level 6D object pose estimation based on monocular RGB images [47] is proposed.
In addition, using radar as a sensor, information loss can occur due to various factors such as occlusions, interference from dense environments, or degradation over long-distance transmission. These limitations can reduce the effectiveness in assessing the environment, posing potential safety risks. To address it, Brucker et al.[48] proposed an unsupervised method for laser-based 3D object detection using infrastructure sensors.
3.1.4 Volume Limitation and Compression
Capacity limitation and compression techniques optimize data transmission by reducing bandwidth usage. These methods enhance transmission efficiency, mitigating network congestion and latency. Typically, these techniques focus on either extracting essential information or compressing the overall data size for streamlined communication. For instance, Liu et al.[49] developed a neural network-based multi-stage handshake mechanism that focuses on compressing key information. This compression serves as a foundation for Hu et al.[50], who introduced Where2Comm, a framework that further optimizes communication by narrowing focus to perceptually critical regions. Building on the theme of optimization, Schwarzbach et al.[51] enhanced GNSS-based positioning through additional sensor data and collaborative positioning (CP) methods, reducing correlated errors.
3.1.5 Trade-Off Between Performance and Bandwidth
There is a delicate balance between communication performance and bandwidth. High-performance communication devices can transmit large volumes of data, but risk consuming much bandwidth, leading to delays or data losses. Conversely, ample bandwidth allows more data transfer, but can overwhelm devices with limited processing capabilities, thereby diminishing communication performance. To achieve the balance, Liu et al.[52] proposed a communication framework which employs an activation-based function to prune less crucial connections, thus enhancing both bandwidth utilization and communication performance.
3.2 During-Communication
During communication, as shown in Fig. 5,real-time data sharing and processing is based on edge computing. Multiple vehicles form a communication domain, and different domains achieve information exchange. Each edge server summarizes the information of the terminal devices and can be linked to form a network, which can support the functions of data processing, filtering and analysis. In this communication stage, live data transmission faces disruptions from errors and delays. Issues like system responsiveness, uncertainties, and perception errors hinder vehicle-infrastructure coordination. Moreover, route planning, network and vehicle security are essential for ICV optimal navigation and safety ( Table 7 ).
3.2.1 Positioning Errors
Positioning errors impact vehicle mapping in connected systems. Accurate relative transform estimation is crucial, as highlighted by OptiMatch [32]. Systems proposed in Refs.[53-56] advocate using multiple data sources for reliable perception. The multi-modal virtual-real fusion (MVRF)-based transformer [57] addresses challenges in traffic, leveraging LiDAR data. Meanwhile, strategies proposed in Ref.[65] blend different data types for better detection performance.
3.2.2 Balance
Maintaining the balance across various facets is crucial for optimizing ICV performance. Firstly, the equilibrium of network resources, such as bandwidth, storage, and computing power, underpins seamless data transfer and processing. Various strategies like ComAp [58], Q-learning [59], and actor-critic frameworks [60] are introduced to address resource balance, emphasizing cloud infrastructures [61]. Furthermore, the balance between delay and cost has paramount importance in ensuring real-time responsiveness without escalating expenses. Lastly, the tandem of bandwidth and precision cannot be overlooked. The limitations of single-vehicle on-board sensors can be addressed through collaborative perception, which enhances accuracy and stability by utilizing shared perception data. Systems like Map Container [53], COOPERNAUT [62], and bandwidth aware multi-domain virtual network embedding (BA-VNE)[63] highlight the significance of optimizing bandwidth usage while not compromising the accuracy of shared perception data in vehicle communication.
3.2.3 Delay
Reducing delay improves operational efficiency. In prediction, unpredictable communication can cause overhead. MASS [64], ComAp [58], FFNet [65], adaptive model predictive control for uncertain model algorithm to predict control (UM-AMPC)[66], and preview steering control [67] boost collaboration and transmission using past data and proactive allocation. Reconfigurable intelligent surface (RIS) technology [68] enhances V2I links and real-time processing. Protocols, like that reffered in Ref.[69] aid in understanding stochastic delays. Meanwhile, synchronization issues arise from delays. Multidimensional Markov models [70] address these by focusing on packet intervals and integrated communication, ensuring reliable ICV synchronization.
Moreover, satellites are used as relay stations to transmit V2X signals, which are then forwarded to ground receivers. Advanced coding techniques and optimized data packet transmission methods are employed to reduce signal transmission delays. SDN and NFV technologies are applied in the V2X system to achieve flexible scheduling and optimization of network traffic, thus reducing signal transmission delays. Multiple satellites are deployed and work together to achieve global coverage and high availability, ensuring V2X communication worldwide with minimal transmission delays. The satellite’s high-altitude position allows it to cover a large geographical area, enabling long-distance communication. Additionally, the use of radio waves for transmission in satellite communication results in a fast propagation in space, thus reducing transmission delays. Given the real-time requirements of V2X communication, the optimization of signaling processes between satellites and ground receivers is essential to minimize unnecessary signaling interactions and transmission delays. In summary, satellite communication systems present an effective solution for V2X communication, providing global coverage, high availability, and minimal transmission delays through the use of advanced technologies and optimization techniques.
3.2.4 Uncertainty
Uncertainty can lead to unpredictable traffic patterns and compromised safety. It can hinder efficient resource allocation and adaptive responses, resulting in traffic congestion and increased travel times, and potentially disrupting real-time decision-making and coordination. In ITS, the unpredictable rewards from cooperative perceptual scheduling among vehicles necessitate algorithms like MASS [64], which predict gains using the restless multi-armed bandit (RMAB) theory to cope with the uncertainty of rewards. Ambiguities introduced by the reliability of varying vehicle data, addressed by map-based frameworks [53].
3.2.5 Route Planning
Route planning addresses challenges posed by urbanization, vehicular congestion, delayed responses, and potential malicious threats. Efficient route planning ensures optimal transportation flow, minimizing computational wastage and enhancing system efficiency. Innovations such as real-time route planner [71], T-PORP [72] that integrates machine learning, and game-theory [73] inspired strategies represent a push towards data-driven, intelligent route planning. The quantum approximate optimization algorithm (QAOA) is presented by Azad et al.[74] for solving the vehicle routing problem.
3.3 Post-communication
The post-communication stage includes improving communication results, alleviating or eliminating the influence of abnormal data from other agents caused by errors or problems in the previous communication process, such as heterogeneous information, temporal asynchrony, localization and coordination errors, and lossy communication ( Table 8 ).
3.3.1 Heterogeneous Information
In V2V scenarios, homogeneous information refers to data or information that is identical or similar across vehicles or devices. However, V2X systems often involve heterogeneous information formed by infrastructure and autonomous vehicles (AVs), which refers to data or information that is diverse or varied across vehicles or devices. The configuration discrepancies between infrastructure and vehicle sensors make the design of a V2X perception system challenging [30].
To solve the challenge of heterogeneity in V2X perception systems, Xu et al.[30] proposed a robust cooperative perception framework with V2X communication using a novel vision transformer called V2X-ViT. Heterogeneous multi-agent self-attention is a key module in the proposed V2X-ViT framework for cooperative perception in autonomous vehicles. It is designed to handle the heterogeneity challenge in V2X perception systems, where different agents have different sensing capabilities and modalities. This module captures inter-agent interaction and per-agent spatial relationships by attending to the features of all agents and fusing them into a single representation.
3.3.2 Temporal Asynchrony
The temporal asynchrony refers to the asynchronous timestamps between the data captured from vehicle sensors and those received from other agents. This can be caused by different sensor initialization time points and the communication delay from the infrastructure to the vehicles [65]. Meanwhile, asynchronous sensor measurements can introduce lagged sensing information [30]. To tackle this problem, Yu et al.[9] proposed TCLF, which estimates the velocity of objects using two adjacent infrastructure frames and estimate the state of infrastructure objects, and then fuse the estimated infrastructure predictions and vehicle predictions using the LiDAR late fusion baseline method [78]. Furthermore, Yu et al.[65] proposed FFNet for VIC3D object detection by utilizing feature flow to transmit compressed intermediate data with critical perceptive information. To further improve the generalization ability in real-world scenarios, a latency-aware collaborative perception system is proposed [34], which leverages feature-attention symbiotic estimation and time modulation techniques to achieve feature-level synchronization. Furthermore, a curriculum learning technology is used to gradually increase the latency time by randomly sampling the latency time with an exponential distribution to accommodate the flexible communication latency. Additionally, Xu et al.[30] proposed a linear projection which warps the learnable embedding to generalize unseen time delay by temporally aligning the features beforehand.
3.3.3 Localization and Coordination Errors
The GPS localization noises can cause the undesirable coordinate transformation, thus leading to errors of localization transmitted to ICVs [30]. The accuracy of transformation matrices is estimated from sensor measurements and the loss or deliberate revision of location and pose data is transmitted via the network; this can lead to the failure of cooperation and serious road safety consequences [75]. To solve this problem, Song et al.[32] proposed the optimal transform estimation, which estimates a correction transform from the matched object pairs and further applies it to the noisy relative transform, which fuses the corrected relative transforms and maps the environment dynamically by global fusion and dynamic mapping. Moreover, in order to guarantee the accuracy of the coordinated localization data as well as the efficiency of the transmission, Song et al.[75] proposed an efficient and robust object-level communication framework involving broadcasting and receiving data of the 3D bounding boxes, location, and pose between connected vehicles. Two matching algorithms based on iterative closest point (ICP) and optimal transport theory are developed to maximize the total correlations between the 3D bounding boxes jointly detected by the vehicles.
3.3.4 Lossy Communication
Lossy communication (LC) can be caused by various factors which can result in incomplete or inaccurate shared intermediate features, thus compromising the effectiveness and efficiency of V2X cooperative perception. Previous research has identified common V2X communication failures, including crowded channels, malicious attacks, hardware failures, and frequency errors [29]. However, most cooperative perception algorithms assume ideal communication and there is a lack of analysis on the impact of lossy communication on feature sharing [76]. To alleviate the impact of lost information, Li et al.[76] proposed a novel approach, including an LC-aware repair network (LCRN) and a V2V-AM, to mitigate the detrimental influence and enhance the interaction between vehicles. In Ref.[6], a novel interruption-aware robust cooperative perception (V2X-INCOP) solution for V2X communication-aided autonomous driving was presented, by utilizing historical information to restore lost information caused by interruption. Moreover, Nguyen et al.[82] proposed a target tracking engine that tracks the position and movement trajectory of the target vehicle based on physical signals. The RSSI-based verification is then used to verify the presence of the target vehicle at its claimed location. Besides, Chen et al.[77] suggested a solution that involves estimating the carrier frequency offset (CFO) by performing autocorrelation between two adjacent identical primary synchronization signal (PSS) symbols and two adjacent identical secondary synchronization signal (SSS) symbols.
4 Senario Analysis
Less
Quality of service (QoS) requirements vary in different communication scenarios. As shown in Fig. 6, four application scenarios are analyzed in this section. Without researching scenario features, misunderstandings would arise in vehicle interactions, leading to poor communication performance, suboptimal resource use, and unmet communication needs.
4.1 Traffic Control
Traffic control is crucial for ICVs regarding safety, efficiency, and reliability. It addresses real-world navigation complexities through robust control methods and machine learning predictions for accurate path tracking. Optimized routing algorithms guide vehicles efficiently, especially in dynamic urban settings, while synchronization ensures timely and reliable data transmission, which is essential for resource allocation and real-time communication.
4.1.1 Event Prediction and Path Tracking
Navigating real-world traffic involves complex challenges such as steering delays, tracking errors, and uncertainties that can affect the ability of following the intended path. Saraiva et al.[36] tackled this problem by using a machine learning engine (MLE) to predict new events based on real-time network packet data. Xu et al.[67] focused on minimizing the impact of communication and steering delays through a preview steering control design, aiming at smoother and more accurate path tracking for AVs. To address vehicle parameter uncertainties and external disturbances, Yao et al.[37] employed robust control strategies, integrating sliding mode control (SMC), H-Control, and robust model predictive control (MPC). Moreover, Sun et al.[83] proposed a MPC path tracking controller with on-off tracking error for autonomous vehicle path tracking.
4.1.2 Route Planning
Route planning in ICV systems aims to guide the efficient vehicular movement between starting and ending points, and its complexity rises in multi-destination scenarios [84]. To address the challenges posed by dynamic urban traffic, recent studies propose different planning algorithms. To illustrate, optimized route planning in multi-destination contexts is tackled using schemes such as Ref.[18] and Ref.[85]. To account for real-time changes in traffic conditions, various optimization algorithms such as the improved ant colony algorithm using congestion avoidance (IACACAA)[86], quantum approximate optimization algorithm (QAOA)[74], and strength Pareto evolutionary algorithm (SPEA)[87] are introduced. In addition, Li et al.[72] introduced a T-PORP model for route planning based on double-layer grid (DLG) index in the face of urban dynamic road network. Al-Hasan et al.[71] proposed an intelligent route planning method for fast AVs driving in large natural terrain.
4.1.3 Information Distribution
Information distribution refers to the process of passing and distributing information within vehicles or between vehicles and external systems. Facing the dynamic traffic network environment, failures such as base station malfunctions and transmission link interruptions may occur. To ensure reliable information distribution, the real-time messaging, corresponding fault recovery mechanisms and backup plans are implemented for continuous information delivery. Mosavat-Jahromi et al.[88] proposed a network coding-based distributed media access control protocol (NC-MAC), which beacons retransmission and network coding, and uses preamble mechanism in the frame structure to report negative acknowledgement (NACK). Moreover, Rasheed et al.[89] proposed an intelligent beam control and a stable routing scheme.
4.1.4 Synchronization
Synchronization implements clock signals, protocols, or other methods to ensure that the transmission and reception of data are consistent at the appropriate time, so as to satisfy the reliability, security, and efficiency demands of ICV communication. GNSS is the commonly used synchronization source and the V2X self-synchronized protocol (SSPV)[46] is presented facing signal-shielding scenarios. Moreover, delay and interference can cause a time difference between the sending and receiving end of messages and lead to signal distortion. A mobile cloud and edge computing delay prediction framework [38] is proposed to improve real-time communication, and frameworks such as Ref.[90] and Ref.[91] are proposed to minimize mutual interference among V2X communication. Additionally, frequency errors that the sender generates and transmits data at a different rate than the receiver, can lead to intersymbol interference, resulting in the different symbol duration than expected by the receiver. To solve this problem, Hu et al.[81] mentioned how both integer and fractional parts of the frequency error can be compensated using short and long training sequences. Furthermore, the synchronous mechanism used to allocate communication resources to vehicles can cause synchronous overhead. Multiple vehicles may request the use of communication resources at the same time, thus the resource allocation should be synchronized to avoid conflicts and confusion. A geographically-based novel scheduling scheme [92] and a new optimization framework [93] are introduced to optimize the synchronization resource coordination and minimize the total transmission power.
4.2 Information Service
Information service is crucial for ICVs to optimize safety, efficiency, and user experience. Real-time data from V2V and V2I communications can support collision avoidance, adaptive speed control, and route optimization by providing timely updates on road conditions, traffic patterns, and potential hazards. They also offer enhanced user services such as navigation assistance, location-based services, and in-vehicle entertainment options, ultimately contributing to the overall acceptability and mass adoption of ICVs.
4.2.1 Communication and Information Sharing
In the realm of V2V communication, research focuses on two core areas: the dynamics of traffic and information flow, and network architecture. Du et al.[94] used mathematical models to assess delays in information propagation, incorporating real-world traffic conditions like congestion and sparsity.
To achieve information sharing, notable advancements have been made. In parallel, Manogaran et al.[95] introduced the information-centric content management framework (ICMF), designed to streamline data acquisition, analysis, and management, thereby optimizing information use with minimal time and overhead.
In the context of space-air-ground integrated networks, a unified scenario combining SDN and network function virtualization (NFV) has been proposed for ICV services across multiple domains: terrestrial, aerial, and space. A space-air-ground integrated network (SAGIN) incorporates low earth orbit (LEO) satellites, unmanned aerial vehicles (UAVs), and ground base stations. LEO satellites establish extensive wireless connections, UAVs use directional wireless communication, and ground stations interconnect through wired links. Coordinating these diverse nodes, an SDN controller at the satellite ground station (SGS) offers centralized oversight. Additionally, Li et al.[96] examined the dynamic mapping and scheduling of virtual network functions (VNFs) within this SAGIN framework. They proposed two Tabu search-based algorithms to solve a mixed-integer linear programming problem that includes specific cost and delay models. This work further enriches the growing body of knowledge on vehicular network optimization and orchestration.
4.2.2 Assistance and Mapping
For proactive traffic safety management (PTSM), the data on vehicle interactions is vital, such as spacing and time-to-collision (TTC). In-vehicle warning systems operate in a four-step framework. Step 1: Collects vehicle interaction data; Step 2: Involve in-car processing for crash risk prediction; Step 3: Integrate a human-machine interface to relay warnings; and Step 4: Disseminate these warnings via V2V and V2I channels. Following this, Jo et al.[97] developed a strategy that anticipates crash risks based on ICV communication data.
In the realm of mapping, current AVs are primarily designed either for lane following on structured highways or urban navigation supported by detailed, manually annotated global maps. These AVs rely on a combination of GPS, inertial navigation systems (INS), and high-definition (HD) maps for localization and route planning. While GPS provides basic location data, INS mitigates GPS limitations, and HD maps offer precise 3D navigation data. To supplement this, technologies like dedicated short-range communication (DSRC)[81] and emerging LTE-V and 5G networks [89] provide real-time road and infrastructure data, further aiding in route planning and map development.
4.3 ICV Safety and Security
In the field of ICV safety, the main concern is how to improve the safety of intelligent connected vehicles by technical means. This includes the safety performance of the vehicle itself, such as body structure, safety configuration, as well as the safety of the vehicle in the process of operation, such as to avoid collision and reduce accident losses [17]. ICV security includes vehicle and network security. As ICVs rely heavily on communication, a single security breach could compromise not only the personal data of passengers but also the functionality of the vehicle itself, leading to severe security risks [29]. Cyberattacks could disrupt real-time decision-making, mislead autonomous driving systems, or compromise route planning and traffic control, thereby endangering lives and infrastructure.
4.3.1 Collision Avoidance
For vehicle and personal safety, collision prevention is a mechanism created to avert accidents by identifying potential hazards and providing appropriate alerts to drivers or executing autonomous maneuvers to evade or reduce collisions [29].
As for traffic supervision and warning, a supervisor system receiving vehicle updates, predicting and addressing collisions is necessary, which helps them receive early warning information for effective rescue and traffic evacuation efforts [98]. Communication technologies such as DSRC, visible light communication (VLC) and 5G are suitable for detecting and reporting unusual driving actions in safety-related applications.
As for cooperative collision avoidance, Cheng et al.[99] proposed the virtual-field flow method-based (VFFM) model to enhance lane-keeping and collision avoidance for ICVs when obstacles are inevitable. By encoding LiDAR information into a point-based compact representation, COOPERNAUT uses cross-vehicle perception for vision-based cooperative driving [62], making more informed driving decisions when line-of-sight perception is limited. In response to the interference deduction problem, Huang et al.[90] introduced a centralized framework to minimize mutual interference between automotive radars using V2X communication. Furthermore, a Scheduling Algorithm for V2X Communication based on NOMA with Interference Mitigation (SAVCN-IM) and a novel geolocation-based scheduling scheme for C-V2X networks are proposed by Refs.[91,92], enabling vehicles to autonomously select wireless resources based on the positions and ordering of neighbouring vehicles on the road.
4.3.2 Attack Prevention
As vehicles become increasingly interconnected, they are vulnerable to unauthorized access, data breaches and safety threats. Taslimasa et al.[4] categorized security threats in ICV networks into two main categories: inter-vehicle and intra-vehicle attacks. Inter-vehicle attacks are usually caused by vulnerabilities in the communication between vehicles or between vehicles and road infrastructures, while the vulnerabilities in the communication between electronic control units (ECUs) and sensors inside a vehicle can result in intra-vehicle attacks. Both attacks can pose great threats to the security of the ICV network.
IDS is a security mechanism that monitors network traffic to detect malicious behavior that threatens the confidentiality, integrity, availability, and authenticity of ICV networks [4]. In ICV networks, the intrusion detection is commonly performed using traditional machine learning (ML) algorithms. Therefore, building a robust IDS for different states requires more accurate ML algorithms. Liu et al.[100] proposed a privacy-preserving two-layered distributed machine learning framework, involving forming vehicle clusters, where each cluster has a road-side unit (RSU) that acts as a local aggregator. The framework uses cryptographic tools and techniques to preserve privacy, protect identities and trajectories of vehicles, and handle packet loss in the application layer. As the supplementary, Kumar et al.[101] proposed IntelligentChain, a blockc hain and ML-based edge server for the ICV system. IntelligentChain enables low-latency ICV device registration and secure traffic accident information communication between them. Moreover, Q-learning-based strategies [102] have been employed to combat eavesdropping and jamming attacks in dynamic wireless networks. Yu et al.[103] further underscored the vulnerabilities by designing the IoV-SMAP protocol specifically to bolster security and provide user anonymity.
By leveraging deep learning (DL) techniques, Oseni et al.[104] proposed an explainable deep learning-based intrusion detection framework that employs a SHapley additive exPlanations (SHAP) mechanism to improve the transparency and resiliency of DL-based IDS. Moreover, to alleviate the black box nature of deep learning models used in IDS, the requirement for powerful processing capabilities, and to meet the need for effective solutions to protect against various types of cyber-attacks in different networks, Almutlaq et al.[105] proposed a two-stage IDS, by combining the advantages of rule extraction technique and two-stage IDS architecture to resource consumption and improve classification accuracy.
To tackle the problems caused by the complexity of new types of threats in the ICV environment and the DL-IDS model dependence on computational resources and datasets, transfer learning is leveraged to convert knowledge from the source task to the target task, making it more efficient in detecting attacks. Haddaji et al.[106] proposed a deep transfer learning based intrusion detection in-vehicle (TRLID) model for the ICV using the CAN bus protocol, which involves data preparation, transfer learning, and evaluation. Additionally, Otoum et al.[107] proposed a security framework for the ICV network to detect intrusions and secure intra/inter-vehicular communications. The proposed framework uses multi-task transfer learning to transfer knowledge gained from two different benchmark datasets, achieving satisfying performance and reduced training and fine-tuning time for the target domains. Similarly, to tackle the vulnerabilities of the ICV as well as the detection speed, Yang et al.[108] proposed a transfer learning and ensemble learning-based IDS for ICV systems using convolution neural networks (CNNs) and hyper-parameter optimization techniques to boost its efficiency as well as its performance.
4.4 Smart City
ICV is implemented in smart cities to ensure smoother traffic flows, reduce congestion, and heighten road safety through seamless integration and real-time data exchange. In this scenario, vehicles can receive updates about traffic conditions, alternative routes, and potential hazards, ensuring better navigation and less accidents. Furthermore, by reducing idle times at traffic signals, vehicles can significantly reduce their carbon footprint.
4.4.1 Traffic Flow Prediction and Control
Integrating traffic flow prediction and control with ICV is pivotal for making urban transportation more efficient, safe, and sustainable. Firstly, as 5G heterogeneous networks (HetNets) integrate varying cell types to cater to expansive urban landscapes, they necessitate agile resource management and accurate traffic prediction. To deal with this, Ref.[109] merges 5G load balancing with AlexNet to allocate resource accordingly. Secondly, as modern ride-sharing platforms optimize urban mobility by efficiently pairing riders, they concurrently reduce traffic congestion and environmental impact. Yet, the bane of traffic congestion, intensified by a myriad of factors ranging from infrastructural inadequacies to unpredictable incidents, underscores the urgency for advanced traffic management solutions. Models harnessing deep learning, such as the EC-DCRFNN [110] and PAG-TSN [111], offer stellar predictive accuracy, highlighting the potential of data-driven approaches in managing urban traffic.
4.4.2 Block Chain and Digital Twin
The integration of block chain and digital twin (DT) technologies offers transformative benefits for smart cities. On the one hand, the virtual simulations of DT leverages real-time and historical data, permitting in-depth monitoring, analysis, and prediction of traffic conditions, enabling proactive management [19]. This is instrumental not only for visualizing and making data-driven decisions concerning traffic patterns, but also for preemptively testing autonomous vehicle behaviors and addressing infrastructure and safety challenges. On the other hand, the blockchain’s decentralized nature ensures data integrity, security and transparency, as evidenced by architectures like DTaaS [112], which dynamically matches transportation needs with DT capabilities. Combined, these technologies foster a safe, efficient, and responsive transportation ecosystem for smart cities.
4.4.3 Parking Solution
As the vehicle numbers surge, efficient parking solutions become vital to combat traffic bottlenecks, reduce pollution, and ensure driver convenience. Traditional parking inefficiencies, marked by time-consuming searches for spaces, contribute significantly to urban congestion. However, with the advent of the ICV, technology-driven parking solutions in Refs.[113-115] employ advanced prediction models and real-time data collection to seamlessly guide drivers to available spots. These systems, ranging from graph neural networks to Internet of Things (IoT) integrated platforms, not only decrease the stress of finding a spot, but also aid in pre-emptive reservation, optimal route planning, and even surveillance. By integrating these solutions, urban transportation infrastructures can be vastly optimized, enhancing the overall driving experience while simultaneously promoting sustainable urban mobility.
4.4.4 Data Collection and Processing
Efficient data collection and processing stand as the backbone of urban mobility optimization. By leveraging tools like sensors, cameras, and IoT devices, cities accumulate vast amounts of real-time data, offering crucial insights into evolving traffic patterns, congestion points, and public transport efficiency. Such granular information, when processed through advanced analytics such as big data, as shown in Ref.[116], facilitates dynamic traffic management, rerouting, and public transportation optimization. However, with this data influx, ensuring robust security measures, as highlighted in Refs.[117,118], becomes paramount to protect user information, establish trust, and ensure the seamless functioning of the integrated transport communication system.
4.4.5 Resource and Energy
The judicious allocation of resources and energy underpins the resilience and efficiency of transportation communication systems. As underscored on the deployment of mobile EVs (MEVs) to recharge sensor nodes [119], ensuring the longevity and efficacy of communication networks is paramount. Coupled with trust evaluation mechanisms, this approach augments data reliability and network performance. Concurrently, the ChaseMe framework [120] emphasizes the value of innovative solutions like the Harris Hawk Optimization in efficiently managing EV charging stations, thereby ensuring enhanced driver experiences and reduced congestion. Collectively, these developments elucidate the pivotal role of resource and energy management in the interconnected matrix of smart cities and the ICVs.
5 Datasets
Less
The perception task of ICVs plays a pivotal role in communication, which enables them to obtain and share sensing data in real time, and realizes information exchange and collaboration between vehicles and other traffic participants. In this section, datasets are classified and analyzed as single view and multi view. Single view includes single vehicle view, drone view and infrastructure view. The multi view is mainly about V2X which includes $\mathrm{V}2\mathrm{\;V}$ and $\mathrm{V}2\mathrm{I}$ scenarios. The classification and evaluation of the datasets are shown in Table 9.
5.1 Single View
5.1.1 KITTI
Reference [121] is the first authoritative dataset widely used for ICVs. The data is collected from vehicles equipped with multiple cameras, sensors, GPS & IMU and then synchronized and calibrated using the method mentioned in Ref.[121]. The team has established a new benchmark: KITTI Vision Benchmark Suite [133], which focuses on the tasks including stereo matching, optical flow estimation, visual odometry, simultaneous localization and mapping (SLAM), 3D object detection and orientation estimation. Additionally, a variety of evaluation metrics are proposed, such as the extended translation metric, the extended ratation error, and the average orientation similarity (AOS).
5.1.2 nuScenes
Reference [122] is the first dataset using the full sensors, including cameras, radars, lidars, GPS and IMU. The dataset synchronizes the data collected from different sensors and provides semantic maps and annotations. It supports multiple tasks including 3D object detection, tracking and localization. For the detection task, it proposes True Positive metrics. For tracking task, nuScenes uses traditional metrics multiple object tracking accuracy (MOTA), multiple object tracking precision (MOTP)[134] and two new metrics: tracking initialization duration (TID) and longest gap duration (LGD). For detection task, the improved standards average multiple object tracking accuracy (sAMOTA) is the main metric.
5.1.3 Argoverse
Reference [124] contains precise map information: vector map of lane geometry, rasterized driveable area map and rasterized ground height map. It has the number of sensors roughly twice than KITTI’s and nuScenes’s and supports multiple tasks including 3D object tracking and forecasting. For the tracking task, argoverse provides precise map data (driveable area, ground removal, lane direction) to support this task. The evaluation metrics used for the tracking task on corresponding benchmark are MOTA and MOTP. For the froecasting task, the evaluation metrics are average displacement error (ADE) and final displacement error (FDE).
5.1.4 Apolloscape
Reference [125] contains large amounts of labeled data. The team uses Riegl VMX-1HA as its acquisition system, and the scanners are capable of acquiring point clouds with high accuracy and precision. LiDAR, GPS and IMU are also used to collect data. After synchronized, the following processes are followed in turn to label the data: moving object removal, 3D labelling, splatting projection. It supports multiple tasks including semantic segmentation, instance segmentation, lane segmentation and self-localization. The team proposes a joint localization and segmentation algorithm with 3D map.
5.1.5 Waymo Open(i) and Waymo Open Motion(ii)
Reference [123] and Ref.[126] are different datasets proposed by the same team. Ref.[123] supports 2D & 3D object detection and tracking task. Ref.[126] only supports motion prediction task. Based on Ref.[123], Ref.[126] added motion data. In Ref.[123], for 2D & 3D object detection and tracking task, the corresponding evaluation metrics are
average precision (AP)[135], average precision weighted by heading (APH, new metric), MOTA and MOTP. In Ref.[126], for the motion prediction task, the evaluation metrics used are minimum average displacement error (minADE), minimum final displacement error (minFDE), overlap rate (OR), miss rate (MR) and mean average precision (mAP). Based on Ref.[126], the team proposes a new AP evaluation metric according to MR.
5.1.6 HighD
Reference [127] is the first dataset collected by a drone. The team uses computer vision algorithms to extract the vehicles out of the scene and annotate them. Infrastructures are manually labeled. The dataset contains traffic scenarios from different types, including intersections, highways, and road fusions, etc.
5.1.7 NGSIM
Reference [129] is a large dataset proposed by the U.S. department of transportation for autonomous driving. Different from the datasets mentioned above, the NGSIM dataset is a simulation dataset, and the raw data was simulated from infrastructure perspective.
5.1.8 WIBAM
Reference [128] collects the data from the infrastructure view only via cameras. The vehicle data in each image in this dataset is automatically labeled to obtain the bounding box of the vehicle, and the test data in the WIBAM dataset is annotated manually. The main task supported by this dataset is 3D object detection and the evaluation metrics are average translation error (ATE), average scale error (ASE) and average orientation error (AOE). In Ref.[128], the team proposes a new loss function: multi-view loss. This loss function is constructed by combining two other loss functions: focal loss [136] and generalised intersection over union (GIoU)[137].
5.2 Multi View
5.2.1 OPV2V
Reference [7] is the first large scale dataset focusing on vehicle to vehicle perception. The raw data is simulated by a co-simulation tool OpenCDA [138], which is integrated with CARLA [139] and SUMO [140]. The main task focused on in this dataset is vehicle detection, and the evaluation metric is AP. Based on OPV2V, the team proposed a novel attentive intermediate fusion pipeline.
5.2.2 V2X-Sim
V2X-Sim is the the first synthetic V2X-aided collaborative perception dataset simulated by CARLA and SUMO. V2X-Sim 1.0 [130] is the earliest version, and the data came from LiDAR in the simulator. The supported task is 3D object detection, which is evaluated by AP. The team proposes distilled collaboration graph (DiscoGraph), a new framework: teacher-student framework and a matrix-valued edge weight in DsicoGraph. The student model in this framework is named distilled collaboration network (DiscoNet). The V2X-Sim 2.0 [131] is a subsequent version. It consists of simulated data from camera, LiDAR, and GPS in the simulator and is more multimodal. It also supports a variety of tasks, including multi-agent collaborative detection, multi-agent collaborative tracking, and semantic segmentation. For these tasks, the corresponding evaluation metrics are AP, higher order tracking accuracy (HOTA), MOTA, MOTP and mean intersection over union (mIoU).
5.2.3 DAIR-V2X
Yu et al.[9] proposed the first real-world dataset about vehicle-to-everything autonomous driving. The raw data is collected by infrastructure and vehicle sensors, and manually labeled. The DAIR-V2X contains three sub-datasets: DAIR-V2X-C (cooperative), DAIR-V2X-V (single vehicle), and DAIR-V2X-I (single infrastructure). The main task supported by the DAIR-V2X dataset is the VIC3D detection. This task involves the communication problem between different vehicles and between vehicles and buildings, which needs to consider the balance between performance and bandwidth. Therefore, AP and average byte measuring the transmission (AB) cost are used as evaluation metrics for this task. The team proposes a new framework: time compensation late fusion (TCLF) framework, to solve the communication delay challenge.
5.2.4 V2X-Seq
Reference [132] is the first large scale sequential V2X dataset from real world. It focuses on tracking and forecasting tasks, which contains: sequential perception dataset (SPD) and trajectory forecasting dataset (TFD). The SPD is built on the basis of DAIR-V2X-C. The main task is VIC3D tracking, the metrics for evaluation are MOTA, MOTP and byte per second (BPS). Tracking tasks supported by TFD include Online-VIC trajectory Forecasting and Offline-VIC trajectory Forecasting. For forecasting task, the evaluation metrics are minADE, minFDE and MR. Based on V2X-Seq, the team proposes new middle fusion method: FF-Tracking, which can effectively complete the VIC3D tracking task with low latency.
5.3 Summary
There are many limitations and challenges in the application of single view dataset in autonomous driving scenarios. The field of view provided by a single sensor is limited, and there may be blind spots that lead to the information loss, which is shown in Fig. 7. At the same time, there is an error in depth estimation due to the single view. It is difficult for a single sensor to accurately model the 3D environment, and at the same time, the reliability under the influence of adverse weather is low.
The multi-view dataset can effectively overcome the above challenges and limitations of the single-view dataset, offering potential improvements in detection and prediction performance [23], and is now widely used in ICV. At the safety level, the application of multi-view sensors can effectively capture the information around the vehicle, eliminate blind spots as Fig. 7 safety showed, reduce the information error obtained by single-view sensors, enhance situational awareness and threat detection [141]. At the efficiency level, multi-view sensors can effectively capture road characteristics, construct high-precision digital maps as Fig. 7 shows, effectively improve vehicle positioning accuracy, and efficiently plan paths. At the communication level, multi-view sensors cooperate with V2X communication devices as shown in Fig. 5. The information obtained by the sensors is shared with the infrastructure, such as traffic lights, so that the vehicle can receive real-time signals and adjust the driving state. The information obtained by the sensors is shared with other vehicles to improve the interaction between vehicles (Tables 10, 11).
6 Experiments
Less
6.1 Platform
The communication process in the ICV systems can be divided into three phases: pre-communication, during-communication and post-communication. In the pre-communication stage, the vehicle and the road-end collect information about traffic scenarios, and most of the experimental platforms use the CARLA simulator in this stage. In the during-communication stage, vehicles and other traffic participants interact with each other, and most of the experimental platforms use OpenCDA integrated by CARLA and SUMO. In the post-communication stage, the decision is made according to the collected and received data, and the experimental platform is the same as that in the during-communication stage. Commonly used V2X datasets are V2X-Sim and DAIR-V2X.
For perception tasks, the commonly used simulation tool is CARLA, and the sensors used for data collection are camera, lidar, GPS, etc. Deep learning algorithms are often used to process data, and the commonly used frameworks are TensorFlow and PyTorch. There are also many tools and libraries in perception tasks experiment, such as OpenCV for image processing, NumPy for data processing, and so on.
6.2 Tasks
Based on the dataset of ICVs, according to the tasks mainly supported by the dataset, the tasks can be classified into the following categories.
6.2.1 Detection
The task of object detection is to accurately detect and identify objects such as vehicles, pedestrians, bicycles, and traffic signs on the road using the data captured by onboard sensors (such as lidar, camera, radar, etc.) while the vehicle is driving. The corresponding reaction and decision is made after the detection result is obtained. The task of object detection is critical to the safety and reliability of autonomous driving systems, which can accurately identify different traffic scenarios, make decisions on the perception results, and ensure the safety and efficiency of vehicles.
6.2.2 Tracking
The object tracking task is to continuously track and identify other vehicles, pedestrians, bicycles and other traffic participants on the road using data captured by on-board sensors (such as lidar, camera and radar, etc.), while the vehicle is moving. In each frame of the data obtained by the sensor, the position and bounding box of the object are detected, and then the tracking algorithm and filtering technology are used to track and update the target. The results of object tracking tasks can provide the continuously changing position and motion information of the target object, which is helpful to sense the surrounding environment in real time.
6.2.3 Forecasting
The task of trajectory forecasting refers to predicting the movement trajectory and behavior intention of other traffic participants such as vehicles, pedestrians, bicycles and so on in the future. In the trajectory prediction task, information in the traffic senarios is obtained from on-board sensors such as lidar, camera, radar, etc. Then, using machine learning techniques, predictive models are built to predict the future movement trajectories and behaviors of traffic participants. Trajectory prediction can be used to make predictive decisions according to the prediction results, avoiding collisions and ensuring the safety and reliability of vehicle driving.
6.2.4 Segmentation
The task of semantic segmentation is to classify each pixel in the driving scenario and label it into different semantic classes, based on the data captured by on-board sensors, while the vehicle is driving. The acquired data is preprocessed and the features are extracted, and then the data is processed through deep learning. The predicted results are used to classify the data at the pixel level to realize the semantic segmentation. The semantic segmentation task can provide more accurate and detailed perceptual information, distinguish different road areas and obstacles, and provide safer decision-making and planning according to the results.
6.3 Evaluation Metrics
6.3.1 AP
$\mathrm{{AP}}$ is a commonly used evaluation metric in object detection tasks, which measures the accuracy and recall of object detection algorithms in predicting object categories and locations. The precision-recall curve is interpolated to calculate the area of the PR curve surrounded by the coordinate axis to obtain the value of AP. In multi-class object detection, the AP value of each class is calculated, and the AP value of all classes is averaged to obtain mAP, which is used to comprehensively evaluate the performance of the algorithm.
6.3.2 MOTA, MOTP, HOTP
MOTA, MOTP, and HOTA accuracy are commonly used evaluation metrics in multiple object tracking (MOT) tasks. In MOTA, a metric used to evaluate the accuracy of the object tracking algorithm over the entire tracking sequence. In MOTP, a metric used to evaluate the position accuracy of the object tracking algorithm in matching process. HOTA is a new target tracking evaluation index, which takes into account the accuracy of different matching stages (low order and high order), which fills the defects of traditional indicators that rely too much on matching.
6.3.3 ADE, FDE, MR
ADE, ADF and MR are commonly used to evaluate trajectory prediction performance metrics in the forecasting task. ADE evaluates the average position error between the predicted trajectory and the true trajectory. The Euclidean distance between all the correct points and the incorrect points is averaged to obtain the ADE. In ADF, there is a special form of ADE that only considers the error between the last point of the predicted trajectory and the last point of true trajectory. MR is used to evaluate the missing rate of predicted trajectory, which is the fraction of points in the predicted trajectory that do not cover the true trajectory.
6.3.4 mloU
$\mathrm{{mIoU}}$ is a commonly used evaluation metric in semantic segmentation tasks to measure the performance of semantic segmentation algorithms. In the semantic segmentation task, the pixels in the test dataset are first classified into different semantic categories. Then, according to the classification results predicted by the algorithm and the real annotation information, the ratio of the intersection area of the predicted category and the real category to the union area is calculated for each semantic category.
7 Summary and Future Perspectives
Less
The works evaluated and the methodologies described in this survey epitomize the contemporary research initiatives and the preliminary strides in elucidating cooperative communication within the realm of ICVs. As the field of autonomous driving technology continues to advance, V2X communication is gaining attention for its potential to enhance the safety and efficiency. Building upon the existing knowledge, several potential research directions are identified to further improve the efficacy and practicality of collaborative communication.
7.1 Optimize Collaborative Sensing
Collaborative sensing, which fosters effective data sharing and cooperation between ICVs and road infrastructure, stands as a pivotal aspect in overcoming the challenges posed by difficult conditions and complex scenarios. Future research endeavors should emphasize the development of advanced data fusion techniques, enabling seamless integration and synthesis of information from multiple sources, including ICVs, roadside sensors, and infrastructure. Such data fusion will lead to enhanced perception capabilities, enabling ICVs to obtain a more comprehensive and accurate understanding of the environment, even in challenging conditions such as adverse weather or occluded views. Moreover, the integration of heterogeneous networks and the reduction of communication delays will be pivotal in ensuring real-time and reliable data exchange, thereby fostering better collaboration between ICVs and the infrastructure, ultimately resulting in safer and more efficient autonomous driving.
7.2 Improve Communication Efficiency
Achieving efficient and high-performance communication is paramount for the success of autonomous driving systems. Future research directions should concentrate on innovative methods to increase communication accuracy and bandwidth, and reduce transmission delays. Advanced resource management strategies can be explored to allocate communication resources optimally, ensuring that ICVs and infrastructure efficiently utilize the available spectrum. Additionally, the adoption of advanced modulation and coding techniques can enhance data transmission rates, thereby increasing the overall communication bandwidth and capacity. Furthermore, research efforts can focus on developing low-latency communication protocols, reducing transmission delays and enabling real-time communication between ICVs and infrastructure. By addressing these key aspects, the communication efficiency of autonomous vehicles can be significantly improved, enabling safer and more responsive autonomous driving systems.
7.3 Enhance Communication Security
With the increasing reliance on V2X communication, ensuring the security and privacy of the transmitted data becomes paramount. Future research should focus on developing robust security mechanisms, encryption techniques, and authentication protocols to safeguard communication channels from potential cyber threats and unauthorized access. A secure communication infrastructure will instill trust in autonomous systems and promote widespread adoption.
7.4 Global Vehicle Networking Development
Standardization is crucial for the widespread deployment of ICVs and their seamless integration into existing transportation systems. Researchers should collaborate on establishing uniform protocols and standards for V2X communication globally. The development of a unified framework will facilitate interoperability between ICVs and various road infrastructure, promoting a more connected and efficient transportation ecosystem.
7.5 Dataset Enhancement
Comprehensive and diverse datasets are essential for the development and validation of V2X communication algorithms. Future research should focus on expanding existing datasets, with the particular emphasis on optimizing collaborative sensing and communication scenarios. These datasets should include various road conditions, weather scenarios, and real-world challenges to ensure the robustness and adaptability of communication systems under different environments, and explore to what extent collaborative communication can solve these problems.
Within the realm of 3D LiDAR autonomous driving applications, the substantial differences among the raw data furnished by various manufacturers pose a challenge in creating a unified dataset. A promising direction is to realize dataset synthesis and standardization, establishing a standardized dataset synthesis process to integrate data from different manufacturers, and processing and annotating the data according to unified standards to ensure data consistency and comparability. Most existing datasets are relatively limited, typically focusing only on perception tasks (traffic elements), while overlooking the topological relationships of the road. Thus, it is urgent to develop a dataset that can understand the structure of the road, the correlations between traffic elements and lane markings, and take topological relationships into account. Existing datasets for autonomous driving (AD) often lack diversity and long-range capabilities, focusing instead on ${360}^{\circ }$ perception and temporal reasoning, resulting in over-specialized solutions. A dataset covering a wide range of real-world driving scenarios as a completion to existing datasets is still to be investigated.
Acknowledgements This work was supported by the National High Technology Research and Development Program of China under Grant No. 2018YFE0204300, and the National Natural Science Foundation of China under Grant No. 62273198, U1964203, 52221005.
Declarations
Less
Conflict of interest The authors declare that they have no conflicts of interest toreport regarding the present study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
References
Less
1
Tan Z. , Dai N. , Su Y. , Zhang R. , Li Y. , Wu D. , Li S. . Human-machine interaction in intelligent and connected vehicles: a review of status quo, issues and opportunities. IEEE Trans. Intell. Transp. Syst. 23 (9): 13954-13975 (2022).https://doi.org/10.1109/TITS.2021.3127217
2
Sarker A. , Shen H. , Rahman M. , Chowdhury M. , Dey K. , Li F. , Wang Y. , Narman H.S. . A review of sensing and communication, human factors, and controller aspects for information-aware connected and automated vehicles. IEEE Trans. Intell. Transp. Syst. 21 (1): 7-29 (2020). https://doi.org/10.1109/TITS.2019.2892399
3
Matin A. , Dia H. . Impacts of connected and automated vehicles on road safety and efficiency: a systematic literature review. IEEE Trans. Intell. Transp. Syst. 24 (3): 2705-2736 (2023). https://doi.org/10.1109/TITS.2022.3227176
4
Taslimasa H. , Dadkhah S. , Neto E.C.P. , Xiong P. , Ray S. , Ghorbani A.A. . Security issues in internet of vehicles (iov): a comprehensive survey. Internet Things. 22 100809-100809 (2023). https://doi.org/10.1016/j.iot.2023.100809
5
Duan W. , Gu J. , Wen M. , Zhang G. , Ji Y. , Mumtaz S. . Emerging technologies for 5g-iov networks: applications, trends and opportunities. IEEE Netw. 34 (5): 283-289 (2020)
6
Ren S. , Lei Z. , Wang Z. , Dianati M. , Wang Y. , Chen S. , Zhang W. . Interruption-aware cooperative perception for V2X communication-aided autonomous driving (2023)
7
Xu R. , Xiang H. , Xia X. , Han X. , Li J. , Ma J. . Opv2v: an open benchmark dataset and fusion pipeline for perception with vehicle-to-vehicle communication.In: 2022 International Conference on Robotics and Automation (ICRA), pp. 2583–2589. IEEE (2022)
8
Kuang X. , Zhu H. , Yu B. , Li B. . Fast clustering for cooperative perception based on lidar adaptive dynamic grid encoding. Cogn. Comput. 16 (2), 546-565 (2024).
9
Yu H. , Luo Y. , Shu M. , Huo Y. , Yang Z. , Shi Y. , Guo Z. , Li H. , Hu X. , Yuan J. , Nie Z. . Dair-v2x: a large-scale dataset for vehicle-infrastructure cooperative 3d object detection.In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 21329-21338 (2022). https://doi.org/10.1109/CVPR52688.2022.02067
10
Ramli N.I. , Shawal S. , Ibrahim M.F. , Rawi M.I.M. . Network performance analysis of content delivery for v2r communication in vanet urban mobility.In: 2021 IEEE International Conference on Computing (ICOCO). pp.129-133 (2021). https://doi.org/10.1109/ICOCO53166.2021.9673560
11
Shin H. , Baldick R. . Plug-in electric vehicle to home (v2h) operation under a grid outage. IEEE Trans. Smart Grid. 8 (4), 2032-2041 (2017). https://doi.org/10.1109/TSG.2016.2603502
12
He Z. , Khazaei J. , Freihaut J.D. . Optimal integration of vehicle to building (v2b) and building to vehicle (b2v) technologies for commercial buildings. Sustain. Energy Grids Netw. 32, 100921-100921 (2022)
13
Hussein A. , Garcia F. , Armingol J.M. , Olaverri-Monreal C. . P2v and v2p communication for pedestrian warning on the basis of autonomous vehicles.In: 2016 IEEE 19th International Conference on Intelligent Transportation Systems (ITSC), pp. 2034–2039. IEEE (2016)
14
Yakusheva N. , Proletarsky A. , Basarab M. . Implemented vehicular communication systems: overview and comparative analysis.In: 2019 27th Telecommunications Forum (TELFOR), pp.1-4.IEEE(2019)
15
Zhou C. , Qian K. , Allan M. , Zhou W. . Modeling of the cost of ev battery wear due to v2g application in power systems. IEEE Trans. Energy Convers. 26 (4), 1041-1050 (2011). https://doi.org/10.1109/TEC.2011.2159977
16
Khelil A. , Soldani D. . On the suitability of device-to-device communications for road traffic safety.In: 2014 IEEE World Forum on Internet of Things (WF-IoT), pp.224–229. IEEE(2014)
17
Fu Y. , Li C. , Yu F.R. , Luan T.H. , Zhang Y. . A survey of driving safety with sensing, vehicular communications, and artificial intelligence-based collision avoidance. IEEE Trans. Intell. Transp. Syst. 23 (7), 6142-6163 (2022). https://doi.org/10.1109/TITS.2021.3083927
18
Hu W.C. , Wu H.T. , Cho H.H. , Tseng F.H. . Optimal route planning system for logistics vehicles based on artificial intelligence. J. Internet Technol. 21 (3), 757-764 (2020)
19
Feng H. , Chen D. , Lv Z. . Blockchain in digital twins-based vehicle management in vanets. IEEE Access. 9 100090-100100 (2021)
20
IEEE standard for low-rate wireless networks-amendment 1: Enhanced ultra wideband (UWB) physical layers (PHYS) and associated ranging techniques. IEEE Std 802.15.4z-2020 (Amendment to IEEE Std 802.15.4-2020), pp.1-174 (2020). https://doi.org/10.1109/IEEESTD.2020.9179124
21
IEEE standard for information technology-telecommunications and information exchange between systems-local and metropoli-tan area networks-specifi c requirements-part 11: Wireless LAN medium access control (MAC) and physical layer (PHY) specifi-cations amendment 3: Enhancements for very high throughput in the 60 ghz band. IEEE Std 802.11ad-2012 (Amendment to IEEE Std 802.11-2012, as amended by IEEE Std 802.11ae-2012 and IEEE Std 802.11aa-2012), pp.1-628 (2012). https://doi.org/10.1109/IEEESTD.2012.6392842
22
IEEE p2020 automotive imaging. IEEE P2020 Automotive Imag-ing, pp.1-32 (2018).
23
Weir N. , Lindenbaum D. , Bastidas A. , Etten A.V. , McPherson S. , Shermeyer J. , Kumar V. , Tang H. . Spacenet mvoi: a multi-view overhead imagery dataset.In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2019)
24
Xu Y. , Ge Y. , Wei D. , Yu R. , Wang J. , Yao Z. . V2v test scenario-study on intersection collision warning.In: 2021 IEEE 93rd Vehicular Technology Conference (VTC2021-Spring). (2021). 1-5
25
He K. , Li B. . Automotive v2x communication security key technology and test method research.In: 2022 7th International Conference on Cyber Security and Information Engineering (ICC-SIE). pp.40-43 (2022). https://doi.org/10.1109/ICCSIE56462.2022.00017
26
Zeng L. , Xie F. , Ye L. , Fang S. , Han Q. , Chen X. , Wang L. . C-v2x test data exchange progress and test interface definition.In: 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC). 2974-2979 (2022). https://doi.org/10.1109/ITSC55140.2022.9921772
27
Senkus B. , Soyturk M. . Multi-user real-time controllable connected cars testing platform.In: 2023 IEEE International Automated Vehicle Validation Conference (IAVVC). 1-6 (2023). https://doi.org/10.1109/IAVVC57316.2023.10328084
28
Chen D. , Yan Y. , Ye L. , Hu H. , Lei J. , Zeng L. . On-road features based in-chamber c-v2x application test scheme design.In: 2022 IEEE Smartworld, Ubiquitous Intelligence & Computing, Scalable Computing & Communications, Digital Twin, Privacy Computing, Metaverse, Autonomous & Trusted Vehicles (SmartWorld/UIC/ScalCom/DigitalTwin/PriComp/Meta). 2191-2197 (2022). https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00315
29
Sun X. , Yu F.R. , Zhang P. . A survey on cyber-security of connected and autonomous vehicles (CAVS). IEEE Trans. Intell. Transp. Syst. 23 (7), 6240-6259 (2022). https://doi.org/10.1109/TITS.2021.3085297
30
Xu R. , Xiang H. , Tu Z. , Xia X. , Yang M.-H. , Ma J. . V2X-ViT: vehicle-to-everything cooperative perception with vision transformer (2022)
31
Yang L. , Kaicheng Y. , Tang T. , Li J. , Yuan K. , Wang L. , Zhang X. , Chen P. . Bevheight: A robust framework for vision-based roadside 3d object detection.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 14167-14176 (2021).
32
Song Z. , Wen F. , Zhang H. , Li J. . A cooperative perception system robust to localization errors.In: 2023 IEEE Intelligent Vehicles Symposium (IV). 1-6 (2023).
33
Li J. , Xu R. , Ma J. , Zou Q. , Ma J. , Yu H. . Domain adaptive object detection for autonomous driving under foggy weather.In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 612-622 (2023).
34
Lei Z. , Ren S. , Hu Y. , Zhang W. , Chen S. . Latency-aware collaborative perception.In: Computer Vision -ECCV 2022. 316-332 (2022).
35
Wang Z. , Bian Y. , Shladover S.E. , Wu G. , Li S.E. , Barth M.J. . A survey on cooperative longitudinal motion control of multiple connected and automated vehicles. IEEE Intell. Transp. Syst. Mag. 12 (1), 4-24 (2020). https://doi.org/10.1109/MITS.2019.2953562
36
Vale Saraiva T. , Girardi R. , Campos C.A.V. . Machine learning applied to network traffic for vehicular applications to predict the vehicle context.In: 2021 IEEE 94th Vehicular Technology Conference (VTC2021-Fall). 1-5 (2021). https://doi.org/10.1109/VTC2021-Fall52928.2021.9625538
37
Yao Q. , Tian Y. , Wang Q. , Wang S. . Control strategies on path tracking for autonomous vehicle: state of the art and future challenges. IEEE Access. 8 161211-161222 (2020). https://doi.org/10.1109/ACCESS.2020.3020075
38
Zhang W. , Feng M. , Krunz M. , Volos H. . Latency prediction for delay-sensitive v2x applications in mobile cloud/edge computing systems.In: GLOBECOM 2020-2020 IEEE Global Communications Conference. 1-6 (2020).
39
Fan S. , Wang Z. , Huo X. , Wang Y. , Liu J. . Calibration-free BEV representation for infrastructure perception (2023)
40
Lu Y. , Li Q. , Liu B. , Dianati M. , Feng C. , Chen S. , Wang Y. . Robust collaborative 3d object detection in presence of pose errors. arXiv preprint arXiv:2103.01957. (2021)
41
Chen R. , Mu Y. , Xu R. , Shao W. , Jiang C. , Xu H. , Li Z. , Luo P. . Co³: cooperative unsupervised 3d representation learning for autonomous driving. arXiv preprint arXiv:2103.01946. (2021)
42
Wang T. , Chen G. , Chen K. , Liu Z. , Zhang B. , Knoll A. , Jiang C. . Umc: a unified bandwidth-efficient and multi-resolution based collaborative perception framework. arXiv preprint arXiv:2303.12400. (2023)
43
Wang Z. , Fan S. , Huo X. , Xu T. , Wang Y. , Liu J. , Chen Y. , Zhang Y.-Q. . Vimi: vehicle-infrastructure multi-view intermediate fusion for camera-based 3d object detection.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 14154-14163 (2021).
44
Su S. , Li Y. , He S. , Han S. , Feng C. , Ding C. , Miao F. . Uncertainty quantification of collaborative detection for self-driving. arXiv preprint arXiv:2209.08162. (2022)
45
Kumar S. , Tiwari P. , Abdou M. , Kamal H. . Sdc-net: end-to-end multitask self-driving car camera cocoon iot-based system. IEEE Internet Things J. (2021)
46
Elsharief M. , Abd El-Gawad M.A. , Kim H. . Sspv: self-time synchronization protocol for vehicular networks.In: 2020 International Conference on Electronics, Information, and Communication (ICEIC). 1-2 (2020).
47
Fan Z. , Song Z. , Xu J. , Wang Z. , Wu K. , Liu H. , He J. . Object level depth reconstruction for category level 6d object pose estimation from monocular RGB image.In: Computer Vision -ECCV 2022. 220-236 (2022).
48
Brucker M. . Unsupervised lidar-based 3d object detection using infrastructure sensors [D]. Technische Universität München (2022)
49
Liu Y.-C. , Tian J. , Ma C.-Y. , Glaser N. , Kuo C.-W. , Kira Z. . Who2com: collaborative perception via learnable handshake communication.In: 2020 IEEE International Conference on Robotics and Automation (ICRA). 6876-6883 (2020).
50
Hu Y. , Fang S. , Lei Z. , Zhong Y. , Chen S. . Where2comm: communication-efficient collaborative perception via spatial confidence maps. arXiv preprint arXiv:2209.12836. (2022)
51
Schwarzbach P. , Tauscher P. , Michler A. , Michler O. . V2x based probabilistic cooperative position estimation applying GNSS double differences.In: 2019 International Conference on Localization and GNSS (ICL-GNSS). 1-6 (2019).
52
Liu Y.-C. , Tian J. , Glaser N. , Kira Z. . When2com: multi-agent perception via communication graph grouping.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 4106-4115 (2020).
53
Jiang K. , Shi Y. , Wijaya B. , Yang M. , Wen T. , Xiao Z. , Yang D. . Map container: a map-based framework for cooperative perception (2022)
54
Shengyi X. , Xiaowen Tao S. , Xuan H. , Jiang J. , Du Zhu , B. Zhao , J. Deng , W. . Multi-sensor fusion positioning strategy for autonomous driving based on perturbation method. J Ambient Intell Humaniz Comput. 1-12 (2021).
55
Al J. , Hage J.A. , Xu P. , Bonnifait P. , Ibanez-Guzman J. . Factor graph-based dead reckoning for autonomous driving localization. IEEE Trans Intell Transp Syst. (2021)
56
Amodio A. , Ermidoro M. , Savaresi S. , Previdi F. . Automatic vehicle model recognition and lateral position estimation based on magnetic sensors. IEEE Trans. Intell. Transp. Syst. 16 (2), 1075-1086 (2015).
57
Zhang H. , Luo G. , Cao Y. , Jin Y. , Li Y. . Multi-modal virtual-real fusion based transformer for collaborative perception.In: 2022 IEEE 13th International Symposium on Parallel Architectures, Algorithms and Programming (PAAP). 1-6 (2022). https://doi.org/10.1109/PAAP56126.2022.10010640
58
Xue Y. , Zhang Y. , Liu Q. , Chen D. , Han K. . Comap: proactive provision for crowdsourcing map in automotive edge computing. IEEE Trans. Veh. Technol. (2021)
59
Thi T. , Le T. , Moh S. . Comprehensive survey of radio resource allocation schemes for 5g v2x communications. IEEE Access. 8 174-496174516 (2020).
60
Chang B. , Zhang L. , Li L. , Zhao G. , Chen Z. . Optimizing resource allocation in URLLC for real-time wireless control systems. IEEE Trans. Veh. Technol. (2019). https://doi.org/10.1109/TVT.2019.2930153
61
Cao B. , Sun Z. , Zhang J. , Gu Y. . Resource allocation in 5G Iov architecture based on SDN and fog-cloud computing. IEEE Trans. Intell. Transp. Syst. 22 (6), 3832-3840 (2021). https://doi.org/10.1109/TITS.2020.304
62
Cui J. , Qiu H. , Chen D. , Stone P. , Zhu Y. . Coopernaut: end-to-end driving with cooperative perception for networked vehicles.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 17252-17262 (2022).
63
Zhang P. , Wang C. , Aujla G.S. , Kumar N. , Guizani M. . Iov scenario: implementation of a bandwidth aware algorithm in wireless network communication mode. IEEE Trans. Veh. Technol. 69 (12), 15774-15785 (2020). https://doi.org/10.1109/TVT.2020.3035341
64
Jia Y. , Mao R. , Sun Y. , Zhou S. , Niu Z. . Mass: mobility-aware sensor scheduling of cooperative perception for connected automated driving. IEEE Trans. Intell. Transp. Syst. (2021)
65
Yu H. , Tang Y. , Xie E. , Mao J. , Yuan J. , Luo P. , Nie Z. . Vehicle-infrastructure cooperative 3D object detection via feature flow prediction. (2023)
66
Luan Z. , Zhang J. , Zhao W. , Wang C. . Trajectory tracking control of autonomous vehicle with random network delay. IEEE Trans. Veh. Technol. 69 (8), 8140-8150 (2020). https://doi.org/10.1109/TVT.2020.2995408
67
Xu S. , Peng H. , Tang Y. . Preview path tracking control with delay compensation for autonomous vehicles. IEEE Trans. Intell. Transp. Syst. 22 (5), 2979-2989 (2021).
68
Gu X. , Zhang G. , Ji Y. , Duan W. , Ding Z. , Ho P.-H. , Wen M. , Ho H. . Intelligent surface aided d2d-v2x system for low-latency and high-reliability communications. IEEE Trans. Veh. Technol. 70 (5), 4966-4977 (2021)
69
Liu H. , Chen S. , Leng L. . Zhou: protocol optimization for vehicular awareness message broadcasting in c-v2x networks. IEEE Trans. Veh. Technol. 69 (11), 12998-13011 (2020)
70
NBA G.P.W. , Haapola J. , Samarasinghe T. . A markov perspective on c-v2x mode 4.In: 2019 IEEE 90th Vehicular Technology Conference (VTC2019-Fall). 1-6 (2019)
71
Al-Hasan S. , Vachtsevanos G. . Intelligent route planning for fast autonomous vehicles operating in a large natural terrain. J. Field Robot. 27 (1), 1-20 (2010)
72
Li B. , Dai T. , Chen W. , Song X. , Zang Y. , Huang Z. , Lin Q. , Cai K. . T-porp: a trusted parallel route planning model on dynamic road networks. IEEE Trans. Intell. Transp. Syst. 24 (1), 1238-1250 (2023). https://doi.org/10.1109/TITS.2022.3216310
73
Dogra R. , Rani S. , Babbar H. , Verma S. , Verma K. . Deger: dynamic clustering green communication routing for intelligent transportation systems. Wireless Pers. Commun. 118 (2), 1075-1097 (2021)
74
Azad U. , Behera B.K. , Ahmed E.A. , Panigrahi P.K. , Farouk A. . Solving vehicle routing problem using quantum approximate optimization algorithm. IEEE Trans. Intell. Transp. Syst. 24 (7), 7564-7573 (2023). https://doi.org/10.1109/TITS.2022.3172241
75
Song Z. , Wen F. , Zhang H. , Li J. . An efficient and robust object-level cooperative perception framework for connected and automated driving. ArXiv:2210.06289. (2022)
76
Li J. , Xu R. , Liu X. , Ma J. , Chi Z. , Ma J. , Yu H. . Learning for vehicle-to-vehicle cooperative perception under lossy communication. IEEE Trans. Intell. Veh. 8 (4), 2650-2660 (2023). https://doi.org/10.1109/TIV.2023.3260040
77
Chen T. , Shen H. , Hu A. , He W. , Xu J. , Hu H. . Radio frequency fingerprints extraction for lte-v2x: a channel estimation based methodology.In: 2022 IEEE 96th Vehicular Technology Conference (VTC2022-Fall). 1-6 (2022)
78
Lang A.H. , Vora S. , Caesar H. , Zhou L. , Yang J. , Beijbom O. . PointPillars: fast encoders for object detection from point clouds. (2019)
79
Pan B. , Wu H. . Success probability analysis of cooperative c-v2x communications. IEEE Trans. Intell. Transp. Syst. 23 (7), 7170-7183 (2022). https://doi.org/10.1109/TITS.2021.3067048
80
Hasan M. , Mohan S. , Shimizu T. , Lu H. . Securing vehicle-to-everything (v2x) communication platforms. IEEE Trans. Intell. Veh. 5 (4), 693-713 (2020). https://doi.org/10.1109/TIV.2020.2987430
81
Hu J. , Chen S. , Zhao L. , Li Y. , Fang J. , Li B. , Shi Y. . Link level performance comparison between lte v2x and dsrc. J. Commun. Inf. Netw. 2 (2), 101-112 (2017)
82
Nguyen V.-L. , Lin P.-C. , Hwang R.-H. . Physical signal-driven fusion for v2x misbehavior detection.In: 2019 IEEE Vehicular Networking Conference (VNC). 1-4 (2019). https://doi.org/10.1109/VNC48660.2019.9062815
83
Sun C. , Zhang X. , Zhou Q. , Tian Y. . A model predictive controller with switched tracking error for autonomous vehicle path tracking. IEEE Access. 7 53103-53114 (2019). https://doi.org/10.1109/ACCESS.2019.2912094
84
Jiang F. , Wang Z. , Yue G. . A novel cognitively inspired deep learning approach to detect drivable areas for self-driving cars. Cogn. Comput. 16 (2), 517-533 (2024)
85
Basso R. , Kulcsar B. , Egardt B. , Lindroth P. , Sanchez-Diaz I. . Energy consumption estimation integrated into the electric vehicle routing problem. Transp. Res. 69 (APR.), 141-167 (2019)
86
Liu C. , Kou G. , Zhou X. , Peng Y. , Alsaadi F.E. . Time-dependent vehicle routing problem with time windows of city logistics with a congestion avoidance approach. Knowl.-Based Syst. 188 (2019)
87
Ghorai C. , Shakhari S. , Banerjee I. . A spea-based multimetric routing protocol for intelligent transportation systems. IEEE Trans. Intell. Transp. Syst. 22 (11), 6737-6747 (2021). https://doi.org/10.1109/TITS.2020.2994362
88
Mosavat-Jahromi H. , Li Y. , Cai L. , Lu L. . Nc-mac: network coding-based distributed mac protocol for reliable beacon broadcasting in v2x.In: IEEE Global Communications Conference. (2020)
89
Rasheed I. , Hu F. , Hong Y.-K. , Balasubramanian B. . Intelligent vehicle network routing with adaptive 3d beam alignment for mmwave 5g-based v2x communications. IEEE Trans. Intell. Transp. Syst. 22 (5), 2706-2718 (2021). https://doi.org/10.1109/TITS.2020.2973859
90
Huang J. , Fei Z. , Wang T. , Wang X. , Liu F. , Zhou H. , Zhang J.A. , Wei G. . V2x-communication assisted interference minimization for automotive radars. China Commun. 16 (10), 100-111 (2019). https://doi.org/10.23919/JCC.2019.10.007
91
Rigane E.b. , Marouane H. , Trabelsi A. , Zarai F. . Interference mitigation based on noma for scheduling algorithm in v2x environments.In: 2020 International Wireless Communications and Mobile Computing (IWCMC). 1952-1957 (2020). https://doi.org/10.1109/IWCMC48107.2020.9148397
92
Molina-Masegosa R. , Sepulcre M. , Gozalvez J. . Geo-based scheduling for c-v2x networks. IEEE Trans. Veh. Technol. 68 (9), 8397-8407 (2019). https://doi.org/10.1109/TVT.2019.2924698
93
Khan W.U. , Jamshed M.A. , Mahmood A. , Lagunas E. , Chatzinotas S. , Ottersten B. . Backscatter-aided noma v2x communication under channel estimation errors.In: 2022 IEEE 95th Vehicular Technology Conference: (VTC2022-Spring). 1-6 (2022). https://doi.org/10.1109/VTC2022-Spring54318.2022.9860382
94
Du L. , Dao H. . Information dissemination delay in vehicle-to-vehicle communication networks in a traffic stream. IEEE Trans. Intell. Transp. Syst. 16 (1), 66-80 (2015). https://doi.org/10.1109/TITS.2014.2326331
95
Manogaran G. , Saravanan V. , Hsu C.-H. . Information-centric content management framework for software defined internet of vehicles towards application specific services. IEEE Trans. Intell. Transp. Syst. 22 (7), 4541-4549 (2021). https://doi.org/10.1109/TITS.2021.3058452
96
Li J. , Shi W. , Wu H. , Zhang S. , Shen X. . Cost-aware dynamic sfc mapping and scheduling in sdn/nfv-enabled space-air-ground-integrated networks for internet of vehicles. IEEE Internet Things J. 9 (8), 5824-5838 (2022). https://doi.org/10.1109/JIOT.2021.3058250
97
Jo Y. , Jang J. , Ko J. , Oh C. . An in-vehicle warning information provision strategy for v2v-based proactive traffic safety management. IEEE Trans. Intell. Transp. Syst. 23 (10), 19387-19398 (2022). https://doi.org/10.1109/TITS.2022.3156923
98
Ahn H. , Rizzi A. , Colombo A. , Del Vecchio D. . Experimental testing of a semi-autonomous multi-vehicle collision avoidance algorithm at an intersection testbed.In: 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 4834-4839 (2015). https://doi.org/10.1109/IROS.2015.7354056
99
Cheng S. , Li L. , Liu Y.-G. , Li W.-B. , Guo H.-Q. . Virtual-field flow method-based lane-keeping and collision avoidance control for autonomous vehicles. IEEE Trans. Intell. Transp. Syst. (2021)
100
Liu R. , Pan J. . Crs: a privacy-preserving two-layered distributed machine learning framework for iov. IEEE Internet Things J. (2023). https://doi.org/10.1109/JIOT.2023.3287799
101
Kumar A. , Das D. . Intelligentchain: Blockchain and machine learning based intelligent security application for internet of vehicles (iov).In: 2022 IEEE 95th Vehicular Technology Conference: (VTC2022-Spring). 1-5 (2022). https://doi.org/10.1109/VTC2022-Spring54318.2022.9860946
102
Kato N. , Tang F. , Mao B. , Gui G. . Machine learning for vehicular networks: recent advances and future directions. IEEE Veh. Technol. Mag. 13 (3), 94-101 (2018)
103
Yu S. , Lee J. , Park K. , Das A.K. , Park Y. . Iov-smap: secure and efficient message authentication protocol for iov in smart city environment. IEEE Access. 8 167875-167886 (2020)
104
Oseni A. , Moustafa N. , Creech G. , Sohrabi N. , Strelzoff A. , Tari Z. , Linkov I. . An explainable deep learning framework for resilient intrusion detection in iot-enabled transportation networks. IEEE Trans. Intell. Transp. Syst. 24 (1), 1000-1014 (2023). https://doi.org/10.1109/TITS.2022.3188671
105
Almutlaq S. , Derhab A. , Hassan M.M. , Kaur K. . Two-stage intrusion detection system in intelligent transportation systems using rule extraction methods from deep neural networks. IEEE Trans. Intell. Transp. Syst. (2022)
106
Haddaji A. , Ayed S. , Fourati L.C. . A transfer learning based intrusion detection system for internet of vehicles.In: 2023 15th International Conference on Developments in eSystems Engineering (DeSE). 533-539 (2023). https://doi.org/10.1109/DeSE58274.2023.10099623
107
Otoum Y. , Wan Y. , Nayak A. . Transfer learning-driven intrusion detection for internet of vehicles (iov).In: 2022 International Wireless Communications and Mobile Computing (IWCMC). 342-347 (2022). https://doi.org/10.1109/IWCMC55113.2022.9825115
108
Yang L. , Shami A. . A transfer learning and optimized CNN based intrusion detection system for internet of vehicles.In: ICC 2022-IEEE International Conference on Communications. 2774-2779 (2022). https://doi.org/10.1109/ICC45855.2022.9838780
109
Yu M. . Construction of regional intelligent transportation system in smart city road network via 5G network. J. Adv. Transp. 2021. (2021).
110
Yang L. . Short-term traffic flow prediction model of 5G iov based on EC and dl algorithms. J. Ambient. Intell. Humaniz. Comput. 12 (9), 10015-10025 (2021)
111
Li J. , Lin F. , Han G. , Wang Y. , Yu R. , Oguti A. , Li Z. . Pag-tsn: ridership demand forecasting model for shared travel services of smart transportation. IEEE Trans. Intell. Transp. Syst. (2021)
112
Liao S. , Wu J. , Bashir A.K. , Yang W. , Li J. , Tariq U. . Digital twin consensus for blockchain-enabled intelligent transportation systems in smart cities. IEEE Trans. Intell. Transp. Syst. (2021)
113
Xiao X. , Peng Z. , Lin Y. , Jin Z. , Shao W. , Chen R. , Cheng N. , Mao G. , Kao M. . Parking prediction in smart cities: a survey. IEEE Trans. Intell. Transp. Syst. (2021)
114
Aljohani M. , Olariu S. , Alali A. , Jain S. . A survey of parking solutions for smart cities. IEEE Trans. Intell. Transp. Syst. 18 (12), 3186-3205 (2022).
115
Wang Y. , Li Y. , Li X. , Li Y. , Li Y. . Vision-based parking occupancy detection and management using edge computing techniques. Transp. Res. C Emerg. Technol. 112 102598-102598 (2020)
116
Ferreira D. , Astuto B. , Nunes A. , Vieira Campos C.A. , Obraczka K. . User community identification through fine-grained mobility records for smart city applications. IEEE Trans. Mob. Comput. 17 (8), 1928-1941 (2017)
117
Rawat D.B. , Singh A.K. , Chaouchi H. , Bonnin J.-М. . Ecc based privacy-aware secure aco-aodv routing protocol for internet of vehicles. IEEE Internet Things J. 6 (2), 1746-1755 (2019)
118
Mahmood K. , Ferzund J. , Saleem M. , Shamshad S. , Das A.K. , Park Y. , Kumar A. , Asad M. , Das A. . A provably secure mobile user authentication scheme for big data collection in iot-enabled maritime intelligent transportation system. IEEE Internet Things J. (2021)
119
Liu X. , Li X. , Li X. , Zhang Y. , Zhang X. . Mobile electric vehicle-assisted trust evaluation mechanism for wireless sensor networks. J. Ambient. Intell. Humaniz. Comput. 11 (11), 5105-5116 (2020)
120
Kumar N. , Chaudhry R. , Kaiwartya O. , Kumar N. . Chaseme: a heuristic scheme for electric vehicles mobility management on charging stations in a smart city scenario. IEEE Access. 9 101090-101103 (2021)
121
Geiger A. , Lenz P. , Stiller C. , Urtasun R. . Vision meets robotics: the kitti dataset. Int. J. Robot. Res. 32 (11), 1231-1237 (2013)
122
Caesar H. , Bankiti V. , Lang A.H. , Vora S. , Liong V.E. , Xu Q. , Krishnan A. , Pan Y. , Baldan G. , Beijbom O. . nuscenes: a multimodal dataset for autonomous driving.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). (2020)
123
Sun P. , Kretzschmar H. , Dotiwalla X. , Chouard A. , Patnaik V. , Tsui P. , Guo J. , Zhou Y. , Chai Y. , Caine B. , et al. Scalability in perception for autonomous driving: Waymo open dataset.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2446-2454 (2020)
124
Chang M.-F. , Lambert J. , Sangkloy P. , Singh J. , Bak S. , Hartnett A. , Wang D. , Carr P. , Lucey S. , Ramanan D. , et al. . Argoverse: 3d tracking and forecasting with rich maps.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 8748-8757 (2019)
125
Huang X. , Wang P. , Cheng X. , Zhou D. , Geng Q. , Yang R. . The apolloscape open dataset for autonomous driving and its application. IEEE Trans. Pattern Anal. Mach. Intell. 42 (10), 2702-2719 (2019)
126
Ettinger S. , Cheng S. , Caine B. , Liu C. , Zhao H. , Pradhan S. , Chai Y. , Sapp B. , Qi C.R. , Zhou Y. , et al. . Large scale interactive motion forecasting for autonomous driving: The waymo open motion dataset.In: Proceedings of the IEEE/CVF International Conference on Computer Vision. 9710-9719 (2021)
127
Krajewski R. , Bock J. , Kloeker L. , Eckstein L. . The highd dataset: a drone dataset of naturalistic vehicle trajectories on German highways for validation of highly automated driving systems.In: 2018 21st International Conference on Intelligent Transportation Systems (ITSC). 2118-2125 (2018)
128
Howe M. , Reid I. , Mackenzie J. . Weakly supervised training of monocular 3d object detectors using wide baseline multi-view traffic camera data. arXiv preprint arXiv:2110.10966. (2021)
129
Dot U. . Next generation simulation (NGSIM) vehicle trajectories and supporting data [R]. US Department of Transportation, (2018)
130
Li Y. , Ren S. , Wu P. , Chen S. , Feng C. , Zhang W. . Learning distilled collaboration graph for multi-agent perception.In: Adv. Neural. Inf. Process. Syst. 34, 29541-29552 (2021)
131
Li Y. , Ma D. , An Z. , Wang Z. , Zhong Y. , Chen S. , Feng C. . V2x-sim: multi-agent collaborative perception dataset and benchmark for autonomous driving. IEEE Robot. Autom. Lett. 7 (4), 10914-10921 (2022)
132
Yu H. , Yang W. , Ruan H. , Yang Z. , Tang Y. , Gao X. , Hao X. , Shi Y. , Pan Y. , Sun N. , et al. . V2x-seq: a large-scale sequential dataset for vehicle-infrastructure cooperative perception and forecasting.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 5486-5495 (2023)
133
Geiger A. , Lenz P. , Urtasun R. . Are we ready for autonomous driving? The kitti vision benchmark suite.In: 2012 IEEE Conference on Computer Vision and Pattern Recognition. 3354-3361 (2012)
134
Bernardin K. , Elbs A. , Stiefelhagen R. . Multiple object tracking performance metrics and evaluation in a smart room environment.In: 6th IEEE International Workshop on Visual Surveillance, in Conjunction with ECCV. 90, (2006)
135
Everingham M. , Van Gool L. , Williams C.K. , Winn J. , Zisserman A. . The pascal visual object classes (voc) challenge. Int. J. Comput. Vis. 88, 303-338 (2010)
136
Zhou X. , Wang D. , Krähenbühl P. . Objects as points. arXiv preprint arXiv:1904.07850. (2019)
137
Rezatofighi H. , Tsoi N. , Gwak J. , Sadeghian A. , Reid I. , Savarese S. . Generalized intersection over union: a metric and a loss for bounding box regression.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 658-666 (2019)
138
Xu R. , Guo Y. , Han X. , Xia X. , Xiang H. , Ma J. . Opencda: an open cooperative driving automation framework integrated with co-simulation.In: 2021 IEEE International Intelligent Transportation Systems Conference (ITSC). 1155-1162 (2021)
139
Dosovitskiy A. , Ros G. , Codevilla F. , Lopez A. , Koltun V. . Carla: an open urban driving simulator.In: Conference on Robot Learning. 1-16 (2017)
140
Olaverri-Monreal C. , Errea-Moreno J. , Díaz-Álvarez A. , Biurrun-Quel C. , Serrano-Arriezu L. , Kuba M. . Connection of the sumo microscopic traffic simulator and the unity 3d game engine to evaluate v2x communication-based systems. Sensors. 18 (12), 4399-4399 (2018)
141
Caesar H. , Bankiti V. , Lang A.H. , Vora S. , Liong V.E. , Xu Q. , Krishnan A. , Pan Y. , Baldan G. , Beijbom O. . nuscenes: a multimodal dataset for autonomous driving.In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). (2020)
Appendix
Less
Year 2025 volume 8 Issue 1
PDF
374
152
Cite this Article
BibTeX
Article Info
doi: 10.1007/s42154-024-00310-2
  • Receive Date：2024-01-02
  • Online Date：2025-07-21
Article Data
Affiliations
History
  • Received：2024-01-02
  • Accepted：2024-05-28
Affiliations
    1 Tsinghua University State Key Laboratory of Automotive Safety and Energy, School of Vehicle and Mobility Beijing 100084 China

Corresponding:

References
Share
https://castjournals.cast.org.cn/joweb/qccx/EN/10.1007/s42154-024-00310-2
Share to
QR

Scan QR to access full text

Cite this article
BibTeX
Citations
表12种不同金属材料的力学参数
科
Family		 属数
Number of
genus		 种数
Number of
species		 占总种数比例
Percentage of
total species (%)		 属
Genus		 种数
Number of
species		 占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏
  • BibTeX
  • EndNote
  • RefWorks
  • TxT
Links
Articles
Latest Articles
Most Read
Collections
Updates
Events
News
Multimedia
About
About Us
Contact
No. 86 Xueyuan South Road, Haidian District, Beijing
100081
010-62199257
qkjq@cast.org.cn

Copyright © 2025 China Association for Science and Technology. All rights reserved. For all open access content, the relevant licensing terms apply.
Sponsored by the Office of the Leading Group for Cybersecurity and Informatization of CAST, and supported by Science and Technology Review Publishing House