| Citation: | SHANGGUAN Wei, LI Xin, CHAI Lin-guo, CAO Yue, CHEN Jing-jing, PANG Hao-jie, RUI Tao. Research review on simulation and test of mixed traffic swarm in vehicle-infrastructure cooperative environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 19-40. doi: 10.19818/j.cnki.1671-1637.2022.03.002
|
Chinese urban road traffic is facing several problems such as frequent accidents, severe congestion, and environmental pollution, which make it difficult to meet the growing demand for transportation. It is urgent to improve the level of connectivity, intelligence, and collaboration in the transportation system. In this context, vehicle road collaboration technology has become an inevitable development trend in the field of road transportation, which is an effective means to ensure safety, improve efficiency, optimize energy consumption, and reduce emissions.
Vehicle road coordination system is the application of communication technology, artificial intelligence, big data, cloud computing and other new technologies in the transportation system. Based on wireless communication, sensor detection and other technologies, it collects and integrates dynamic traffic information throughout time and space. Through the interaction and sharing of vehicle road and vehicle road information, it achieves active safety control of vehicles and collaborative management of roads[1]The vehicle road collaboration system is a large-scale real-time distributed system involving multiple factors and complex multi-level relationships, and must be based on practical and feasible applications to effectively reflect its performance and value. Testing and verifying the vehicle road collaboration system is an essential part of the development process[2]The vehicle road collaborative simulation testing technology is a key means to accelerate the application and implementation of vehicle road collaborative systems. Looking at the development history of simulation testing for vehicle road collaboration systems both domestically and internationally, it can be divided into three stages: the embryonic stage, the initial stage, and the development stage.
(1) Sprout period (1986~2002). During this period, the actual road traffic environment was dominated by manually driven vehicles, and the safety of vehicle operation mainly relied on the driver's operation and experience. In order to improve the safety and convenience of road operations, the University of California, Berkeley launched the Partners for Advanced Transit and Highways (PATH) program in 1986, which first proposed the concept of vehicle road coordination system and focused on the research of vehicle following theory[3]In 1991, Japan launched the Advance Safety Vehicle (ASV) program[4]The Smartway Plan[5]We focused on researching vehicle road collaborative safe driving and intelligent traffic technology. In 1999, China established the National Intelligent Transportation Engineering Technology Research Center and began the overall layout of the vehicle road coordination system. In 2000, the CarTALK2000 project was launched in Europe[6]Focus on conducting research on driving assistance systems.
Traffic simulation originated in the 1960s, and the Traffic Signal Optimization Program (TSOP) model and Traffic Network Study Tool (TNST) model were typical macro models used for signal control optimization at that time[7]With the proposal of the concept of vehicle road collaboration and in combination with the characteristics of the road environment at that time, domestic and foreign experts and scholars have focused on conducting relevant research on macro and micro traffic flow simulations composed of a single subject, using simulation methods such as time scanning and event scanning[8]In terms of simulation models, British scholars Wada et al[9]Proposed the application of fluid dynamics theory to simulate traffic flow and derived the Payne model[10]、 Papageorgious model[11]、 Ross model[12]Wu Zheng Model[13]Waiting for macro traffic flow models; Pipes[14]Proposed the theory of vehicle following and developed models including cellular automata[15]、 Krauss model[16]Intelligent Driver Model (IDM)[17]Microscopic traffic flow model including. In the early stages, vehicle road collaborative simulation is mainly based on virtual simulation, and typical simulation software includes TRANSYT[18]And VISSIM[19]For virtual simulation software, simulation testing mainly relies on white box testing or black box testing, and there is a lack of research on simulation performance evaluation and simulation result analysis.
(2) Initial period (2003-2014). At this stage, with the rapid development of computer technology and wireless communication technology, vehicle road collaborative systems based on vehicle road and vehicle road information interaction have emerged. The United States, Japan, and Europe have successively launched IntelliDrive[20]Automated Highway System (AHS)[21]、 SafeSpot[22]Research project on communication based vehicle road collaboration technology. China focuses on the forefront of science and technology and attaches great importance to technological innovation at the national policy level. In 2011, the Ministry of Science and Technology established the first key technology research project for intelligent vehicle road collaboration in the "863 Program". In order to achieve the testing and verification of the technical performance and efficiency of the actual vehicle road collaboration system, a communication supported vehicle road collaboration simulation environment was built, and a digital vehicle road collaboration simulation platform with vehicle/vehicle road information interaction as the core was constructed, which promoted the further research and development of key technologies in vehicle road collaboration[23]Based on the "863 Program" of the Ministry of Science and Technology, a research team led by Tsinghua University and participated by 10 units has defined the Intelligent Vehicle Infrastructure Cooperative System (i-VICS) for the first time, and established a vehicle road collaboration simulation testing and verification platform based on High Level Architecture (HLA) and Multi Resolution Modeling (MRM)[24]The relevant technologies and achievements have formed the book "Framework of Intelligent Transportation System System Based on Vehicle Road Collaboration", reflecting the development process of China's intelligent transportation system in this stage.
At this stage, vehicles can obtain more extensive and comprehensive traffic status information based on communication, and achieve optimal behavior control between vehicles through vehicle road interaction. Simulation methods for behavior control such as vehicle following behavior simulation, overtaking behavior simulation, and lane changing behavior simulation are also becoming increasingly complex. Gipps[25]Assuming that the expected braking rate and acceleration rate of each vehicle subject have certain limitations, a new car following model was constructed for subsequent vehicle responses, reproducing the characteristics of real traffic flow; Petrov and others[26]A mathematical model and adaptive controller for automatic overtaking maneuver of automobiles have been proposed, which generates polynomial virtual trajectories for each stage in real time, enabling overtaking to track the expected trajectory even when the speed is unknown; Butakov et al[27]By learning the response characteristics of vehicles before and during lane changes in different driving environments, a two-layer model was established to describe its dependence on the configuration of surrounding vehicles, achieving communication based simulation of lane changing behavior; Cai Bogen and others[28]A multi-resolution information interaction method was proposed, and high-resolution vehicle driving status information models, medium resolution fleet status information models, and low resolution traffic flow information models were established. The simulation process of the HLA based vehicle road coordination system was optimized; In the book "Simulation Theory and Key Technologies of Intelligent Vehicle Road Collaboration System", Shang Guanwei and others systematically proposed a modeling method for vehicle road collaboration simulation system based on multi-resolution and federated structure, and provided the implementation process of the vehicle road collaboration system simulation, testing and verification platform.
Vehicle Ad hoc Network (VANET) is the main way of information distribution in vehicle road collaborative systems. There are many studies on VANET routing protocol algorithms and their simulations at home and abroad. Toutouh et al[29]By incorporating intelligent particle swarm optimization algorithm, ant colony algorithm, etc. into the Optimized Link State Routing (OLSR) routing protocol, and evaluating it through a large number of real experimental scenarios, it was found that the optimized protocol can provide better performance; Khokhar et al[30]A vehicle self-organizing network self selection clustering algorithm based on optimization design is proposed. The algorithm uses communication between on-board devices and other nodes, evaluates communication indicators using optimization algorithms, and reasonably divides vehicle nodes into different groups; Zhou Lianke and others[31]A VANET clustering broadcast protocol based on comprehensive weights is proposed, which takes into account the traffic characteristics of vehicle nodes to generate a stable clustering structure suitable for VANET traffic scenarios; Li Sihui and others[32]We have studied various vehicular self-organizing network routing protocols and designed a clustering routing protocol optimization method based on vehicle location using the minimum distance routing competition mechanism. We have also constructed an information exchange platform for the vehicle road collaboration system based on OPNET. Some scholars have also proposed VANET clustering algorithm by combining cellular automata and the level of interest of vehicles in different interactive information[33].
(3) Development period (2015 present). At this stage, artificial intelligence technology and vehicle networking technology are becoming increasingly mature, promoting the rapid development of vehicle road coordination systems. The transportation environment is in a mixed state of artificial/intelligent/connected/autonomous vehicles (CAV)[34]To cope with large-scale vehicle road group simulation testing, the United States, Japan, and Europe have successively carried out activities including Mcity[35]Projects such as the Smart Mobility Advanced Research Test Center (SMART Center), Vehicle Information and Communication System Center (VICS), and Cooperative Intelligent Transport Systems (C-ITS) were established in China in 2018, following closely behind with the establishment of the national key project "Theory and Testing Verification of Vehicle Swarm Intelligent Control in Vehicle Road Collaborative Environment". The Highway Science Research Institute of the Ministry of Transport, Chang'an University, and others have successively set up testing bases to carry out vehicle road collaborative autonomous driving testing[36].
At this stage, the vehicle road collaborative simulation testing method has gradually evolved from traditional single vehicle intelligent simulation to vehicle road group simulation, and its testing method system has also undergone technological upgrades. For the modeling methods of heterogeneous mixed traffic entities, relevant work at home and abroad mainly focuses on the macro characteristics analysis and optimization of mixed traffic flow composed of multiple intelligent level vehicles[37-38]Group control of vehicles in pure autonomous driving scenarios[39-40]And individual autonomous vehicle control in specific mixed traffic scenarios[41-42]Three directions. Ge and others[43-46]Based on heterogeneous vehicle motion simulation models such as artificial/networked/autonomous driving, preliminary research has been conducted on the optimization theory of mixed traffic flow efficiency; Gong et al[47-49]On this basis, the construction of dynamic models for mixed traffic flow, regional traffic flow control based on the time-varying characteristics of macroscopic traffic flow, and allocation of vehicle road rights resources in mixed traffic scenarios were discussed; Chai Lingguo and others[50-52]We studied the construction method of the basic environment for kinematic simulation of networked intelligent vehicles, and tested and verified the optimization control methods for signalized intersections and bidirectional two lane intelligent fleet gap optimization control in mixed traffic environments. In terms of optimizing testing mechanisms, Feng et al[53]A virtual real information interaction testing system for intelligent level testing of bicycles was constructed based on the Mcity testing field; Qiu et al[54-55]We have explored the construction of a large-scale virtual real integration testing environment, built a virtual real integration simulation analysis platform for heterogeneous vehicle collaborative behavior, and proposed a parallel layered control virtual reality simulation testing method to improve simulation testing efficiency; Li et al[56]An intelligent vehicle state parallel monitoring and control system was constructed based on parallel system theory. However, there are still bottleneck problems in related research, such as limited complexity of testing scenarios and poor stability of testing systems, and there is still a significant gap from large-scale applications.
In summary, with the rapid development of technologies such as intelligent connected vehicles and vehicle road collaboration, the urban transportation environment is undergoing tremendous changes. The group characteristics of vehicle road self-organization, networking, nonlinearity, strong coupling, pan randomness, and heterogeneous granularity are highlighted, and there is an urgent need for future oriented heterogeneous transportation entities to integrate into the mixed transportation environment, and to study the theory and methods of group intelligent collaborative behavior simulation control with vehicle road collaboration as the core.
This article focuses on the simulation testing and validation methods of vehicle road collaborative swarm intelligence in mixed traffic environments, summarizing the requirements and key technologies for vehicle road collaborative simulation testing in three stages; To break through the bottleneck of traditional vehicle road simulation, a heterogeneous traffic subject simulation model was constructed, and the operation mechanism of mixed traffic was analyzed. A mixed traffic swarm intelligence simulation method was proposed, and experimental results were given, providing guidance for the research and development of practical vehicle road collaboration systems.
During the embryonic, initial, and developmental stages of vehicle road collaborative simulation, the vehicle road collaborative system underwent a positive transformation towards heterogeneity, collectivization, and intelligence. The changing testing requirements directly propelled the development of vehicle road collaborative simulation testing technology, resulting in significant differences in simulation methods, simulation objects, and testing architectures among simulation testing platforms at different stages. The research on vehicle road collaborative simulation testing methods will shift from traditional single vehicle intelligent macro and micro simulations to hybrid traffic entity simulations; Transforming from traditional virtual simulation to virtual real interactive simulation driven by artificial intelligence technology and digital twin technology; From small-scale vehicle collaborative simulation in traditional simple scenarios to large-scale vehicle group behavior simulation supported by ubiquitous Internet. Overall,Figure 1Summarized the development path of traffic group simulation,Table 1Summarized the simulation testing requirements and feature evolution at each stage.
| 仿真阶段 | 仿真手段 | 仿真对象 | 典型测试方法/架构 | 特征 |
| 萌芽期 | 虚拟仿真 | 交通个体/ 宏观交通流 | 黑盒测试/ 元胞自动机 | 低智化个体化 |
| 起步期 | 视景一体化仿真 | 小规模群体 | 高层体系架构 | 智能化分布式 |
| 发展期 | 虚实交互 | 大规模群体 | 硬件在环/ 实车在环 | 群智化规模化 |
DownLoad:
CSV
At the beginning of the development of the concept of vehicle road collaboration, its feasibility urgently needs to be verified by scientific means. It can be said that the simulation testing system is an inevitable result driven by the demand for vehicle road collaborative development: the theoretical design framework and structure, wireless network communication protocol, and security control technology all need to be tested and verified reasonably; The inability to ensure the safety of people and vehicles during on-site testing may result in the paralysis of the urban road network; Vehicles, drivers, and road facilities themselves have high hardware costs, and actual on-site testing is very complex, requiring a large amount of funds; Simulation methods can ensure the safety of vehicle road collaborative testing through digital simulation of scenarios, while reducing the consumption of manpower, material resources, and financial resources. It can be seen that the feasibility of vehicle road collaboration needs to be verified through simulation testing methods.
In this context, simulation testing methods have emerged. At the beginning of the development of vehicle road collaborative simulation, the research on traffic simulation models was limited to the establishment of simple micro or macro models. The simulation objects were mainly individual or macro traffic flows such as vehicles and signal lights. That is, by constructing circuits and data simulations of a signal light, a vehicle, a certain functional hardware, or the overall traffic flow, the collaborative control performance in simple scenarios can be verified. The main goal of the development of vehicle road coordination system simulation in this stage is to achieve coordinated operation of vehicles under the reasonable design of traffic signals. The simulation model has certain limitations, and the individual's mobility and expressiveness are not ideal, resulting in poor authenticity.
At the same time, regarding the verification of vehicle road collaboration functions, due to incomplete sensor and hardware protocol interfaces, architectures based on black box testing can often directly verify the functions of the tested vehicles and roadside devices. The black box simulation testing architecture is as follows:Figure 2As shown. It is of great significance to verify the functionality of sensors such as cameras, stereo cameras, and radars through a black box testing architecture. Unfortunately, at this stage, the overall characteristics are characterized by individualization and low intelligence, and the technology used is relatively primitive. Its core still lies in verifying the feasibility of the vehicle road collaboration concept, which is limited to the technical feasibility verification of a single transportation entity. In addition, virtual simulation testing is a purely digital simulation testing method that heavily relies on the accuracy and correctness of the simulation model, and is also difficult to simulate the interaction characteristics between multiple traffic entities.
With the rapid development of autonomous driving technology and high-performance computing technology, vehicle road collaborative simulation testing technology has undergone a transformation. In addition to considering individual vehicles, it can also simulate the traffic environment in which vehicles are located and verify small-scale groups in typical scenarios. At this stage, a large number of emerging technologies such as vehicle behavior control, information exchange simulation, environmental visual simulation, simulation testing and evaluation have developed rapidly. Simulation testing technology based on distributed architecture has been widely applied, which utilizes the full spatiotemporal dynamic information of traffic participants to dynamically and realistically reflect various traffic phenomena such as traffic flow and accidents, reproduce the spatiotemporal changes of traffic flow, analyze the evolutionary characteristics of individual subjects supported by the vehicle road coordination system, and verify the effectiveness of vehicle road coordination in traffic efficiency and vehicle safety travel. The simulation architecture, such asFigure 3As shown.
The behavior control method of vehicles is an important component of the vehicle road coordination system. It collects external environmental information (including external vehicle interference and road information) as well as the vehicle's own position, speed, and other information, and controls the vehicle's driving in the simulation system in real time. It can not only simulate the normal operation of the vehicle, but also simulate the occurrence of accidents, save resources, and improve the realism of the simulation.
In the initial stage of the vehicle road collaboration system, the related work of behavior control simulation mainly focuses on three directions: macro characteristic analysis and optimization of traffic flow involving autonomous driving, meso level control of homogeneous autonomous driving vehicles, and individual behavior control of autonomous driving vehicles in typical traffic scenarios. The vehicle behavior control method can simulate the next running state of the vehicle by judging its acceleration, deceleration, lane changing, steering and other behavior states, loading different car following models, lane changing models and other models, providing warning information in advance when danger may occur, and sending it to the traffic simulation module and simulation management and parameter evaluation module in the form of data to avoid dangerous situations. The specific functions are as follows:Figure 4As shown.
Information exchange simulation realizes the information exchange between vehicles and roadside devices in the road network. It is responsible for receiving vehicle status, roadside status, and scene execution location information forwarded by simulation manager members; The vehicle to vehicle/vehicle to road interaction information in the vehicle to road collaboration system is integrated into the 4G network of vehicles Wifi、 Simulate communication in Dedicated Short Range Communication (DSRC) mode and provide feedback on the simulated communication network status; Send the simulation results to the traffic simulation team as the basis for deciding how to execute the scenario.
The vehicle road collaborative vehicle networking communication simulation program includes modules for data exchange with external onboard communication units (On Board Units, OBU) and roadside communication units (RSUs). The communication modules of the simulation system are simulated in hardware in the loop, and some environmental vehicle and intersection signal light data are mapped in hardware to achieve joint communication simulation of multi-agent nodes. This joint simulation scheme is as follows:Figure 5As shown.
As an important branch of the vehicle road coordination system, the visual simulation system can output real-time traffic operation status that is close to reality. However, different traffic components have different focuses and levels of detail on vehicles, roads, onboard devices, roadside devices, and the process of vehicle ground information exchange. Therefore, in the initial stage of vehicle road collaboration, distributed interactive visual simulation platforms are often used to conduct in-depth research on vehicle information exchange, vehicle road information exchange, road condition information collection, traffic flow control, and other aspects combined with visual simulation technology. HLA based Run Time Infrastructure (RTI) visual simulation technology is a typical representative of it. It defines the vehicle road collaboration system as a federated system, taking into account the principle of modularity. Each subsystem in the vehicle road collaboration system is defined as a federated member, and the system is divided into six entity simulation federations based on their functions. Each federation is responsible for completing its own simulation tasks and providing interactive information required for simulation to other federations, forming a distributed vehicle road collaboration simulation system. The specific composition is as followsFigure 6As shown.
The performance testing during the initial stage of the system mainly focuses on single vehicle control with the support of roadside equipment, and is carried out in a small-scale multi vehicle collaborative vehicle road collaborative environment to complete the simulation or real system instantiation of various elements of the vehicle road collaborative system, such as vehicles, roads, people, and information exchange networks, with specific functional characteristics of units/components. The performance testing architecture in typical vehicle road collaborative application scenarios is shown inFigure 7.
On the basis of completing the above test scenario design, for the testing of the vehicle road collaborative simulation system, it is necessary to build an excellent set of vehicle road collaborative testing cases to achieve the goal of maximizing the number of system defects discovered while minimizing testing work. It is worth mentioning that in this stage, the application scenarios of vehicle road collaboration are mainly based on the collaboration between roadside devices and single or small-scale entities. The communication methods are mainly vehicle to infrastructure (V2I) and vehicle to vehicle (V2V) communication. Some typical scenarios includeTable 2As shown.
| 控制模式 | 场景名称 | 通信 | 分类 |
| 单车控制 | 绿波车速引导 | V2I | 效率 |
| 长直路段车路通讯 | V2I | 服务 | |
| 限速预警 | V2I | 安全 | |
| 多车协同 | 前向碰撞预警 | V2V | 安全 |
| 盲区预警/变道预警 | V2V | 安全 | |
| 协作式变道 | V2V | 效率 |
DownLoad:
CSV
After generating test cases, it inevitably involves the generation and optimization of test sequences. The test sequence is formed by concatenating the test sub sequences according to the working mode conversion and operation scenario of the vehicle road collaborative simulation system. The main methods include time automata based test sequence acquisition methods, genetic optimization based test sequence acquisition methods, ant colony optimization based test sequence acquisition methods, and firefly immune algorithm based test sequence acquisition methods.
In addition, after testing the vehicle road coordination system, it is necessary to verify and evaluate it to confirm its reliability. This is often achieved by analyzing and verifying the correlation between the evaluation requirements and the characteristics of the vehicle road coordination system using methods such as Analytic Hierarchy Process, Data Envelopment Analysis, Fuzzy Comprehensive Evaluation, Neural Network, and Grey Evaluation. This establishes an evaluation index system that organizes and levels a large number of mutually coupled and constraining factors. Taking early warning assistance as an example,Figure 8The impact of different Connected Vehicle (CV) penetration rates on the success rate of early warning under small-scale transportation was demonstrated,Figure 9The changes in the minimum distance between vehicles under different distance positioning errors (measured value minus actual value) during warning braking were demonstrated, indicating that the vehicle road collaborative simulation technology has greatly improved in traffic simulation and functional verification at this stage.
From this, it can be seen that in the initial stage, the vehicle road collaborative distributed simulation testing technology has developed rapidly, with more intelligent performance, laying a solid theoretical foundation for subsequent testing technologies. However, with the explosion of autonomous driving technology, modern transportation will undergo earth shattering changes. Traditional vehicle road collaborative simulation systems are often limited to the technical feasibility verification of a single vehicle subject, lacking in-depth understanding of the behavioral characteristics of traffic groups. They have not yet considered vehicle road collaboration as a large-scale system involving the entire traffic environment, and urgently need more accurate vehicle models, more intelligent control methods, more realistic testing environments, more flexible testing methods, and richer testing scenarios.
With the rapid development of technologies such as 5G communication, artificial intelligence, and multi-sensor, autonomous driving technology is being fully applied, and vehicle driving modes are in the process of developing from manual driving, assisted driving, autonomous driving, human vehicle hybrid driving, to advanced unmanned driving. Modern transportation will see a large number of autonomous vehicles with different standards, architectures, and levels of intelligence. At this stage, the demand for vehicle road collaborative simulation testing is developing towards swarm intelligence and scale.
Therefore, technologies such as virtual real interaction and hardware in the loop based on digital twins can effectively conduct comprehensive testing and evaluation of autonomous vehicles in the simulation process of multiple physical quantities, scales, and probabilities, providing good testing and evaluation data for autonomous driving testing. Therefore, the intelligent simulation testing technology of vehicle road collaborative group has become the key to breaking through the bottleneck of existing vehicle road collaborative simulation testing technology.
The method architecture proposed in this article is as follows:Figure 10As shown in the figure, based on the underlying theory of traditional vehicle road collaborative simulation testing, heterogeneous traffic subject modeling is used as the first level, mixed traffic group intelligent behavior simulation is used as the second level, and virtual real interaction testing is used as the third level, aiming to achieve comprehensive simulation testing of vehicle road collaboration for future large-scale transportation. Specifically, this method architecture can study asynchronous data synchronization driven mechanisms for twin testing subjects, construct a compact scene space, build a scalable, highly reliable, and large-scale virtual real interactive operating environment that integrates virtual and real data, map the physical space of autonomous vehicle testing, reflect the corresponding life cycle of autonomous driving capabilities of physical vehicles, and simulate and test the group perception ability, communication ability, decision-making ability, control ability, and positioning ability of autonomous vehicles in mixed traffic environments.
For the new hybrid transportation system composed of vehicle road collaboration technology integrated into road traffic, subject heterogeneity is one of its core features, reflected in the differences in perception, interaction, decision-making, control and other characteristics of vehicles with different levels of intelligence, which leads to the heterogeneity of vehicle motion behavior and makes the operation mechanism of hybrid transportation complex and diverse, difficult to analyze. In this context, traditional single vehicle simulation models can no longer meet the behavioral simulation needs of heterogeneous traffic entities, and it is difficult to reflect the intelligent collaborative behavior characteristics of mixed traffic groups, nor can it reflect the operating mechanism of new mixed traffic. Therefore, it is necessary to deeply analyze the functional characteristics differences and motion behavior characteristics of heterogeneous traffic entities, and establish their kinematic simulation models separately.
On the basis of a profound understanding of the functional characteristics of heterogeneous traffic entities, this section constructs a simulation model from the perspective of vehicle kinematics, analyzes the operation mechanism of mixed traffic, and establishes an intelligent behavior simulation framework for heterogeneous traffic groups（Figure 11）To provide vehicle behavior model support for the intelligent simulation testing system of mixed traffic group in the vehicle road collaborative environment.
According to the different forms of perception, interaction, and decision control of vehicles in road traffic, heterogeneous traffic entities are divided into Human Driven Vehicles (HDVs), Autonomous Vehicles (AVs), CVs, and CAVsFigure 12As shown. To simulate the motion behavior of heterogeneous traffic entities, it is necessary to analyze the functional characteristics of different types of vehicles, as well as study the impact relationship between traffic entities and road environments.
Artificially driven vehicles do not have automatic control and networking functions. The perception of the environment during the entire driving process relies on human senses, and their driving safety and efficiency are closely related to the driver's driving experience; Intelligent vehicles, equipped with multiple sensors and automatic control technologies, can assist drivers to some extent in driving; Connected vehicles can achieve communication with roadside devices or other vehicles with connected functions; CAV refers to autonomous driving vehicles with networking capabilities, which can share status information between vehicles through vehicle networking technology and achieve autonomous perception, decision-making, and control.
The operation of conventional manually driven vehicles mainly relies on the driver's behavioral decisions and the vehicle's underlying kinematics. IDM is currently a relatively advanced dynamic model for describing manually driven vehicles, with meaningful physical parameters that can better reflect the car following behavior characteristics of HDVs. It has been widely applied in various research fields. The IDM following function is
|
˙vn(t)=a{1−[vn(t)v0]4−[s0+vn(t)T−vn(t)Δvn(t)(2√ab)−1hn(t)−l]2} |
(1) |
In the formula:vn(t）For vehiclesnexisttThe speed of time; Δvn(t）For vehiclesn-1 andnexisttThe speed difference of time;hn(t）For vehiclesnexisttThe distance between the front of the car at the moment;aThe maximum acceleration of the vehicle;s0The minimum braking distance for the vehicle;v0For free flow velocity;TFor safe headway;bExpected deceleration for the vehicle;lFor the length of the vehicle body.
Before the development of advanced autonomous driving technology matured, the application of intelligent driving assistance technology enabled vehicles to improve driving performance in road cruising scenarios. Although traditional cruise control can allow vehicles to drive at a fixed speed, it cannot actively avoid collision risks. Therefore, an adaptive cruise control module has been constructed to assist and automatically adjust the desired speed by autonomously sensing the position and speed information of the target vehicle. Adaptive Cruise Control (ACC) can exhibit high stability and low fluctuation vehicle operating characteristics that are not possessed by manual driving, and can reflect the differences in motion behavior between intelligent vehicles and manually driven vehicles. Therefore, an adaptive cruise control model for intelligent vehicles is proposed, and its formula is as follows
|
˙vn(t)=k1[hn(t)−l−s0−tavn(t)]+k2Δvn(t) |
(2) |
In the formula:taExpected workshop time for ACC vehicles;k1Control coefficient for distance error between vehicles;k2Control coefficient for speed difference.
Connected vehicles can operate efficiently with the expected distance and speed through information exchange between vehicles, improving road carrying capacity and driving safety, and reducing fuel consumption and environmental pollution. The operational characteristics of connected vehicles are mainly reflected in the situation of vehicle queue driving. Therefore, a time-varying communication delay connected vehicle queue model is proposed. Multiple information flow topologies are established through vehicle to vehicle communication and emerging 5G solutions. The communication topology is designed as a bidirectional leader following type, which not only enables any following vehicle to receive the position, speed, and acceleration information of the leading vehicle, but also enables information flow interaction between adjacent vehicles, making vehicle to vehicle communication interaction more frequent, communication topology more complex, and simulation model output results more realistic. The distributed control model for networked vehicle queues with time-varying delays is
|
ui=−ksN∑j=1aij{xi[t−τ(t)]−xj[t−τ(t)]−dij}−kvN∑j=1aij{vi[t−τ(t)]−vj[t−τ(t)]}−kaN∑j=1aij{ai[t−τ(t)]−aj[t−τ(t)]} |
(3) |
In the formula:ui(t）After feedback linearization, the vehicle dynamics model for connected vehiclesiexisttTime control input;xi[t－τ(t）]Thevi[t－τ(t）]Andai[t－τ(t）]Vehicles in the connected vehicle queueiexistt－τ(t）The position, velocity, and acceleration of time;τ(t)∈(0, ˆτ], time-varying delay, for anyt≥ 0 continuously differentiable,ˆτThe upper bound of the maximum delay allowed for the networked vehicle queue;ks、kvandkaControl coefficients for inter vehicle distance, velocity, and acceleration errors of the vehicle node controller;aijThe adjacency matrix elements have a value range of {0,1},aij=0 represents the vehicleiUnable to obtain vehiclejOtherwise, it indicates the vehicleiCan obtain vehiclesjThe information;dijFor two adjacent connected vehiclesiandjExpected distance between nodes;NThe total number of vehicles.
Connected autonomous vehicles combine the functional features of intelligent vehicles and connected vehicles, and can achieve efficient adaptive target detection and precise control through real-time information sharing. When driving information can be transmitted through V2X technology, connected autonomous vehicles drive in Cooperative Adaptive Cruise Control (CACC) mode, that is, CACC vehicles; Otherwise, the CAV operates in adaptive cruise control (ACC) mode and is referred to as an ACC vehicle. At each time step, ACC/ACC vehicles will determine the acceleration or velocity to be output based on their own and the target preceding vehicle's speed and position in the previous time step, as well as their expected inter vehicle time distance and cruise speed. The CACC model is
|
{vn(t+Δt)=vn(t)+ksen(t)+kd˙en(t)en(t)=hn(t)−l−s0−tcvn(t) |
(4) |
In the formula: ΔtControl step size for CACC system;en(t）For vehiclesnexisttThe error between the actual distance between vehicles at the moment and the expected distance between vehicles;tcExpected workshop time distance parameters for CACC vehicles;kdControl coefficient for the differential term of distance error between vehicles.
In summary, the characteristics of heterogeneous traffic subject models are important for, for exampleTable 3As shown. The differentiated micro motion characteristics of heterogeneous traffic entities will affect the operation mechanism of traffic flow at the macro level. Based on the heterogeneous traffic entity model, the interaction relationship between mixed traffic flow volume, density, average speed, and heterogeneous vehicle penetration rate can be analyzed, providing traffic flow mechanism model support for intelligent simulation testing of mixed traffic groups in a vehicle road collaborative environment.
| 类型 | 模型输入 | 反馈形式 | 模型输出 |
| HDV模型 | 车头间距、本车/前车速度 | 非线性 | 加速度 |
| AV模型 | 车头间距、本车/前车速度 | 线性 | 加速度 |
| CV模型 | 队列车辆位置、速度、加速度、时延 | 线性 | 加速度变化量 |
| CAV模型 | 车头间距、前车速度、本车速度/加速度 | 线性 | 加速度/速度 |
DownLoad:
CSV
In road traffic simulation, the behavior simulation of all heterogeneous traffic entities reflects the operating mechanism of mixed traffic flow. In the mixed traffic flow composed of HDV and CAV, the heterogeneity of CAV's motion behavior is reflected in the driving modes of CACC and ACC, mainly manifested in the expected inter vehicle time distance, power response type, and feedback form. settingpmThe proportion of HDV in mixed traffic flow;pcThe proportion of CACC vehicles;paThe proportion of ACC vehicles. In the equilibrium state of homogeneous traffic flow, all vehicles have the same equilibrium speed and headway; In mixed traffic flow, due to the maximum speed limit, it can be assumed that all vehicles have the same equilibrium velocity, but the equilibrium headway of the vehicles is determined by their respective kinematic models. By establishing a function relationship between the equilibrium state of traffic flow and the distance between vehicles, the basic graph function relationship between mixed traffic flow density and speed can be obtained as follows:
|
k=1pcfc(ve)+pafa(ve)+pmfm(ve) |
(5) |
In the formula:kFor mixed traffic flow density;fm(ve）Thefc(ve）Andfa(ve）The equilibrium velocity headway functions for HDV, CACC, and ACC vehicles, respectively,veFor equilibrium velocity.
Due to the lack of communication function in HDV, it is unable to exchange information with CAV. Therefore, CAV is set to only monitor the driving status of vehicles ahead. According to the dynamic characteristics of traffic flow, the relative spatial position of vehicles in mixed traffic flow is random. If CAV follows HDV closely, it will degrade to ACC vehicle. If CAV follows ACC or CACC vehicle closely, CAV will maintain CACC driving mode.
The penetration rate of CAV in mixed traffic flow is defined aspThe proportion of different vehicles appearing in the mixed traffic flow can be obtained
|
{pm=1−ppc=p2pa=p(1−p) |
(6) |
Based on the relationship between the equilibrium speed, density, and headway functions of traffic flow, the basic flow density diagram model of mixed traffic flow can be obtained as follows:
|
{k=1(1−p)[(s0+Tve)/√1−(ve/v0)4+1]+p(1−p)(tave+s0+l)+p2(tcve+s0+l)q=kve |
(7) |
In the formula:qFor traffic flow.
From this, mixed traffic flow density curves can be obtained for different CAV penetration rates, such asFigure 13As shown, it can be seen that with the increase of CAV penetration rate, the maximum capacity of mixed traffic flow gradually increases, and traffic efficiency gradually improves. Therefore, for mixed traffic environments composed of heterogeneous traffic entities, road traffic efficiency can be improved by increasing the proportion of connected autonomous vehicles in the mixed traffic flow.
The traditional transportation environment has formed a new type of mixed traffic environment after the integration of heterogeneous vehicles with different levels of intelligence. Therefore, traditional vehicle road collaborative simulation and testing technology can no longer solve the new traffic problems caused by mixed traffic flow phenomena, and typical application scenarios of vehicle road collaboration cannot fully cover the functional characteristics of mixed traffic. Therefore, this section proposes to use virtual real interaction simulation methods and swarm intelligence algorithms to test and analyze vehicle behavior in different mixed traffic scenarios under vehicle road collaborative environments.
Virtual reality interaction test technology integrates some/all actual test objects in the auto drive system into the virtual test environment to achieve debugging, low-cost, multi scenario comprehensive test. Virtual real interaction testing technology can be mainly divided into two types: hardware in the loop testing and real vehicle in the loop testing. The overall system or partial components of hardware in the loop testing are real, such as module testing for environmental perception, decision planning, and control execution. In this mode, most of the functions of the testing subject are reflected by software, and the application scope is still relatively limited. Real vehicle in the loop testing integrates all components and systems onto the vehicle, and connects the entire vehicle as physical hardware to a virtual testing environment for testing. This article designs a traffic simulation platform framework suitable for real vehicle in the loop virtual real interaction testing, mainly composed of virtual space, data space, and physical space. The overall design framework is as follows:Figure 14As shown, where:vi′′andpi′′respectivelyiInformation in both virtual and physical spaces at all times.
Starting from scene categories, traffic subjects, road types, natural environments, and behavior control, virtual space constructs a standardized testing scene library and a scalable and granularity adjustable simulation road network based on safety and efficiency guidance. Data space fusion of multi-source sensor data to extract traffic feature data, and bidirectional synchronous transmission of state data between virtual and real spaces. The physical space utilizes real-time state feature data transmitted from the physical space to dynamically map and reconstruct the measured object in the virtual space, and uses multi-source sensing technology to collect the full cycle operation data of the traffic subject, ensuring the integrity of the data source. By integrating virtual space, data space, and physical space through wireless communication, group collaborative behavior simulation, testing, and verification of virtual and real traffic states can be achieved.
In order to improve the synchronization performance of the vehicle road collaborative virtual real combination simulation testing system, based on the virtual real combination simulation testing process, the operating mechanism of the testing subject is clarified. In response to the time offset problem of the testing subject in data transmission, a clock error estimation strategy is designed to correct the measurement time deviation of the virtual real state transmission. On this basis, a multi-scale filtering synchronization method for the virtual real testing system state is proposed by combining Kalman filtering and constant rotation rate and constant acceleration models, dynamically compensating for the real-time operating state of the virtual real subject and maintaining good synchronization stabilityFigure 15As shown.
Due to the wireless communication network connection between physical space and virtual space, frequent data exchange will greatly increase the communication burden of the testing system; The simulation event calculation scale in virtual space needs to maintain a high frequency in order to more closely approximate the real traffic scene in physical space. known numberfvFor the sampling frequency of virtual space state transition,frIf the sampling frequency observed for the physical space system satisfiesfv>frTwin testing subject in virtual spacemThe operational status of timeXmCan be described as
|
Xm∈{ˆXm,v,ˆXm,r} |
(8) |
In the formula:ˆXm,vTest subjects for twins in virtual spacemPredicted value of time state;ˆXm,rTest subjects for twins in virtual spacemThe predicted value of the state is adjusted based on real spatial information at all times.
The synchronization method of virtual and real system states is: when the virtual space ismWhen no updated observations from real space are received, the system calculates the next state at a high-resolution time scale, i.e., selectsfvVirtual state for frequency propulsion system; When the virtual space ismWhen receiving updated observations from real space at all times, select the sampling frequency for observing the physical space systemfrPromote the virtual state of the system, i.e
|
Xm←{ˆXm,v=g(Xm−1,fv)ˆXm,r=g(Xm−1,fr) |
(9) |
In the formula:g(·) is the multi-scale filtering synchronization function for the state of the virtual real testing system.
Selecting horizontal and vertical test trajectories, model-based trajectory estimation method, discrete Kalman filtering method, and multi-scale filtering method were used to optimize the synchronization performance of the twin testing system. The specific trajectory synchronization error results are as followsFigure 16As shown.
Compared with model-based trajectory estimation methods, the multi-scale filtering synchronization method has significant advantages in lateral trajectory synchronization; The discrete Kalman filter method is limited by the communication interval between virtual and real information, and cannot provide continuous and effective synchronization information. The continuity and smoothness of the twin subject's trajectory are poor, therefore, the synchronization effect of this method is relatively poor.
Based on the multi-scale filtering synchronization method, the simulation testing system is organically integrated with the real traffic environment, ultimately forming a virtual real combination testing and verification environment that includes real traffic operation, twin state synchronization, and virtual traffic operation. This supports the collaborative effectiveness testing and verification of the vehicle road traffic test equipment, core component functional modules, and virtual simulation subjects in multiple environments.
The testing of vehicle road collaboration system based on simulation scenarios can flexibly deconstruct and reconstruct scenarios, achieve targeted testing of dangerous scenarios, thereby reducing on-site testing costs and ensuring the efficiency and completeness of vehicle road collaboration system simulation. The core of testing is the construction of testing scenarios, and the analysis of scenario elements is a quantitative assessment of the coverage ability of scenario testing. The scenarios used for virtual testing should meet the requirements that the characteristics of each element of the scenario can be quantified, the scenario can be reproduced on the testing software, and can reflect the real traffic conditions to a certain extent. In short, the scene is composed of each frame of scene combination, describing a series of actions and events that lead to the result, and the scene includes static elements, dynamic elements, and autonomous driving vehicles.
With the improvement of autonomous driving levels, traditional simulation testing environments and methods can no longer meet the needs of mixed traffic simulation testing in vehicle road collaborative environments. There are no clear regulations at home and abroad for the rapid construction and simulation testing of vehicle road collaborative mixed traffic scenarios. Therefore, based on the simulation requirements of intelligent behavior of vehicle road collaborative mixed traffic groups composed of heterogeneous traffic entities, the functional characteristics of different traffic scenarios are extracted layer by layer and scene elements are decomposed, such asFigure 17As shown, among themFI, j′doILayer by LayerjOne element.
Based on the importance of scene elements, reconstruct the vehicle road collaborative mixed traffic scene test case, and finally execute the test case in the simulation environment to support the scene functional features required for group behavior simulation. The relationship between the demand for intelligent behavior simulation of vehicle road collaborative mixed traffic groups and scenario cases is as follows:Figure 18As shown.
The commonly used research methods for scenario testing case generation include testing matrix method, Monte Carlo method, combination testing method, worst-case scenario evaluation method, and game theory method. The combination testing method can achieve scene coverage through the interaction between scene elements, generating as few scene test cases as possible to cover more dimensional scene elements. It can not only meet the needs of testing and verifying the functional characteristics of vehicle road collaborative mixed traffic scenes, but also improve the coverage of test cases based on the combination of all scene elements; The Monte Carlo method can extract key scene elements based on real natural driving data, simulate the probability distribution of scene element parameters, restore real scenes, and randomly generate elementary scene test cases.
With the support of the above methods, in order to ensure the comprehensiveness of the testing process in the digital twin operating environment, a virtual real interaction testing mechanism for typical application scenarios of vehicle road collaborative group control is constructed. Based on factors such as test objects, road types, natural environments, and behavioral control objectives, more than 90 traffic scenarios are constructed, greatly improving the coverage of traditional simulation testing scenarios and basically meeting the testing coverage requirements of intelligent vehicle road collaborative systems. Some scenarios includeTable 4As shown.
| 场景功能分类 | 场景功能特征 | 仿真构建需求 |
| 协作式交叉口通行 | 信号控制交叉口车辆常规通行 | 安全/效率 |
| 信号控制交叉口处理异常停车 | 安全/效率 | |
| 交叉口信号自适应控制 | 效率 | |
| 交叉口动态车道管理 | 效率/管理 | |
| 无信号控制交叉口组织通信异构车辆常规通行 | 安全/效率 | |
| 协作式城市快速路匝道控制 | 城市快速路单匝道控制 | 安全/效率 |
| 城市快速路主线多匝道协同控制 | 安全/效率 | |
| 城市快速路与高速公路结合部控制 | 安全/效率 | |
| 协作式城市路段控制 | 城市常规路段组织车辆编队 | 安全/效率 |
| 常规路段组织紧急车辆优先通行 | 安全/效率 |
DownLoad:
CSV
By constructing mixed traffic operation scenarios, a testing foundation and environment have been provided for swarm intelligence simulation. In the process of scenario testing, the interaction between vehicles and the environment forms a closed loop with significant mutual influence. The testing scenario will have an impact on the vehicle state. By calling various mixed traffic scenarios in the virtual real interaction scenario library of different scenarios to simulate the vehicle group, combined with the simulation method of virtual real interaction, the goal of vehicle group decision-making and control can be achieved, promoting the development and improvement of intelligent transportation systems.
Based on the modeling of single vehicle dynamics/kinematics and multi vehicle collaborative behavior, this section analyzes and models the optimization of operation control for heterogeneous traffic groups. Due to the close relationship between the intelligent control methods of heterogeneous traffic groups and the involved traffic scenarios, this section uses the concept of scene slicing to decompose the operation scenarios of large-scale mixed traffic groups into traffic operation scenario elements with different levels of complexity, such as road sections and intersections, to finely reflect the regional group simulation control mechanism. In addition, due to the composability between scenes, basic scene elements can be combined and reconstructed on the basis of scene slicing, increasing the types of regional scenes, expanding the testing scope, and forming various regional testing scenes with different structures and comprehensive functions. So this article reviews group simulation methods from typical basic scenarios.
In the current level intersection control mode, the standard 8-lane 4-direction intersection has a traffic capacity of approximately 3600 veh · h-1In the urban transportation system, except for the maximum traffic capacity that may reach the intersection during rush hour, the traffic flow may not reach such a high level during other periods. Therefore, signal controlled intersections can basically divert the current traffic flow in the city.
Considering the changes in mixed traffic environments, a Coupled Vehicle Signal Control (CVSC) method is proposed to optimize the traffic signal timing and driving trajectory of CAVs with the goal of improving traffic efficiency and energy conservation. Meanwhile, continuously optimize signal timing to minimize the total delay at intersections. The CVSC method uses a mixed traffic flow basic diagram model as a connection, with efficiency and energy conservation as the goals, to achieve coordinated control between signalized intersections and CAVs at both ends.
Combining the above signal control methods, based on the static characteristics of level intersections, the signal control characteristics of intersections, the kinematic models of different types of intelligent vehicles, and the operation mode of intelligent vehicle road coordination systems, a CAV/HDV signal coordination control method for hybrid gap coupling intersections is proposed. Through the coupling of signal/non signal and different types of vehicle motion processes, combined with the coupling process of lane level signal control and parallel gap control at intersections, the gap passage mechanism of CAV/HDV at intersections is established. Based on the vehicle start stop delay model and driver reaction time, the HDV gap selection mechanism using vehicle behavior prediction and gap delay is used to determine the lane level signal control strategy of HDV and achieve dynamic coupling of CAV/HDV in the passage gap at intersectionsFigure 19As shown, where:TFCThe minimum time for the first conflict between vehicles;SLTo determine the length of the lane change decision area;SSAdjust the length of the speed adjustment zone. Based on the control method of a single intersection, continuous intersection signal collaborative optimization is carried out "from point to surface", radiating to regional signal control, achieving intelligent simulation and control optimization of mixed traffic groups, and improving the traffic capacity of the regionFigure 20As shown.
In the operating environment of urban road sections, the operation control of heterogeneous traffic entities is an important method to improve traffic efficiency and safety. Simulating the optimization method of intelligent fleet collaborative control can effectively alleviate urban traffic pressure, reduce congestion and accidents. The operation mode of intelligent fleet in road segments has multiple characteristics. This article focuses on the mixed traffic flow environment where manual driving and autonomous driving are parallel, and optimizes the fleet formation for different driving behaviors of vehicles during road segment driving.
In terms of optimizing the ranking of vehicles in the fleet, based on the continuous following behavior characteristics of drivers, taking into account the differences in driving behavior characteristics of the drivers in front of the following vehicles, fully understanding the behavioral characteristics of the following vehicle drivers in the process of following different types of drivers, a fleet vehicle ranking optimization method based on deep cognition of following behavior characteristics is proposed. In terms of vehicle entry control strategy, based on the characteristics of occasional lane changing behavior of drivers, and driven by the gap control theory, a customized vehicle entry control strategy optimization method based on lane changing behavior cognition is proposed. In terms of fleet vehicle separation control, a fleet vehicle separation control method based on stress driving behavior cognition is proposed by studying the influence of the behavior characteristics of single bicycle drivers around the fleet on reducing the working pressure of roadside equipment at intersections, while fully considering the actual driving characteristics of the fleet in the front section of the intersection and the objective problems that exist.
Based on the three different driving behavior operation control methods of intelligent fleet mentioned above, combined with multi fleet driving scenarios, the research on multi fleet collaborative operation control methods is carried out, thus forming a mechanism for collaborative operation control of road sections in mixed traffic environments, such asFigure 21As shown.
To meet the demand for rapid testing of the effectiveness of vehicle road collaborative group control, a traffic simulation platform combining virtual and real is designed, and simulation testing and evaluation of group intelligent behavior in traffic operation scenarios such as road sections and intersections are carried out around this platform. The virtual real simulation platform mainly consists of a data generation system, an algorithm testing system, and a scene verification system. The data generation system is integrated with the real space, and the scene verification system is integrated with the virtual space. The overall system architecture is as followsFigure 22As shown.
Based on the above simulation platform, the scale of vehicle road collaborative simulation testing in mixed traffic scenarios has been significantly improved. From the initial mixed operation scenario simulation that could only meet the needs of 9 nodes and 500 traffic subjects, it has now been expanded to large-scale group collaborative behavior simulation with 150 nodes and 2000 traffic subjectsFigure 23As shown.
Based on a virtual real combination simulation platform, the proposed hybrid gap coupled intersection signal cooperative control method was validated and evaluated for a typical 8-lane intersection, and compared with the classical CACC model. The experimental results are as follows:Figure 24 (a)As shown, it has been verified that this method can improve the overall performance of traffic flow at signalized intersections, perform well under moderate traffic flow conditions, and can adapt to unbalanced traffic demands.
Comparative experiments were conducted on intersections with parallel gap control, timed control, and stop sign control under the same road channelization conditions. For an 8-lane intersection, the input was 350 veh · h for each type of turning vehicle-1FromFigure 24 (b)It can be seen that the delay of vehicles at intersections controlled by parking signs continues to increase, which reflects the non periodicity of the parking sign control method and also indicates its limitations in controlling traffic flow; The delay of vehicles at signal controlled intersections has a certain periodicity, and this periodicity depends on the signal control period. When vehicles pass through the intersection, the delay of vehicles with green lights decreases, and when vehicles pass through the intersection, the delay of vehicles with red lights increases.
Combining the designed CAV/HDV signal collaborative control method for hybrid gap coupled intersections, the road segment subnet is divided and implemented to continuously induce the maximum number of vehicles passing through the green wave zone, thereby controlling and optimizing the green wave zone signal of continuous intersections, achieving traffic group control radiating from the point to the area.
Based on the combination of virtual and real simulation platforms and the acquisition of the characteristics of occasional lane changing behavior of drivers, simulation tests were conducted in a mixed traffic environment using gap control theory for single vehicle entry, fleet cruising, and fleet separation control in a two-way two lane environment.
In order to select the same vehicle entry speed, the entry speeds of the vehicles are set to 10, 20, and 25 m · s, respectively-1Analyze the relationship between different speed combinations and safety clearances of overtaking vehicles, vehicles in front of safety clearances, and vehicles behind safety clearances in four different situationsFigure 25As shown, it can be seen that the average speed of overtaking vehicles during the step-by-step overtaking process is 14.02 m · s-1Compared to the initial following state, the vehicle speed has increased by 16.8%, and the maximum average speed of overtaking vehicles is 14.94 m · s-1.
When overtaking vehicles are following a convoy, there is an overall acceleration phase, where the overall acceleration speed of the convoy is less than the maximum average speed. That is, the following speed of the vehicle is less than the speed when overtaking in the opposite lane. Therefore, the acceleration speed of the convoy is set to 14.92 m · s-1.
Calculate the average speed of vehicles in the convoy during the process of a step-by-step bicycle overtaking the convoy, as shown inFigure 26It can be seen that the average speed of the fleet vehicles is 16.27 m · s-1Compared to the free acceleration driving state, it has increased by 9.04%.
In the fleet cruising simulation test, CACC mode and fleet operation optimization mode based on the depth of following behavior characteristics were selected as the following control modes for the fleet in mixed traffic scenarios. Data within 120 meters of the highway simulation scenario was selected to obtain the multi vehicle motion mode. The experimental results are as follows:Figure 27As shown, it can be seen that the optimized cruising speed and surrounding multi vehicle operating speed of the fleet operation mode are significantly higher than before optimization, indicating that this method is beneficial for improving the operational efficiency of the fleet.
The scenario for fleet separation testing is set as follows: a fleet consisting of 7 connected autonomous vehicles runs on the road section in front of the intersection, with each vehicle set as a separated vehicle, each with a length of 5 meters, a headway of 1 second between vehicles, and a fleet speed of 20 m · s-1The remaining green light duration is 5 seconds. The position of the manually driven vehicle is parallel to the fourth vehicle in the convoy at this moment, and the acceleration of the separated vehicles in the convoy is obtained when the convoy approaches the intersection. The experimental results indicate that compared to the vehicle separation control method with unknown driving behavior（Figure 28 (a)）Application of a fleet vehicle separation control method based on stress driving behavior cognition（Figure 28 (b)）The range of acceleration variation of the separated vehicles has been reduced by 37% to 72%.
Select vehicle groups with different average operating speeds to conduct simulation testing and analysis of their operating trajectories. Select 160 sets of trajectories that meet the speed conditions within the same operating range. The movement behavior of the traffic group is as follows:Figure 29As shown. Analyze group behavior by statistically analyzing the distribution of vehicle trajectories at different average speeds on two scales, horizontal and vertical, and conduct simulation tests on important indicators such as communication delay and traffic operation efficiency in the current mixed traffic scenario using vehicle trajectory data combined with a virtual real simulation platform.
Communication delay is one of the key indicators in vehicle road cooperative group control. In order to ensure the normal operation of intelligent vehicles and connected vehicles, it is necessary to simulate and test communication delay. The transmission delay results under different traffic simulation scenarios are as follows:Figure 30As shown in the figure, it can be seen that when there are 100 connected vehicle nodes in the scene, the transmission delay is relatively large. This is because when there are fewer vehicles, the nodes are scattered and sparse, and the reachable routes from the source node to the destination node are relatively few, which may result in a larger transmission delay; With the increase of vehicles, the delay effect has relatively improved; When the number of vehicles continues to increase, it will cause channel congestion, increase the probability of transmission failure, and increase the possibility of an increase in the backoff counter of the Data Flow Control (DFC) channel access protocol in the Media Access Control (MAC) layer, resulting in an increase in latency; The overall performance of the simulation is good, with an average delay of around 0.059 seconds for all traffic scenarios.
The network throughput mainly depends on the number of connected vehicles. When there are many vehicle nodes, there are relatively more vehicles engaged in information exchange. Although communication media may be relatively congested, network throughput will still increase. As the number of vehicles increases, the throughput increases significantly, and the impact of speed on throughput is relatively small. The main reason is that the multi communication mode competition mechanism based on neural networks will switch to suitable communication modes such as DSRC in real time. The average throughput of all traffic simulation scenarios is approximately 104862 bit · s-1.
When conducting traffic operation efficiency index testing, changing the CAV penetration rate while keeping other scene element parameters unchanged, analyzing the trend of efficiency changes in mixed traffic scenarios, and verifying the relationship between efficiency and heterogeneous traffic subject penetration rate in mixed traffic scenarios. Efficiency is characterized by traffic flow, with higher traffic flow indicating higher traffic efficiency. The test scenario is when a vehicle in the road network experiences an emergency (such as a vehicle malfunctioning and stopping on the road or a traffic accident). Vehicles with vehicle road coordination receive a notification of the sudden accident on the road ahead through vehicle to vehicle or vehicle to road communication in advance to change lanes or detour. Vehicles without vehicle road coordination only begin to slow down and change lanes when they approach the incident site. The test results were selected for CAV permeability of 0, 20%, 40%, 60%, 80%, and 100%, respectivelyFigure 31As shown, the experimental results indicate that the lower the penetration rate of CAV, the higher the vehicle delay and parking frequency, and therefore, the lower the traffic efficiency; When the penetration rate of CAV is 0, it represents that all vehicles in the simulated road network are manually driven, and their driving behavior is completely uncontrollable, which will inevitably lead to a decrease in traffic efficiency; When the CAV penetration rate is 1, it represents that all vehicles in the simulated road network are CAVs. At this time, these vehicles are all in CACC driving mode, maintaining real-time communication with roadside equipment. The roadside equipment guides the vehicles to avoid congestion and accident sections in advance, efficiently driving in formation and passing through intersections, and achieving the highest efficiency; The vehicle road collaborative mixed traffic group simulation method has good ability to predict short-term traffic flow and improve prediction accuracy, providing scientific and effective data support for short-term traffic flow analysis and prediction.
The mixed traffic group virtual real simulation can provide quantitative basis for road traffic management and theoretical guidance for urban construction. In the virtual real combined traffic group simulation system, traditional traffic simulation indicators have been significantly improved, and the simulation time granularity of the intelligent vehicle road collaborative system has been shortened from less than 0.5 m to less than 0.1 m for fast single solving time of group collaborative control in different scenarios; The simulation scale has also been significantly improved, from simulating a vehicle road coordination system with 9 nodes and 500 traffic entities to currently supporting more than 90 simulation scenarios related to traffic safety and efficiency, dynamically adjustable heterogeneous traffic entity penetration rate of 0-100%, and simulating group collaborative behavior of 150 node road networks and 2000 traffic entities.
With the development and application of autonomous driving technology, vehicles with different levels of intelligence are integrated into road traffic to form a new hybrid traffic environment. In recent years, simulation testing technologies such as information physics systems, digital twins, metaverse, and virtual real interaction have emerged as important development trends in the field of intelligent transportation simulation testing. Therefore, this article deeply analyzes the research status and development trends in the field of vehicle road collaborative simulation testing, and draws the following conclusions.
(1) The vehicle road coordination system is a large-scale real-time distributed system involving multiple factors and complex multi-level relationships, which must be based on practical and feasible applications. Field testing has key issues such as high cost, poor safety, limited testing scenarios, and single testing methods. Therefore, in-depth research on simulation testing technology for vehicle road coordination systems will effectively accelerate the application and implementation of vehicle road coordination systems.
(2) The development of simulation testing for vehicle road coordination systems can be summarized into three stages: the embryonic stage, the initial stage, and the development stage. The core of the budding stage is to verify the feasibility of the vehicle road collaboration system, and the overall performance of the testing technology is low intelligence and individualization; The key technologies in the initial stage are more efficient and intelligent, and distributed simulation of high-level system architecture is one of the representative technologies in this stage, which promotes the development of visual integration simulation of vehicle road collaboration systems; With the explosive development of autonomous driving technology, the demand for vehicle road collaborative simulation testing is rapidly evolving towards collectivization, intelligence, and scale. Virtual real interaction and hardware in the loop technology are becoming powerful simulation testing methods for the future development of vehicle road collaborative technology.
(3) The current transportation environment has entered the stage of mixed traffic. In order to cope with the challenges of large-scale vehicle road group simulation testing, a mixed traffic group intelligent simulation testing method based on virtual real interaction is proposed. A heterogeneous traffic subject behavior model is established, and the operation mechanism of mixed traffic is analyzed. Combined with group intelligence methods, the simulation and control of mixed traffic group intelligent collaborative behavior are realized; Relying on the advantages of virtual real interaction technology, we have overcome the challenges of generating and simulating intelligent scenarios for mixed traffic groups, and broken through the synchronization mechanism of simulation testing; Through test results verification, this method can effectively improve the efficiency, scale, and coverage of vehicle road collaborative mixed traffic simulation testing.
To cope with the new challenges brought by the simulation testing of future intelligent vehicle road collaboration systems, and to meet the requirements of large-scale open/semi open traffic environments, typical traffic scenarios such as highways/multi intersections, and cross scenario testing of group behavior, it is necessary to combine the virtual real interaction of vehicle road entities and the simulation technology of operating environment data to form a simulation testing verification logic that adapts to the testing needs of scenarios, accelerate the construction of large-scale group intelligent simulation systems and key technology breakthroughs in mixed traffic environments. The future research trends of vehicle road collaborative simulation testing are as follows.
(1) Accelerate the construction of new digital infrastructure, build a ubiquitous networked perception system with full business, full time and space, and full coverage, form fully covered digital scene information, reconstruct real scenes in virtual space, and verify testing algorithms.
(2) Developing high-level fully autonomous vehicles using brain like intelligence technology, deploying the swarm intelligence decision control method integrated with brain like intelligent vehicles on the vehicle road collaborative virtual real simulation platform based on the collection of vehicle road spatiotemporal state information, and conducting testing and verification work.
(3) Build a multi-level real-time simulation platform that supports bidirectional state synchronization execution between virtual and real spaces, accelerate the "near zero on-site" mapping and reconstruction of large-scale mixed traffic group collaborative behavior, and promote the parallel development of vehicle road collaborative testing/operation.
(4) Promote the research and breakthrough implementation of decision-making control methods and related technologies for vehicle road groups, establish a comprehensive, mature, deeply applied, and integrated virtual and real data demonstration area for vehicle road collaborative applications, form a scientific and complete theoretical system and promotion plan for intelligent virtual and real simulation testing of vehicle road collaborative groups, provide simulation testing theoretical guidance and platform environment support for China to independently master the key core technologies of vehicle road collaborative systems, and help achieve the goal of building a strong transportation country.
Plagued by frequent accidents, serious congestion, and environmental pollution, China's urban road traffic system finds it difficult to meet people's increasing demand for travel. It is urgent that measures should be taken to make the traffic system more connected, intelligent, and cooperative. In this context, vehicle–infrastructure cooperation (VIC) technology becomes an inevitable trend, which is an effective way to guarantee security, improve efficiency, optimize energy consumption, and lower emissions.
The vehicle–infrastructure cooperative system (VICS) applies communication technology, artificial intelligence (AI), big data, and cloud computing to the traffic system. Supported by wireless communication and sensing detection, it temporally and spatially acquires and integrates traffic information. It realizes vehicles' active control of safety and coordinated management of roads by the vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) information interaction and sharing [1]. A VICS is a large-scale real-time distributed system involving multiple factors and layers, and only on a feasible basis can it effectively exhibit its performance and value. Its testing and verification are one of the essential parts of research and development [2]. The technology for the simulation and test of VIC is the very means to expedite the application and implementation of VICSs. The evolution of the simulation and test of VICSs can be divided into three stages, namely, the rudiment, infancy, and developing stages.
(1) The rudiment stage (1986–2002). During this stage, the actual road traffic environment features human-driven vehicles (HDVs), whose operation safety relies on drivers' driving experience. To improve the safety and convenience of road operation, the University of California, Berkeley, launched the Partners for Advanced Transit and Highways (PATH) plan in 1986, which was the first plan putting forward the concept of VICS and put the focus on theoretical research on the car-following theory [3]. In 1991, Japan initiated the Advanced Safety Vehicle (ASV) plan [4]and Smartway plan [5], which focused on safe cooperative vehicle–infrastructure driving and smart traffic technologies. In 1999, China founded the National Center of ITS Engineering & Technology (ITSC) to make overall arrangements for VICSs. In 2000, Europe started the CarTALK 2000 program [6], which mainly aimed at studying drivers' assistance systems.
Traffic simulation dates back to the 1960s. The traffic signal optimization program (TSOP) model and the traffic network study tool (TNST) model were the typical macro models for signal control optimization [7]then. Given the VIC concept and taking into account the road environment back then, experts mainly studied the simulation of macro-traffic and micro-traffic flows. They mainly employed such simulation methods as time scan and event scan algorithms [8]. Regarding simulation models, British experts Wada et al. [9]proposed a fluid model of traffic flow, from which macrotraffic flow models such as the Payne's model [10], Papageorgious's model [11], Ross's model [12], and Wuzheng's model [13]were derived. Pipes [14]proposed the car-following theory, and on this basis, developed micro-traffic flow models including the cellular automation model [15], Krauss's model [16], and the intelligent driver model (IDM) [17]. At the rudiment stage, virtual VIC simulation was in the dominant position. Typical simulation software included TRANSYT [18]and VISSIM [19], and simulation and tests mainly relied on the white box test or black box test. There was a lack of research on the analysis of simulation effectiveness and results.
(2) The infancy stage (2003–2014). During this stage, the advance in computer technology and wireless communication technology gave rise to VICSs based on V2V and V2I information interaction. The United States, Japan, and Europe launched research programs on communication-based VIC technology in succession, such as IntelliDrive [20], the automated highway system (AHS) [21], and SafeSpot [22]. China emphasized science and technology frontier research, especially technological innovation. In 2011, China's Ministry of Science and Technology set the first research program for the key technologies of intelligent VIC under its National High-tech R & D Program (863 Program). A digital VIC platform with V2V/V2I information interaction at its core was modeled to test the actual performance and efficiency of VICSs and create a VIC environment underpinned by communication. It propelled the research and development of key technologies for VIC[23]. Supported by the 863 Program, a tenmember science research team led by Tsinghua University became the first team to define the intelligent vehicle– infrastructure cooperative system (i-VICS), and the team established a VICS test platform on the basis of high level architecture (HLA) and multi-resolution modeling (MRM) [24]. Relevant technical achievements were recorded in a book entitled Architecture for Intelligent Transportation Systems Based on Intelligent Vehicle–Infrastructure Cooperative Systems, which narrated the development of China's intelligent transportation system in this stage.
During this stage, vehicles could use communication to gain access to more information about the traffic state and achieve optimal behavior control between vehicles through V2V/V2I interaction. Methods for vehicle following simulation, overtaking simulation, and lane changing simulation became increasingly sophisticated. Assuming that there were certain limitations to each vehicle subject's expected braking rate and acceleration rate, Gipps [25]constructed a new car-following model and applied it to the response of following vehicles, thus reproducing the genuine features of traffic flow. Petrov et al. [26]presented a mathematical model and adaptive controller for an autonomous vehicle overtaking maneuver to generate polynomial virtual trajectories for every phase in real time so that the overtaking can follow the desired trajectory with unknown speed. Butakov et al. [27]developed a method that learns the characteristics of vehicle response before and during lane changes under different driving environments. They developed a two-layer model to describe the dependence of the response characteristics on surrounding vehicle configuration to achieve communication-based lane change simulation. Cai et al. [28]proposed a multiresolution information interaction method. They built a high-resolution information model of vehicle running states, a middle-resolution information model of platoon status, and a low-resolution information model of traffic flow, thus optimizing the simulation process of the HLA-based VICS. In their book Simulation Theory and Key Technology for Intelligent Vehicle–Infrastructure Cooperative Systems, Shangguan et al. systematically proposed modeling methods for VICS simulation based on multiple resolutions and federated architecture and described the flow for VICS simulation, test, and verification.
The vehicular ad-hoc network (VANET) is a major way for information distribution of VICSs. There have been plenty of studies on the VANET routing protocol algorithm and its simulation. Toutouh et al. [29]added the intelligent particle swarm algorithm and ant colony algorithm to the optimized link state routing (OLSR) routing protocol. It was found that the optimized protocol, given the evaluation results of a number of real test scenarios, could provide better performance. Khokhar et al. [30]proposed a self-election clustering algorithm in VANET on the basis of optimal design, which uses vehicle devices to communicate with other nodes and the optimal algorithm to assess communication indices so as to divide vehicle nodes into different groups. Zhou et al. [31]proposed a clustering broadcast protocol for VANETs based on overall weights, which considers node-traffic characteristics for the generation of stable clustering structures applicable to VANET traffic scenarios. Li et al. [32]studied different types of VANET routing protocols. They applied the minimum distance routing competition mechanism to design the optimization methods for clustering routing protocols based on vehicle position and constructed a VICS information exchange platform based on OPNET. Some scholars proposed the VANET grouping algorithm by integrating cellular automata clustering with the interests of vehicles in different interaction information [33].
(3) The developing stage (from 2015 to now). In this stage, the advance of AI technology and Internet of Vehicles (IoV) technology has spurred the rapid development of VICSs, and the traffic environment is in a mixed state of HDVs/autonomous vehicles (AV)/connected vehicles (CVs)/connected and autonomous vehicles (CAVs) [34]. To cope with large-scale vehicle–infrastructure swarm simulation and tests, the United States, Japan, and European countries have launched programs such as Mcity [35], Smart Mobility Advanced Research Test Center (SMART Center), Vehicle Information and Communication System Center, and Cooperative Intelligent Transport Systems (C-ITS) in succession. China, following their steps immediately, established the National Key Project of Intelligent Control Theory and Test Verification of Vehicle Swarms in VICS in 2018. The Research Institute of Highway of the Ministry of Transport and Chang'an University successively set up test bases to carry out autonomous driving tests of VIC[36].
During this stage, the test methods of VIC simulation gradually evolve from traditional single-vehicle intelligent simulation to vehicle–infrastructure swarm simulation, and the test method system has also undergone a technological upgrade. For the modeling method of heterogeneous mixed traffic subjects, the related work focuses on analyzing and optimizing the macro-characteristics of mixed traffic flow composed of vehicles with different levels of intelligence [37–38], the vehicle swarm control in sheer autonomous driving scenarios [39–40], and the control of individual AVs in specific mixed traffic scenarios [41–42]. On the basis of the motion simulation models of heterogeneous vehicles such as HDVs/CVs/AVs, Ge et al. [43–46]carried out preliminary research on the efficiency optimization theory of mixed traffic flow. On this basis, Gong et al. [47–49]discussed the kinetics modeling of mixed traffic flow, regional traffic flow control based on the macroscopic time-varying characteristics of traffic flow, and right-of-way resource allocation of vehicles in mixed traffic scenarios. Chai et al. [50–52]studied the construction method of kinematic simulation for CAVs and tested and verified the optimal control method of signalized intersections and gap control optimization of the two-way dual-lane intelligent platoon in a mixed traffic environment. In terms of test mechanism optimization, Feng et al. [53]built a virtual–real information interactive test system for singlevehicle intelligence level tests with Mcity test scenarios. Qiu et al. [54–55]carried out technical exploration for the construction of a large-scale virtual–real test environment, built a virtual– real simulation analysis platform for heterogeneous vehicle cooperation behavior, and proposed a method for parallel hierarchical control-based efficiency enhancement for large-scale virtual reality traffic simulation and tests. By the parallel system theory, Li et al. [56]built an intelligent parallel monitoring and control system for vehicle states. However, relevant studies are still subjected to limited test scenario complexity and poor testing system stability. There is still a long way to go before large-scale applications can be realized.
To sum up, in the wake of CAVs and VICSs, great changes have taken place in the urban traffic environment. Vehicle swarms are characterized by vehicle–infrastructure ad hoc, networking, non-linearity, strong coupling, generalized stochastic nature, and heterogranity. It is urgent to integrate future-oriented heterogeneous traffic subjects into the mixed traffic environment to study the theory and the methodology for the simulation control of intelligent swarm cooperation centered on VIC.
This paper focuses on the simulation, test, and verification methods of swarm intelligence of VIC in the mixed traffic environment and summarized the requirement and key technologies of vehicle–infrastructure cooperative simulation and tests in three stages. To overcome the bottleneck of traditional vehicle–infrastructure simulation, we construct a heterogeneous traffic subject simulation model, analyze the operation mechanism of mixed traffic, and present a swarm intelligence simulation method for mixed traffic as well as experimental results. In this way, we can steer the research and development of actual VICSs.
In the rudiment, infancy, and developing stages mentioned above, VICSs are becoming more heterogeneous, swarm-based, and intelligent. The change in its test requirements directly propels the development of the technologies for vehicle–infrastructure cooperative simulation and tests, which leads to vast differences in simulation approaches, objects, and test architecture among different stages. Hence, the research on vehicle–infrastructure cooperative simulation and tests will evolve from traditional singlevehicle intelligent macro-simulation and micro-simulation to mixed traffic subject simulation, from traditional virtual simulation to virtual–real interaction driven by AI and digital twin technology, and from traditional small-scale vehicle cooperative simulation under a simple scenario to large-scale vehicle-swarm behavior simulation underpinned by the Internet. Figure 1demonstrates the summary of the simulation development routes of traffic swarms, and Table 1presents the summary of the simulation and test requirements and feature evolution of VIC at different stages.
|
DownLoad:
CSV
In this stage, the feasibility of VIC was in urgent need of verification through a scientific approach. Thus, it was safe to claim that the simulation and test system was the inevitable product of the development of: Theoretical design framework and structure, wireless network communication protocol, and security control technology all need to be tested and verified. In the field test, if the safety of people and vehicles could not be guaranteed, the urban road infrastructure might be paralyzed. The costs of vehicles, drivers, and infrastructure were high, and the actual field test was rather complicated and involves a large sum of capital. By means of digital simulation of scenarios, the safety of VIC tests can be guaranteed while the consumption of manpower, material, and financial resources can be reduced. Therefore, the feasibility of VIC has to be verified by simulation and tests.
Against such a backdrop, simulation and test approaches emerged. In the rudiment stage, the research on the traffic simulation model was limited to the construction of a micro-model or a macro-model, and the simulation objects were mainly vehicles, signal lamps, and other individual traffic flow or macro-traffic flow. In other words, through the circuit construction and data simulation of a signal lamp, a vehicle, some functional hardware, or the overall traffic flow, the coordinated control performance in simple scenarios was verified. In this stage, the main goal was to realize the cooperative operation of vehicles under the reasonable design of traffic signals. The simulation models were subjected to some limitations, and the individual model's maneuverability, formulation, and authenticity remain poor.
Meanwhile, in terms of the verification of VIC functions, due to the imperfect sensor and hardware protocol interfaces, the framework based on black box testing, shown in Fig. 2, could directly verify the functions of the tested vehicles and roadside equipment. It was of great significance to verify the functions of cameras, stereo cameras, radars, and other sensors under the framework based on black box testing. Unfortunately, this stage was generally featured by individuation and low intelligence, and the technology employed was primary. The core was to verify the feasibility of VIC, and the verification was limited to the technical feasibility of an individual traffic subject. In addition, a virtual simulation test was a sheer digital simulation test method heavily dependent on the accuracy of the simulation model, which could hardly simulate the interaction between multiple traffic subjects.
With the rapid development of autonomous driving technology and high-performance computing technology, disruptive changes took place in vehicle–infrastructure cooperative simulation and tests. Apart from individual vehicles, the traffic environment of vehicles could also be simulated, and simulation and verification of small-scale swarms of vehicles under typical scenarios could be conducted. In this stage, the development of a wide range of emerging technologies picked up paces of development, such as vehicle behavior control, information interaction simulation, environment visual simulation, and simulation and test evaluation. Simulation and test technologies based on distributed architecture were extensively applied. These new technologies employed the full space-time dynamic information of traffic participants to dynamically and realistically represent various traffic phenomena such as traffic flow and traffic accidents, reproduce the space-time changes in traffic flow, and analyze the evolution of a single subject supported by a VICS. Hence, they could further verify the effectiveness of VIC in traffic efficiency and vehicle motion safety. The simulation framework is shown in Fig. 3.
The vehicle behavior control method is an important part of VICSs. Collecting external environment information, including external vehicle interference and road information and the vehicle's own position, speed, and other information, to control the vehicle's driving in the simulation system in real time can not only simulate the normal operation of the vehicle but also simulate the occurrence of accidents, save resources, and enhance the realism of the simulation.
In the infancy stage of VICSs, related work on behavior control simulation focused on the analysis and optimization of the macro-characteristics of traffic flow involving autonomous driving, the meso-control of isomorphic AVs, and the individual behavior control of AVs under typical traffic scenarios. The vehicle behavior control method could load different vehicle-following models, lane-changing models, and the like to preview the next-step operation state of vehicles by judging the behavior states such as acceleration, deceleration, lane changing, and steering and issue early warning information in advance in the possible event of danger. Moreover, it could send early warning information to the traffic simulation module and the simulation management and parameter evaluation modules in the form of data to avoid danger. Specific functions are shown in Fig. 4.
Information interaction simulation realizes the information interaction between vehicles and roadside equipment in the road network. It is responsible for receiving the vehicle states, roadside states, and execution location information of scenarios forwarded by simulation manager members. Communication simulation was carried out on V2V/V2I interaction information in the VICS under the IoV 4G, Wi-Fi, and the dedicated short range communication (DSRC) modes to feedback simulated communication network states. The simulation results were sent to the traffic simulation members as a basis for determining how to enforce the scenario.
The IoV communication simulation program in a VICS comprises a module for data interaction with the onboard unit (OBU) and the roadside unit (RSU). Hardware-in-the-loop simulation was undertaken in a communication module of the simulation system. In the simulation, the data of vehicles in some contexts and intersection signal lamps were mapped by hardware to realize the joint communication simulation of multi-subject nodes. The co-simulation scheme is shown in Fig. 5.
The visual simulation system, which is an important branch of VICSs, can output real traffic states in real time. However, different traffic elements have different emphases and details for vehicles, roads, onboard equipment, roadside equipment, and vehicle-ground information interaction process. Therefore, in the infancy stage, a distributed interactive visual simulation platform was often used to conduct in-depth research on V2V information interaction, V2I information interaction, road condition collection, and traffic flow control in combination with visual simulation technology. Run time infrastructure (RTI) based on HL, a visual simulation technology, is one of the typical representatives. It treats the VICS as a federated system while giving due regard to the principle of modulation, and each subsystem of the VICS is defined as a member. In terms of functions, the VICS is divided into six physical simulation federations, which not only have to complete their respective simulation tasks but also have to provide interactive information for other federations so as to form a distributed vehicle–infrastructure cooperative simulation system. The specific structure is shown in Fig. 6.
The performance test during the infancy stage was mainly based on single vehicle control under the support of roadside equipment and was carried out in the VICS of small-scale multi-vehicle coordination, so as to realize instantiation of each element of the VICS including vehicles, roads, humans, and information interaction networks in the simulation or real system as a unit/component with a specific function. The performance test framework in typical vehicle–infrastructure cooperative application scenarios is shown in Fig. 7.
On the basis of the above test scenario design, for the test of the vehicle–infrastructure cooperative simulation system, it is also necessary to construct a set of excellent vehicle– infrastructure cooperative test cases to maximize the number of system defects spotted on the premise of minimized test work. It is worth pointing out that in this stage, the application scenario of VIC was dominated by the cooperation between roadside equipment and single subjects and small-scale subjects, with the main communication modes being V2I and V2V communication. Some typical application scenarios are shown in Table 2.
|
DownLoad:
CSV
After the test cases were generated, the generation and optimization of test sequences became inevitable. According to the working mode conversion and operation scenarios of the vehicle–infrastructure cooperative simulation system, the test sequence was formed by the connection of the test sub-sequences in series. Methods to acquire a test sequence were based mainly on timed automata, genetic optimization, ant colony optimization, and the immune firefly algorithm (IFA).
In addition, the reliability of VICSs should be verified and evaluated after the testing. Analytic hierarchy process (AHP), data envelopment analysis (DEA), fuzzy comprehensive evaluation, neural networks, and the gray evaluation method were generally used to analyze and verify the feature correlation between evaluation requirements and VICS. In this way, the evaluation index system could be established to make a large number of factors that were coupled and restricted each other orderly and hierarchical. In the case of early warning assistance, Fig. 8shows the impact of different penetration rates of CVs on the number of successful early warnings in small-scale traffic, and Fig. 9displays the change in the minimum vehicle distance under different vehicledistance positioning errors (measured value minus the real value) during early warning for braking. It can be seen that in this stage, the vehicle–infrastructure cooperative simulation technology has been greatly improved in traffic simulation and function verification.
In the infancy stage, the distributed simulation and test technology of VIC achieved rapid development, and its performance became more intelligent, which laid a solid theoretical foundation for the follow-up test technology. However, with the disruptive development of autonomous driving technology, modern transportation would undergo tremendous changes. The traditional vehicle–infrastructure cooperative simulation system was often limited to the technical feasibility verification of a single vehicle. It lacked an in-depth understanding of the characteristics of traffic swarm behaviors and had not yet considered VIC as a large system covering the whole traffic environment. More accurate vehicle models, more intelligent control methods, more realistic testing environments, more flexible test approaches, and richer test scenarios were in urgent need.
In the wake of the rapid development of 5G communication, AI, multi-sensors, and other technologies, autonomous driving technology is applied on all fronts. The vehicle driving mode is evolving from human driving and assisted driving to autonomous driving, human-vehicle mixed driving, and advanced unmanned driving. Modern transportation will usher in a large number of AVs with different standards, architectures, and different levels of intelligence. In this stage, the simulation and test of VIC call for swarm intelligence and large-scale development.
Therefore, digital technologies such as virtual–real interaction and hardware-in-the-loop technology can well test and evaluate AVs in the simulation process of multi-physical quantity, multi-scale, and multi-probability and can well provide test and evaluation data for the autonomous driving test. Therefore, the simulation and test technology of vehicle– infrastructure cooperative swarm intelligence becomes the key to overcoming the bottleneck of the existing vehicle– infrastructure cooperative simulation and test technology.
The method framework proposed in this paper is shown in Fig. 10. It is supported by the underlying theory of traditional vehicle–infrastructure cooperative simulation and tests. It takes heterogeneous traffic subject modeling as its first layer, the simulation of mixed traffic swarm intelligence as its second layer, and the virtual–real interaction test as the third layer. It aims to realize large-scale vehicle–infrastructure cooperative simulation and tests in the future. Specifically, this method framework can study the synchronized driving mechanism of asynchronous data for twin test subjects, construct compact scene space, build a large-scale virtual–real interactive operation environment that is expandable and highly reliable and integrate virtual and real data, map the physical space of AV tests, and reflect the full cycle of the autonomous driving capability of corresponding physical vehicles. Simulation and tests are carried out on the perception capability, communication capability, decision-making capability, control capability, and positioning capability of AV swarms in the mixed traffic environment.
Subject heterogeneity is one of the core characteristics of the new mixed traffic system formed by the integration of VIC technology into road traffic. It reflects the difference in perception, interaction, decision-making, and control between vehicles of different levels of intelligence, which further results in the heterogeneity of vehicle movement. As a result, the operation mechanism of mixed traffic becomes too complicated to be analyzed. In this context, the traditional single vehicle simulation model cannot meet the behavior simulation needs of heterogeneous traffic subjects. It cannot represent the cooperative behavior characteristics of mixed traffic swarm intelligence, nor can it reflect the operation mechanism of the new mixed traffic. Therefore, it is necessary to thoroughly analyze the differences in functional characteristics and motion behavior characteristics between heterogeneous traffic subjects and build respective kinematic simulation models.
On the basis of grasping the main functional characteristics of heterogeneous traffic subjects, this section constructs a simulation model from the perspective of vehicle kinematics, analyzes the operation mechanism of mixed traffic, and establishes the behavior simulation framework for heterogeneous traffic swarm intelligence (Fig. 11). It is expected to provide vehicle-behavior model support for the simulation and test system of mixed traffic swarm intelligence in VICSs.
According to the different forms of perception, interaction, and decision-making control of vehicles in road traffic, heterogeneous traffic subjects are divided into HDVs, AVs, CVs, and CAVs, as shown in Fig. 12. For the simulation of the motion of heterogeneous traffic subjects, it is essential to analyze the function of vehicles of different types of subjects and study the relationship between traffic subjects and road environment.
Without the functions of autonomous control and connection, HDVs have to rely on human perception of the environment in the course of driving, and thus driving safety and efficiency are closely related to the human driving experience. Equipped with multiple sensors and autonomous control technology, AVs can, to some extent, assist drivers. CVs can realize the communication with roadside devices or other vehicles equipped with the connecting function. CAVs refer to AVs equipped with a connecting function, which can share state information through IoV technology and realize autonomous perception, decision-making, and control.
The operation of a conventional HDV depends mainly on the driver's decision-making and the underlying kinematics of the vehicle. IDM is a sophisticated dynamic model to describe HDVs. With well-defined physical parameters, it can well present the following behavior of HDVs. Therefore, it has been employed in all kinds of research. The vehiclefollowing function of IDM is written as
|
|
(1) |
where vn(t) denotes the speed of the vehicle nat time t; Δvn(t) denotes the difference in speed between vehicle n− 1 and vehicle nat t; hn(t) denotes the space headway of the vehicle nat t; ais the maximum acceleration of the vehicle; s0is the minimum braking distance of the vehicle; v0is the free flow speed; Tis the safe time headway; bis the expected deceleration of the vehicle, and lis the body length.
Before high-grade autonomous driving technology becomes fully-fledged, the application of autonomous driving assisting technology enables vehicles to improve driving performance in road cruise scenarios. Although the traditional constant-speed cruise allows a vehicle to drive at a fixed speed, the risk of a crash cannot be avoided on its own. Hence, it is equipped with an adaptive cruise module to automatically adjust its expected speed by autonomously perceiving the position and speed of the targeted vehicle.
Adaptive cruise control (ACC) can make a vehicle run more stably with less fluctuation, which is what is lacking inhuman driving, and ACC mirrors the difference in motion between AVs and HDVs. Thus, an adaptive cruise model for AVs is proposed, whose equation is given by
|
|
(2) |
where tadenotes the expected time headway of the ACC vehicle; k1denotes the control coefficient of the vehicle distance error, and k2denotes the control coefficient of the speed difference.
CVs can operate efficiently with the expected spacing and speed through information interaction between vehicles, which can improve the road carrying capacity and driving safety and reduce fuel consumption and environmental pollution. CV running characteristics are mainly manifested in vehicle platoons. Hence, a CV platoon model with time-varying delays is proposed. Information-flow topologies are established through V2V communication and emerging 5G, and the communication topology is designed as a two-way leader-follower pattern. This enables any following vehicle to receive information on the position, speed, and acceleration of the leading vehicle, and it also makes it possible for vehicles next to each other to exchange information with each other. Moreover, it makes V2V communication interaction more frequent, communication topology more complex, and the output of the simulation model closer to reality. The distributed control model of a CV platoon with time-varying delays is as follows:
|
|
(3) |
where ui(t) represents the control input of the vehicle iat tafter the feedback linearization of the dynamic CV model; xi[t− τ(t)], vi[t− τ(t)], and ai[t− τ(t)] denote the position, speed, and acceleration of the vehicle iin the CV platoon at t− τ(t), respectively. 
CAVs integrate the functions and features of AVs and CVs. CAVs can achieve efficient adaptive target detection and precise control through real-time information sharing. When the driving information can be transmitted through V2X technology, a CAV runs in a cooperative ACC (CACC) mode. In this case, the vehicle is known as a CACC vehicle; otherwise, the CAV runs in an ACC mode, and in this case, the vehicle is known as an ACC vehicle. Within each time step, the ACC/CACC vehicle determines the acceleration or velocity to be output according to the speed and position of itself and its targeted vehicle in the previous time step, as well as its desired time headway and cruise speed. The CACC model is expressed as
|
|
(4) |
where Δtdenotes the control step of the CACC system; en(t) denotes the error between the actual vehicle distance and the expected vehicle distance of the vehicle nat t; tcis the expected inter-vehicle time headway parameter of the CACC vehicle; kddenotes the differential control coefficient of the vehicle distance error.
Characteristics comparison of heterogeneous traffic subject models is shown in Table 3. The differential micromovement characteristics of heterogeneous traffic subjects affect the operation mechanism of traffic flow at the macro level. The heterogeneous traffic subject model can analyze the interaction among flow, density, average speed, and penetration rates of heterogeneous vehicles in mixed traffic flow, which can provide traffic flow model support for the simulation and test of mixed traffic swarm intelligence in the VIC context.
|
DownLoad:
CSV
In road traffic simulation, the behavior simulation of all heterogeneous traffic subjects reflects the operation mechanism of mixed traffic flow. In the mixed traffic flow composed of HDVs and CAVs, the heterogeneity of CAVs in motion is manifested in the driving modes of CACC and ACC mainly in terms of expected time headway, dynamic response types, and feedback forms. Let pmbe the proportion of HDVs in the mixed traffic flow, pcthe proportion of CACC vehicles, and pathe proportion of ACC vehicles. In the equilibrium state of homogeneous traffic flow, all vehicles have the same equilibrium speed and space headway. In mixed traffic flow, it is believed that all vehicles have the same equilibrium speed due to the limitation of maximum speed, but the equilibrium space headway is determined by respective kinematic models. By establishing the functional relationship between equilibrium state speed and space headway in traffic flow, we can obtain the fundamental diagram function of mixed traffic flow density and speed, which can be expressed as
|
|
(5) |
where kis the density of mixed traffic flow; fm(ve), fc(ve), and fa(ve) denote the equilibrium speed-space headway functions of HDVs, CACC vehicles, and ACC vehicles, respectively, and vedenotes the equilibrium speed.
In the absence of a communication function, HDVs cannot exchange information with CAVs. Therefore, it is set that CAVs can only monitor the driving condition of the vehicle ahead. In accordance with the dynamic characteristics of traffic flow, the relative spatial positions of vehicles in mixed traffic flow are random. If a CAV follows an HDV, it degrades to an ACC vehicle, and if a CAV closely follows an ACC or CACC vehicle, the CAV maintains the CACC driving mode.
When the penetration rate of CAVs in the mixed traffic flow is set as p, the proportion of different types of vehicles in the mixed traffic flow can be obtained, namely,
|
|
(6) |
When the functional relationship between equilibrium speed, density, and space headway in traffic flow is established, the mixed traffic flow–density curve is given by
|
|
(7) |
where qdenotes the traffic flow.
In this way, the mixed traffic flow–density curve under different CAV penetration rates can be obtained, as shown in Fig. 13. It can be seen that with the increase in CAV penetration rates, the maximum capacity of mixed traffic flow is gradually increased, and the traffic efficiency is gradually improved. Therefore, the mixed traffic environment composed of heterogeneous traffic subjects can improve road traffic efficiency by raising the proportion of CAVs in the mixed traffic flow.
The traditional traffic environment becomes a new mixed traffic environment after the integration of heterogeneous vehicles with different levels of intelligence. Therefore, the traditional simulation and test technology for VIC is unable to solve the new traffic problems caused by mixed traffic flow. In addition, typical application scenarios of VIC cannot fully cover the functional characteristics of mixed traffic. In this section, the virtual–real interactive simulation method and swarm intelligence algorithm are employed to test and analyze vehicle behavior in different mixed traffic scenarios.
By integrating some/all of the actual test objects in the autonomous driving system into the virtual test environment, the virtual–real interactive test technology can carry out comprehensive low-cost and multi-scenario tests that can be debugged. The virtual–real interactive test technology can be categorized as the hardware-in-the-loop test and vehiclein-the-loop test. The whole system or some parts of the hardware-in-the-loop test are real, such as the module tests of environment perception, decision-making planning, and control execution. In this mode, most of the functions of the test subject are implemented by software, and the scope of application is still relatively limited. The vehicle-in-the-loop test integrates all components and systems into the vehicle and connects the whole vehicle as physical hardware to the virtual test environment for testing. In this paper, a traffic simulation platform framework suitable for vehiclein-the-loop virtual–real interaction tests is designed. It is mainly composed of virtual space, data space, and physical space, with the overall design framework shown in Fig. 14, where 
The virtual space constructs a standardized test scenario library and a simulation road network with easy expansion and adjustable granularity in terms of the scenario category, traffic subject, road type, natural environment, behavior control, etc., for safety and efficiency. The traffic characteristic data are extracted by the integration of the multi-source sensor data in the data space, and the bidirectional synchronous transmission of the state data is carried out in the virtual space and the real space. The physical space uses the realtime state characteristic data transmitted from the physical space to dynamically map and reconstruct the measured objects in the virtual space in real time and uses the multisource sensor technology to collect the full-cycle operation data of the traffic subjects to ensure the integrity of data sources. Virtual space, data space, and physical space are integrated through wireless communication to realize the simulation, testing, and verification of swarm cooperation behavior in virtual–real traffic states.
If the synchronization of the vehicle–infrastructure cooperative virtual–real simulation and test system is to be improved, the operation mechanism of the test subject has to be confirmed on the basis of the virtual–real simulation and test process. For the problem of time migration in the data transmission of test subjects, a clock-error estimation strategy is designed to correct the measurement time deviation in the transmission of virtual and real states. On this basis, a multiscale filtering synchronization method is proposed for the virtual–real test system by the combination of Kalman filtering and constant rotation rate and constant acceleration models. The method dynamically compensates for the realtime running states of the virtual–real subjects and maintains good synchronization stability, as shown in Fig. 15.
As physical space and virtual space are connected through wireless communication networks, frequent data interaction will greatly increase the communication burden of the test system. However, the calculation scale of simulation events in virtual space needs to maintain a high frequency to be close to the real traffic scenario in physical space. Given that fvis the sampling frequency transferred by the virtual space, and fris the sampling frequency observed by the physical space, then, fv> fr. The running state Xmof the twin test subjects at time min the virtual space can be described as
|
|
(8) |
where 
The synchronized mode of the virtual–real system state is as follows: When the virtual space does not receive the updated observations of the real space at m, the system calculates the state at the next moment with a high-resolution time scale, namely that fvis selected to propel the virtual state of the system. When the virtual space receives updated observations of the real space at m, the sampling frequency frobserved by the physical space system is selected to propel the virtual state of the system, i.e.,
|
|
(9) |
where g(·) denotes the multi-scale filtering synchronization function of the state of the virtual–real test system.
The lateral and longitudinal test trajectories are selected to optimize the synchronization performance of the twin test system by the model-based dead reckoning method, the discrete Kalman filtering method, and the multi-scale filtering method separately. Specific trajectory synchronization errors are shown in Fig. 16.
Compared with the model-based dead reckoning method, the multi-scale filtering synchronization method of states has obvious advantages in lateral trajectory synchronization. Limited by the communication gap between virtual and real information, the discrete Kalman filtering method cannot provide continuous and effective synchronization information, and the continuity and smoothness of the twin subject trajectories are poor. Thus, the synchronization effect of this method is relatively poor.
The multi-scale filtering synchronization method integrates the simulation test system and the real traffic environment and eventually forms a virtual–real test and verification environment including real traffic operation, twin state synchronization, and virtual traffic operation. In this way, the method supports the test and verification of multivariate-in-the-loop cooperation of vehicle–infrastructure traffic test equipment, core component function modules, and virtual simulation subjects.
The VICS test based on simulation scenarios can deconstruct and reconstruct scenarios flexibly and realize the targeted test of dangerous scenarios, thus reducing the cost of field tests and ensuring the efficiency and completeness of VICS simulation. The core of testing is the construction of test scenarios, and the analysis of scenario elements is the quantitative assessment of the coverage ability of scenario testing. The scenario for virtual tests should meet the requirements that the characteristics of each element of the scenario can be quantified, and the scenario can be reproduced on the test software; the real traffic state can be reflected by it to a certain extent. In short, the scenario is composed of each frame of the scenario, describing a series of actions and events leading to the outcome, which includes static elements, dynamic elements, and AVs.
As the autonomous driving level upgrades, the traditional simulation test environment and test methods cannot meet the needs of mixed traffic simulation and tests in the VICS. Moreover, there are no clear regulations on the rapid construction and simulation test of vehicle–infrastructure cooperative mixed traffic scenarios. Therefore, according to the behavior simulation requirements of the vehicle– infrastructure cooperative mixed traffic swarm intelligence formed by the mixing of heterogeneous traffic subjects, the functional characteristics of different traffic scenarios are extracted hierarchically, and the scenario elements are decomposed, as shown in Fig. 17, where FI, j′is the j'-th element of Layer I.
Given the importance of scenario elements, the test case of the mixed traffic scenario of VIC is reconstructed. Finally, the test case is executed in the simulated environment to support the functional characteristics of the scenario required by swarm behavior simulation. Figure 18shows the relationship between simulation requirements and scenario cases of vehicle–infrastructure cooperative mixed traffic swarm intelligence.
Common research methods for scenario test-case generation mainly include the test matrix method, Monte Carlo method, combinatorial testing method, worst scenario evaluation method, and game theory method. The combinatorial testing method can achieve scenario coverage through the interaction between scenario elements and generate as few scenario test cases as possible to cover scenario elements of more dimensions. It can satisfy the need for testing and verifying the functional characteristics of the mixed traffic scenarios of VIC. In addition, it can improve the coverage of test cases based on the combination of all elements of the scenario. The Monte Carlo method can extract key scenario elements on the basis of real natural driving data, simulate the probability distribution of the parameters of scenario elements, restore the real scenario, and randomly generate test cases for primitive scenarios.
Under the support of the above method, if the comprehensiveness of the test process in the digital twin running environment is to be guaranteed, a virtual–real interaction test mechanism has to be constructed for typical application scenarios of vehicle–infrastructure cooperative swarm control. Considering the test object, road type, natural environment, and behavior control objectives, more than 90 traffic scenarios are constructed. This greatly improves the coverage of traditional simulation and test scenarios and meets the test coverage requirements of i-VICS. Some scenarios are shown in Table 4.
|
DownLoad:
CSV
The construction of mixed traffic operation scenarios provides the basis and environment for the test of swarm intelligence simulation. In the course of scenario testing, the interaction between vehicles and environment forms a closed loop, which has a great influence on each other, and the test scenario has an impact on the state of the vehicle. By calling various mixed traffic scenarios in the virtual–real interactive scenario library of different scenarios, we can simulate the vehicle swarms, and considering the virtual–real interactive simulation mode, we can achieve the purpose of vehicleswarm decision-making control and promote the development and perfection of intelligent transportation systems.
On the basis of single-vehicle dynamics/kinematics and multi-vehicle cooperative behavior modeling, this section carries out analysis and modeling for the operation control optimization of heterogeneous traffic swarms. As the intelligent control mode of heterogeneous traffic swarms is closely related to the involved scenarios, in this section, the idea of scenario slicing is adopted to break the large-scale mixed traffic swarm operation scenarios down into elements with different complexity levels, such as road sections and intersections, to precisely reflect the regional swarm simulation control mechanism. As scenarios can be combined with each other, by scenario slicing, basic scenario elements can be combined and reconstructed. In this manner, regional scenario types can be increased, and the test range can be expanded, so as to form regional testing scenarios with different structures and comprehensive functions. That is why we started with typical basic scenarios to review swarm simulation methods.
In the current intersection control mode, the capacity of a standard eight-lane four-way intersection is about 3 600 veh·h−1. In the urban traffic system, except for the morning and evening rush hours during which the maximum capacity of the intersection can be reached, other periods may not see such a high traffic flow. Therefore, the signalized intersection can basically divert the current urban traffic flow.
Given the change in the mixed traffic environment, a coupled vehicle signal control (CVSC) method is proposed to optimize the traffic signal timing and driving trajectory of CAVs so as to improve traffic efficiency and save energy. Meanwhile, the signal timing is optimized continuously to minimize the total delay of the intersection. The CVSC method uses the fundamental diagram model for mixed traffic flow as the connection to realize the coordinated control of signalized intersections and CAVs with the objectives of efficiency improvements and energy saving, respectively.
In line with the above signal control methods, a CAV/HDV signal cooperative control method for mixed gap coupled intersection is proposed considering the static characteristics of intersections, the characteristics of intersection signal control, the kinematics models of different types of AVs, and the operation mode of i-VICS. Through the coupling of the signalized/non-signalized and different types of vehicle movement processes, and in light of the coupling process of lane-level signal control and intersection parallel-gap control, a gap traffic mechanism of CAVs/HDVs at intersections is established. Given the start-stop delay model of vehicles and driver reaction time, the lane-level signal control strategy of HDVs is determined by the vehicle behavior prediction and the HDV gap selection mechanism of gap delays to achieve the dynamic gap coupling of CAVs/HDVs at intersections. As shown in Fig. 19, TFCdenotes the minimum first conflict time between vehicles; SLdenotes the length of the lane-changing decision-making area, and SSdenotes the length of the speed adjustment area. With the control method for one individual intersection as the basis, we could achieve consecutive intersection signal coordination and optimization from points to areas, which then spreads to regional signal control. Thus, the intelligent simulation and control optimization of mixed traffic swarms is realized to improve the traffic capacities of areas, as shown in Fig. 20.
In the urban road operation environment, the operation control of the platoon of heterogeneous traffic subjects is an important method to improve traffic efficiency and safety. The simulation of the intelligent platoon cooperative control optimization method can effectively alleviate urban traffic pressure and reduce congestion and accidents. The operation modes of intelligent platoons in road sections have multiple characteristics. In this paper, the platoon marshaling of different driving behavior of vehicles in road sections is optimized for mixed traffic flow environment with parallel autonomous driving and autonomous driving.
In terms of the optimization of vehicle sequencing in a platoon, the characteristics of drivers' continuous vehiclefollowing behavior are obtained, and then, the differences in driving behavior characteristics of the drivers following the vehicles ahead are taken into full account for the full understanding of the behavior characteristics of the drivers following different types of drivers. As a result, a sequencing optimization method for vehicles in a platoon is presented on the basis of an in-depth understanding of vehicle-following behavior characteristics. In terms of control strategy for vehicles entering the queue, upon the acquisition of drivers' sporadic lane-changing behavior characteristics and the drive of gap control theory, a customized control strategy optimization method is proposed. In terms of vehicle separation control of a platoon, efforts are made to study the impact of the understanding of vehicle platoons in the surrounding individual drivers' behavior characteristics on reducing the pressure of roadside equipment at intersections; the actual driving characteristics of platoons in the front road section of intersections and the existing objective problems are under full consideration. A platoon vehicle separation control method based on stress driving behavior cognition is proposed according to the characteristics of drivers' lanechanging behavior.
Given the operation control methods of different driving behavior for the above three types of intelligent platoons and the research on the control methods of multi-platoon cooperative operation in multi-platoon driving scenarios, a control mechanism is proposed for the cooperative operation of the formation of road sections in the mixed traffic context. It is shown in Fig. 21.
To meet the requirements of rapid testing of the effectiveness of vehicle–infrastructure cooperative swarm control, we designed a virtual–real traffic simulation platform, through which simulation, test, and evaluation of swarm intelligence behaviors were carried out in traffic operation scenarios such as road sections and intersections. The virtual–real simulation platform was mainly composed of a data generation system, an algorithm testing system, and a scenario verification system. The data generation system was connected to real space, and the scenario verification system was connected to virtual space. The overall system framework is shown in Fig. 22.
On the basis of the above simulation platform, the test scale of vehicle–infrastructure cooperative simulation in mixed traffic scenarios has been significantly improved, evolving from the initial simulation of mixed traffic scenarios with only nine nodes and 500 traffic subjects to the current simulation of large-scale swarm cooperative behavior with 150 nodes and 2 000 traffic subjects, as shown in Fig. 23.
With the virtual–real simulation platform, we verified and evaluated the effectiveness of the proposed mixed gap-coupled intersection signal cooperative control method for a typical eight-lane intersection and compared it with the classical CACC model. The test results are shown in Fig. 24 (a). It is proved that the method can improve the overall performance of traffic flow at signalized intersections and performs well under moderate traffic flow, which can adapt to unbalanced traffic requirements.
A comparison experiment was conducted on the intersections with parallel gap control, timing control, and stop sign control under the same road channelization conditions. The intersections were also eight-lane, and the input was 350 veh·h−1for each steering vehicle. It can be seen from Fig. 24 (b)that the delay of vehicles at intersections controlled by stop signs continues to rise, which reflects the non-periodicity of stop sign control methods and the limitations of their effects on traffic flow control. Vehicle delays at signalized intersections have a certain periodicity, the repetition of which depends on the signal control period. When vehicles passed an intersection when the signal light was green, vehicle delays decreased. When vehicles passed an intersection when the signal light was red, vehicle delays increased.
Given the CAV/HDV signal cooperative control method of the designed mixed gap coupled intersection, the road sections were divided into sub-networks, and the speed of the sub-network within the green-wave band with the maximum number of continuous passing vehicles was guided, so as to control and optimize the green-wave band signal of the continuous intersections. Then, this method extended to regional swarm control from points to areas.
By the virtual–real simulation platform, we obtained drivers' sporadic lane-changing behavior characteristics. On this basis, in the mixed traffic environment, the simulation and tests were carried out for individual vehicles entering the queue and platoon cruising in the bidirectional dual-lane environment and platoon separation control in the intersection environment according to the gap control theory.
To select the same queue-entering speed of vehicles, we set it to 10, 20, and 25 m·s−1separately and analyzed the relationship between safety gap and different speed combinations of the overtaking vehicle and the vehicles in the lead and following behind with a safety gap, as shown in Fig. 25. The average speed of step-by-step vehicles overtaking to enter the queue is 14.02 m·s−1, which is 16.8% higher than that in the initial following state, and the maximum average speed of overtaking vehicles is 14.94 m·s−1.
When the overtaking vehicle was following the platoon, there was an overall acceleration phase, but the speed of the whole accelerated platoon was less than the maximum average speed. In other words, the speed of the following vehicle was less than the overtaking speed in the opposite lane. Here, the speed of the platoon after acceleration was set to 14.92 m·s−1.
The average speed of vehicles in the platoon in the process of step-by-step overtaking by an individual vehicle was counted, as shown in Fig. 26. It can be seen that the average vehicle speed of the platoon was 16.27 m·s−1, which is 9.04% higher than that of free acceleration.
In the platoon cruising simulation and test, the selected platoon-following control modes in the mixed traffic scenario were the CACC mode and the platoon-operation optimization mode based on the characteristic depth of vehicle-following behavior. The data within 120 m of the freeway simulation scenario were selected to obtain the multi-vehicle movement mode, and the test results are shown in Fig. 27. It can be seen that the cruising speed and the running speed of the surrounding vehicles after the optimization of the platoon operation mode are significantly higher than those before the optimization, which indicates that the method is conducive to improving the operational efficiency of the platoon.
The testing scenario of platoon separation was designed as follows: The platoon consisting of seven CAVs ran in the road sections in front of the intersection. It was set that each vehicle was a separated vehicle, with the length being 5 m on average, the time headway between vehicles being 1 s, the platoon speed being 0 m·s−1, and the remaining green light duration being 5 s. The position of an AV was in parallel with the fourth vehicle of the platoon, and the acceleration of the vehicle separated from the platoon approaching the intersection was obtained. According to the results, compared with the results of the vehicle-separation control method in the unknown driving behavior situation (Fig. 28(a)), the application of the platoon-vehicle separation control method based on the stress driving behavior recognition (Fig. 28(b)) reduces the acceleration variation range of the separated vehicles by 37% to 72%.
Trajectories of vehicle swarms with different average running speeds were selected for simulation and test analysis. A total of 160 groups of trajectories meeting the speed conditions in the same running range were screened. The motion behavior of traffic swarms is shown in Fig. 29. The swarm behavior was analyzed by the distributions of vehicle trajectories at different average speeds in both lateral and longitudinal directions. The important indicators such as the communication delay and traffic operation efficiency in the current mixed traffic scenarios were simulated and tested through vehicle trajectory data in line with the virtual–real simulation platform.
Communication delay is one of the key indicators in vehicle–infrastructure cooperative swarm control. If the normal operation of AVs and CVs is to be guaranteed, it is necessary to simulate and test communication delays. The transmission delay results under different traffic scenarios are shown in Fig. 30. When there are 100 CV nodes in the scenario, the transmission delay is relatively greater. This is because when there are few vehicles, the nodes are scattered and sparse, and there are relatively few reachable routes from the source node to the destination nodes, which results in a greater transmission delay. When the number of vehicles increases, the delay becomes less obvious. When the number of vehicles continues to increase, however, the channel will become congested, and the probability of transmission failure will grow. As a result, the probability of a greater back-off counter for the data flow control (DFC) channel connecting to the protocol of the media access control (MAC) layer will increase to lead to a larger delay. The overall performance of the simulation is good, and the average delay of all traffic scenarios remains at about 0.059 s.
Network throughput mainly hinges on the number of CVs. When there are plenty of vehicle nodes, there are relatively more vehicles for information interaction. Although the communication media will be relatively congested, the network throughput will still grow. The throughput will prominently increase along with the rise in the number of vehicles. Speed has little impact on the throughput mainly because the multi-communication-mode competition mechanism based on neural networks will switch to a suitable communication mode such as DSRC in real time. The average throughput of all traffic simulation scenarios is about 104 862 bit·s−1.
When the traffic operation efficiency index was tested, and the parameter values of other scenario elements remained unchanged, the penetration rate of CAVs was changed to analyze the changing trend of the efficiency of the mixed traffic scenario and verify the relationship between efficiency and penetration rate of heterogeneous traffic subjects in the mixed traffic scenario. Efficiency is represented by traffic flow, and greater traffic flow means higher traffic efficiency. In the test scenario, an emergency happened to a vehicle in the road network (for example, the vehicle broke down on the road, or a traffic accident took place). Vehicles equipped with VIC were in advance informed of the emergency ahead through V2V or V2I communication and accordingly changed lanes or chose to bypass the road section. Vehicles without VIC began to slow down and changed lanes when they approached the site of the accident. Test results were selected with the penetration rates of CAVs being 20%, 40%, 60%, 80%, and 100%, as shown in Fig. 31. The followings are what test results reveal. As the penetration rate decreases, vehicle delays and stop times increase, which indicates lower traffic efficiency. When the penetration rate is zero, all the vehicles in the simulated road network are HDVs, whose driving behavior is completely uncontrollable and inevitably leads to a decrease in traffic efficiency. When the penetration rate is 1, all the vehicles in the simulated road networks are CAVs. At this time, these vehicles are all in the CACC driving mode and maintain real-time communication with roadside equipment. Vehicles are guided by roadside equipment to avoid road sections suffering from congestion and accidents and drive in formations and eventually pass through intersections efficiently, with the highest efficiency achieved. The vehicle–infrastructure cooperative swarm simulation method enjoys a good ability to predict short-term traffic flow and can improve prediction accuracy, which provides scientific and effective data support for short-term traffic flow analysis and prediction.
The virtual–real simulation of mixed traffic swarms can provide a quantitative basis for road traffic management and theoretical guidance for urban construction. In the virtual– real traffic swarm simulation system, the traditional traffic simulation index has been significantly improved, and the simulation time granularity of i-VICS has been shortened from less than 0.5 m to less than 0.1 m for a single fast solution for swarm cooperation control in different scenarios. With the above simulation platform, the simulation has been significantly improved, evolving from the VICS simulation with only nine nodes and 500 traffic subjects to behavior simulation of swarm cooperation with more than 90 simulated scenarios related to traffic safety and efficiency, the penetration rates of heterogeneous traffic subjects adjustable from 0% to 100%, 150 nodes, and 2 000 traffic subjects.
With the development and application of autonomous driving technology, vehicles with different intelligence levels are integrated into road traffic to form a new mixed-traffic environment. Recent years has witnessed emerging simulation and testing technologies, such as cyber-physical system (CPS), digital twins, meta-universe, and virtual–real interaction, which have become an important development trend in the field of intelligent transportation simulation and tests. Therefore, this paper thoroughly analyzes the research status and development trends of vehicle–infrastructure cooperative simulation and tests, and the following conclusions are reached.
(1) A VICS is a large-scale real-time distributed system involving multiple factors and layers. Only on a feasible basis can it effectively exhibit its performance and values. Field testing is subjected to some key problems, such as high costs, poor security, limited test scenarios, and monotonous testing methods. Therefore, the vehicle–infrastructure cooperative simulation and test technology is the very means to expedite the application and implementation of VICSs.
(2) The evolution of the simulation and test of VICSs can be divided into three stages, namely, the rudiment, infancy, and developing stages. The key to the rudiment stage was to verify the feasibility of VICSs. Overall, the test technologies during this stage remained low-intelligence and individualized. The key technologies in the infancy stage were more efficient and intelligent, represented by the distributed simulation of high-level architecture. These technologies promoted the development of the visual integrated simulation of VICSs. Upon the explosive development of autonomous driving technology, the requirements of vehicle– infrastructure cooperative simulation and tests in the new mixed traffic environment are rapidly evolving towards a larger swarm, higher intelligence, and larger scale.Virtual–real interaction and hardware-in-the-loop are becoming powerful simulation and test approaches for the development of VIC technology in the near future.
(3) The current traffic environment has entered the stage of mixed traffic. In order to rise to the challenge of large-scale vehicle–infrastructure swarm simulation and tests, we propose a simulation and test method for mixed traffic swarm intelligence based on virtual–real interaction. Through this method, a behavior model of heterogeneous traffic subjects is built, and the operation mechanism of mixed traffic is analyzed. The simulation and control of the cooperative behavior of mixed traffic swarm intelligence are realized according to the swarm intelligence method. Owing to the advantages of virtual–real interaction technology, scenario generation and simulation technology for mixed traffic swarm intelligence came into being, and breakthroughs are made in the synchronization mechanism of simulation and tests. Verification based on test results indicates that this method can improve the efficiency, scale, and coverage of vehicle–infrastructure cooperative mixed traffic simulation and tests.
We will confront new challenges brought by the simulation and test of i-VICS and the test requirements in the context of typical traffic scenarios such as large-scale open/semi-open traffic environment and freeway/multi-intersections and the cross-scenario test requirements of swarm behavior. It becomes necessary to integrate the virtual–real interaction of vehicle–infrastructure subjects and the data simulation technology for the operating environment to form simulation and test logic that meets the needs of scenario tests and accelerate the construction of large-scale simulation systems for swarm intelligence and key technology research for mixed traffic environments. The future research trends of vehicle–infrastructure cooperative simulation and tests are as follows.
(1) Efforts should be made to accelerate the construction of new digital infrastructure, build a full-service, full- spatiotemporal, and full-coverage ubiquitous network perception system, form fully-covered digital scenario information, reconstruct real scenarios in virtual space, and verify test algorithms.
(2) Brain-inspired intelligence technology should be employed to develop sophisticated fully-autonomous vehicles. After full spatiotemporal vehicle–infrastructure information is obtained, the swarm intelligence decision-making control method integrated with brain-inspired intelligent vehicles is to be deployed on the vehicle–infrastructure cooperative virtual–real simulation platform, and test verification is to be carried out.
(3) Measures should be taken to build a multi-layer virtual–real real-time simulation platform supporting the synchronous execution of two-way states in virtual and real space, accelerate the "near-zero scenario" mapping and reconstruction of large-scale mixed traffic swarm cooperation behavior, and promote the parallel development of vehicle–infrastructure cooperative test/operation.
(4) Steps should be taken to promote the research on vehicle– infrastructure swarm decision-making control method and the enforcement of related technologies, build a fully-fledged demonstration area of vehicle–infrastructure cooperative application that upholds deep application and integration of virtual and real data, and form a set of scientific and complete theoretical systems and promotion schemes of the virtual– real simulation and tests of vehicle–infrastructure cooperative swarm intelligence. These measures will provide theoretical and methodological guidance and platform environment support for China to independently grasp the key core technologies of VICSs to propel the realization of the goal of building China's strength in transport.
| [1] |
ZHANG Yi, YAO Dan-ya, LI Li, et al. Technology and application of intelligent vehicle-infrastructure cooperation systems[J]. Journal of Transportation Systems Engineering and Information Technology, 2021, 21(5): 40-51. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YSXT202105006.htm
|
| [2] |
TIAN Bin, ZHAO Xiang-mo, XU Zhi-gang, et al. NRT-V2X: adaptive data dissemination protocol for traffic efficiency of connected and automated highways[J]. China Journal of Highway and Transport, 2019, 32(6): 293-307. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL201906030.htm
|
| [3] |
PETTY K F, NOEIMI H, SANWAL K, et al. The freeway service patrol evaluation project: database support programs, and accessibility[J]. Transportation Research Part C: Emerging Technologies, 1996, 4(2): 71-85. doi: 10.1016/0968-090X(96)00001-0
|
| [4] |
HIROSHIMA Y. Development of collision danger reasoning algorithm, Part 5: the basic technology for ASV (advanced safety vehicle)[J]. JSAE Review, 1995, 16(1): 221-224.
|
| [5] |
FUKUSHIMA M. The latest trend of V2X driver assistance systems in Japan[J]. Computer Networks, 2011, 55(14): 3134-3141. doi: 10.1016/j.comnet.2011.03.012
|
| [6] |
WEIB C. V2X communication in Europe—from research projects towards standardization and field testing of vehicle communication technology[J]. Computer Networks, 2011, 55(14): 3103-3119. doi: 10.1016/j.comnet.2011.03.016
|
| [7] |
ZOU Zhi-jun, YANG Dong-yuan. Preliminary study on dynamic traffic microsimulation[J]. Journal of Tongji University (Natural Science), 1999, 27(3): 305-308. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-TJDZ199903011.htm
|
| [8] |
CHUNG-MAN HO C M, GENTLE J E. A comparison of clock pulse and event algorithms for simulation of traffic flow[J]. ACM SIGSIM Simulation Digest, 1976, 8(1): 53-55. doi: 10.1145/1103189.1103204
|
| [9] |
WADA S, HAYAKAWA H. Kink solution in a fluid model of traffic flow[J]. Journal of the Physical Society of Japan, 1998, 67(3): 763-766. doi: 10.1143/JPSJ.67.763
|
| [10] |
DEL CASTILLO J M, PINTADO P, BENITEZ F G. The reaction-time of drivers and the stability of traffic flow[J]. Transportation Research Part B: Methodological, 1994, 28(1): 35-60. doi: 10.1016/0191-2615(94)90030-2
|
| [11] |
PAPAGEORGIOU M. Some remarks on macroscopic traffic flow modelling[J]. Transportation Research Part A: Policy and Practice, 1998, 32(5): 323-329. doi: 10.1016/S0965-8564(97)00048-7
|
| [12] |
DANIŠOVIČ P, JANČAŘÍKOVÁ E, ŠRÁMEK J, et al. Fire spread models and tunnel traffic & operation simulator[J]. Procedia Engineering, 2017, 19(2): 92-95.
|
| [13] |
KRONJÄGER W, KONHÄUSER P. Applied traffic flow simulation[J]. IFAC Proceedings Volumes, 1997, 30(8): 777-780. doi: 10.1016/S1474-6670(17)43916-4
|
| [14] |
PIPES L A. An operational analysis of traffic dynamics[J]. Journal of Applied Physics, 1953, 24(3): 274-281. doi: 10.1063/1.1721265
|
| [15] |
FUKUI M, ISHIBASH Y. Traffic flow in 1D cellular automaton model including cars moving with high speed[J]. Journal of the Physical Society of Japan, 1996, 65(6): 1868-1870. doi: 10.1143/JPSJ.65.1868
|
| [16] |
KRAUSS, WAGNER P, GAWRON C. Metastable states in a microscopic model of traffic flow[J]. Physical Review E, 1997, 55(5): 5597-5602. doi: 10.1103/PhysRevE.55.5597
|
| [17] |
BARLOVIC R, SANTEN L, SCHADSCHNEIDER A, et al. Metastable states in cellular automata for traffic flow[J]. The European Physical Journal B—Condensed Matter and Complex Systems, 1998, 5(3): 793-800. doi: 10.1007/s100510050504
|
| [18] |
WONG S C, WONG W T, LEUNG C M, et al. Group-based optimization of a time-dependent TRANSYT traffic model for area traffic control[J]. Transportation Research Part B: Methodological, 2002, 36(4): 291-312. doi: 10.1016/S0191-2615(01)00004-2
|
| [19] |
PARK B B, SCHNEEBERGER J D. Microscopic simulation model calibration and validation: case study of VISSIM simulation model for a coordinated actuated signal system[J]. Transportation Research Record, 2003(1856): 185-192.
|
| [20] |
KARAGIANNIS G, ALTINTAS O, EKICI E, et al. Vehicular networking: a survey and tutorial on requirements, architectures, challenges, standards and solutions[J]. IEEE Communications Surveys and Tutorials, 2011, 13(4): 584-616. doi: 10.1109/SURV.2011.061411.00019
|
| [21] |
MANIVANNAN P V, RAMAKANTH P. Vision based intelligent vehicle steering control using single camera for automated highway system[J]. Procedia Computer Science, 2018, 133: 839-846. doi: 10.1016/j.procs.2018.07.111
|
| [22] |
VIVO G, DALMASSO P, VERNACCHIA F. The European Integrated Project "SAFESPOT"—How ADAS applications co-operate for the driving safety[C]//IEEE. 2007 IEEE Intelligent Transportation Systems Conference. New York: IEEE, 2007: 534-539.
|
| [23] |
SIKDAR B. Comparison of broadcasting schemes for infrastructure to vehicular communications[J]. IEEE Transactions on Intelligent Transportation Systems, 2012, 13(2): 492-502. doi: 10.1109/TITS.2011.2182193
|
| [24] |
ZHANG Han, CAI Bai-gen, SHANGGUAN Wei, et al. MR-based CVIS scenario design and implementation[J]. System Simulation Technology, 2013, 9(1): 52-60. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-XTFJ201301010.htm
|
| [25] |
GIPPS P G. A behavioural car-following model for computer simulation[J]. Transportation Research Part B: Methodological, 1981, 15(2): 105-111. doi: 10.1016/0191-2615(81)90037-0
|
| [26] |
PETROV P, NASHASHIBI F. Modeling and nonlinear adaptive control for autonomous vehicle overtaking[J]. IEEE Transactions on Intelligent Transportation Systems, 2014, 15(4): 1643-1656. doi: 10.1109/TITS.2014.2303995
|
| [27] |
BUTAKOV V A, IOANNOU P. Personalized driver/vehicle lane change models for ADAS[J]. IEEE Transactions on Vehicular Technology, 2015, 64(10): 4422-4431. doi: 10.1109/TVT.2014.2369522
|
| [28] |
CAI Bai-gen, WANG Cong-cong, SHANGGUAN Wei, et al. Simulation method of information interaction in CVIS[J]. Journal of Traffic and Transportation Engineering, 2014, 14(3): 111-119. (in Chinese) http://transport.chd.edu.cn/article/id/201403015
|
| [29] |
TOUTOUH J, GARCÍA-NIETO J, ALBA E. Intelligent OLSR routing protocol optimization for VANETs[J]. IEEE Transactions on Vehicular Technology, 2012, 61(4): 1884-1894. doi: 10.1109/TVT.2012.2188552
|
| [30] |
KHOKHAR R H, NGADI M A, LATIFF M S, et al. Multi-criteria receiver self-election scheme for optimal packet forwarding in vehicular ad hoc networks[J]. International Journal of Computers Communication and Control, 2014, 7(5): 865. doi: 10.15837/ijccc.2012.5.1342
|
| [31] |
ZHOU Lian-ke, ZUO De-cheng, CUI Gang, et al. A node-traffic characteristics considered clustering broadcast protocol for VANETs[J]. Chinese High Technology Letters, 2012(5): 468-476. (in Chinese)
|
| [32] |
LI Si-hui, CAI Bai-gen, SHANGGUAN Wei, et al. Multi-resolution information exchange method in cooperation vehicle-infrastructure system[J]. Journal of Transportation Systems Engineering and Information Technology, 2014, 14(6): 50-57. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YSXT201406008.htm
|
| [33] |
CHENG S T, HORNG G J, CHOU C L. Using cellular automata to form car society in vehicular ad hoc networks[J]. IEEE Transactions on Intelligent Transportation Systems, 2011, 12(4): 1374-1384. doi: 10.1109/TITS.2011.2158211
|
| [34] |
KALRA N, PADDOCK S M. Driving to safety: how many miles of driving would it take to demonstrate autonomous vehicle reliability?[J]. Transportation Research Part A: Policy and Practice, 2016, 94: 182-193. doi: 10.1016/j.tra.2016.09.010
|
| [35] |
BRIEFS U. Mcity grand opening[J]. Research Review, 2015, 46(3): 1-2.
|
| [36] |
XU Hui-le, ZHANG Yi, LI Li, et al. Cooperative driving at unsignalized intersections using tree search[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(11): 4563-4571. doi: 10.1109/TITS.2019.2940641
|
| [37] |
GUO Qiang-qiang, BAN Xue-gang. Macroscopic fundamental diagram based perimeter control considering dynamic user equilibrium[J]. Transportation Research Part B: Methodological, 2020, 136: 87-109. doi: 10.1016/j.trb.2020.03.004
|
| [38] |
GHIASI A, HUSSAIN O, QIAN Zhen, et al. A mixed traffic capacity analysis and lane management model for connected automated vehicles: a Markov chain method[J]. Transportation Research Part B: Methodological, 2017, 106: 266-292. doi: 10.1016/j.trb.2017.09.022
|
| [39] |
YANG Chao, LOU Wei, LIU Yi, et al. Resource allocation for edge computing-based vehicle platoon on freeway: a contract-optimization approach[J]. IEEE Transactions on Vehicular Technology, 2020, 69(12): 15988-16000. doi: 10.1109/TVT.2020.3039851
|
| [40] |
XU Li-wei, ZHUANG Wei-chao, YIN Guo-dong, et al. Energy-oriented cruising strategy design of vehicle platoon considering communication delay and disturbance[J]. Transportation Research Part C: Emerging Technologies, 2019, 107: 34-53. doi: 10.1016/j.trc.2019.07.019
|
| [41] |
WANG Zhu-wei, GAO Yu, FANG Chao, et al. Optimal control design for connected cruise control with stochastic communication delays[J]. IEEE Transactions on Vehicular Technology, 2020, 69(12): 15357-15369. doi: 10.1109/TVT.2020.3040321
|
| [42] |
ZHONG Zi-jia, LEE J Y. The effectiveness of managed lane strategies for the near-term deployment of cooperative adaptive cruise control[J]. Transportation Research Part A: Policy and Practice, 2019, 129: 257-270. doi: 10.1016/j.tra.2019.08.015
|
| [43] |
GE J I, OROSZ G. Connected cruise control among human- driven vehicles: experiment-based parameter estimation and optimal control design[J]. Transportation Research Part C: Emerging Technologies, 2018, 95: 445-459. doi: 10.1016/j.trc.2018.07.021
|
| [44] |
HAO Liu, KAN Xing-an, SHLADOVER S E, et al. Modeling impacts of cooperative adaptive cruise control on mixed traffic flow in multi-lane freeway facilities[J]. Transportation Research Part C: Emerging Technologies, 2018, 95: 261-279. doi: 10.1016/j.trc.2018.07.027
|
| [45] |
ZHANG Lin-jun. Cooperative adaptive cruise control in mixed traffic with selective use of vehicle-to-vehicle communication[J]. IET Intelligent Transport Systems, 2018, 12(10): 1243-1254. doi: 10.1049/iet-its.2018.5235
|
| [46] |
XIAO Lin, WANG Meng, SCHAKEL W, et al. Unravelling effects of cooperative adaptive cruise control deactivation on traffic flow characteristics at merging bottlenecks[J]. Transportation Research Part C: Emerging Technologies, 2018, 96: 380-397. doi: 10.1016/j.trc.2018.10.008
|
| [47] |
GONG Si-yuan, DU Li-li. Cooperative platoon control for a mixed traffic flow including human drive vehicles and connected and autonomous vehicles[J]. Transportation Research Part B: Methodological, 2018, 116: 25-61. doi: 10.1016/j.trb.2018.07.005
|
| [48] |
LIN Gui-hua, HU Yu, ZOU Yuan-yang. A mixed-mode traffic assignment model with new time-flow impedance function[J]. International Journal of Modern Physics B, 2018, 32(3): 173-185.
|
| [49] |
HUANG Dong-dong, CUI Miao, ZHANG Guang-chi, et al. Trajectory optimization and resource allocation for UAV base stations under in-band backhaul constraint[J]. EURASIP Journal on Wireless Communications and Networking, 2020, 2020: 831-845.
|
| [50] |
CHAI Lin-guo, CAI Bai-gen, SHANGGUAN Wei, et al. A construction approach based on kinematic simulation environment for networked intelligent vehicle[J]. Journal of South China University of Technology (Natural Science Edition), 2018, 46(1): 66-77. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-HNLG201801010.htm
|
| [51] |
CHAI Lin-guo, CAI Bai-gen, SHANGGUAN Wei, et al. Connected and autonomous vehicles coordinating approach at intersection based on space-time slot[J]. Transportmetrica A: Transport Science, 2018, 14(10): 929-951. doi: 10.1080/23249935.2018.1452308
|
| [52] |
CHEN Jun-jie, CAI Bai-gen, SHANGGUAN Wei, et al. Slot control optimization of intelligent platoon for dual-lane two-way overtaking behavior[J]. Journal of Traffic and Transportation Engineering, 2019, 19(2): 178-190. (in Chinese) http://transport.chd.edu.cn/article/id/201902016
|
| [53] |
FENG Yi-heng, YU Chun-hui, XU Shao-bing, et al. An augmented reality environment for connected and automated vehicle testing and evaluation[C]//IEEE. 2018 IEEE Intelligent Vehicles Symposium (IV). New York: IEEE, 2018: 1549-1554.
|
| [54] |
QIU Wei-zhi, SHANGGUAN Wei, CAI Bai-gen, et al. Advance estimate-based traffic state synchronization for parallel testing[C]//IEEE. 2020 IEEE 23rd International Conference on Intelligent Transportation System. New York: IEEE, 2020: 1-6.
|
| [55] |
QIU Wei-zhi, SHANGGUAN Wei, CHAI Lin-guo, et al. Parallel hierarchical control-based efficiency enhancement for large-scale virtual reality traffic simulation[J]. IEEE Intelligent Transportation Systems Magazine, 2021, DOI: 10.1109/MITS.2021.3051473.
|
| [56] |
LI Li, WANG Xiao, WANG Kun-feng, et al. Parallel testing of vehicle intelligence via virtual-real interaction[J]. Science Robotics, 2019, 4(28): 4106. doi: 10.1126/scirobotics.aaw4106
|
| [1] | CUI Jian-xun, YAO Jia, ZHAO Bo-yuan. Review on short-term traffic flow prediction methods based on deep learning[J]. Journal of Traffic and Transportation Engineering, 2024, 24(2): 50-64. doi: 10.19818/j.cnki.1671-1637.2024.02.003 |
| [2] | ZHAO Xiang-mo's team supported by the National Key Research and Development Program of China (2021YFB2501200). Research progress in testing and evaluation technologies for autonomous driving[J]. Journal of Traffic and Transportation Engineering, 2023, 23(6): 10-77. doi: 10.19818/j.cnki.1671-1637.2023.06.002 |
| [3] | GUO Yan-yong, LIU Pei, YUAN Quan, LIU Pan, XU Jin, ZHANG Hui. Review on research of road traffic safety of connected and automated vehicles[J]. Journal of Traffic and Transportation Engineering, 2023, 23(5): 19-38. doi: 10.19818/j.cnki.1671-1637.2023.05.002 |
| [4] | LI Rui, RAN Bin, QU Xu. Traffic capacity enhancement strategy for urban expressway diversion area under vehicle-infrastructure cooperative environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 126-138. doi: 10.19818/j.cnki.1671-1637.2022.03.010 |
| [5] | PEI Hua-xin, YANG Jing-xuan, HU Jian-ming, ZHANG Yi. Distributed cooperative decision-making method for vehicle swarms in large-scale road networks[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 174-183. doi: 10.19818/j.cnki.1671-1637.2022.03.014 |
| [6] | CHEN Ting, YAO Da-chun, GAO Tao, QIU Hui-hui, GUO Chang-xin, LIU Zhan-wen, LI Yong-hui, BIAN Hao-yi. A fused network based on PReNet and YOLOv4 for traffic object detection in rainy environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 225-237. doi: 10.19818/j.cnki.1671-1637.2022.03.018 |
| [7] | WANG Run-min, ZHANG Xin-rui, ZHAO Xiang-mo, WU Xia, FAN Hai-jin. Vehicle-infrastructure cooperative control method of connected and signalized intersection in mixed traffic environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 139-151. doi: 10.19818/j.cnki.1671-1637.2022.03.011 |
| [8] | QU Xu, GAN Rui, AN Bo-cheng, LI Lin-heng, CHEN Zhi-jun, RAN Bin. Prediction of traffic swarm movement situation based on generalized spatio-temporal graph convolution network[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 79-88. doi: 10.19818/j.cnki.1671-1637.2022.03.006 |
| [9] | ZHANG Yi, PEI Hua-xin, YAO Dan-ya. Research review on cooperative decision-making for vehicle swarms in vehicle-infrastructure cooperative environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 1-18. doi: 10.19818/j.cnki.1671-1637.2022.03.001 |
| [10] | QIU Wei-zhi, SHANGGUAN Wei, CHAI Lin-guo, CHU Duan-feng. Multi-scale filtering synchronization method for vehicle-infrastructure cooperative twin-simulation testing[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 199-209. doi: 10.19818/j.cnki.1671-1637.2022.03.016 |
| [11] | SHANGGUAN Wei, ZHA Yuan-yuan, FU Yao, ZHENG Si-fa, CHAI Lin-guo. Vehicle-infrastructure cooperative credible interaction method based on traffic business characteristics understanding[J]. Journal of Traffic and Transportation Engineering, 2022, 22(4): 348-360. doi: 10.19818/j.cnki.1671-1637.2022.04.027 |
| [12] | ZHAO Hang, ZHAO Min, SUN Di-hua, DU Cheng. Mixed traffic group throttling control strategy for traffic bottleneck of expressway[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 162-173. doi: 10.19818/j.cnki.1671-1637.2022.03.013 |
| [13] | SHANGGUAN Wei, PANG Xiao-yu, LI Qiu-yan, CHAI Lin-guo. Guidance and cooperative operation method for group vehicles in vehicle-infrastructure cooperative environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 68-78. doi: 10.19818/j.cnki.1671-1637.2022.03.005 |
| [14] | ZHAO Tong, SHANGGUAN Wei, CHAI Lin-guo, GUO Peng. Scenario factor analysis and test case generation for vehicle-infrastructure cooperative mixed traffic[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 263-276. doi: 10.19818/j.cnki.1671-1637.2022.03.021 |
| [15] | MA Yong-jie, CHENG Shi-sheng, MA Yun-ting, MA Yi-de. Review of convolutional neural network and its application in intelligent transportation system[J]. Journal of Traffic and Transportation Engineering, 2021, 21(4): 48-71. doi: 10.19818/j.cnki.1671-1637.2021.04.003 |
| [16] | CHEN Xiao-bo, CHEN Cheng, CHEN Lei, WEI Zhong-jie, CAI Ying-feng, ZHOU Jun-jie. Interpolation method of traffic volume missing data based on improved low-rank matrix completion[J]. Journal of Traffic and Transportation Engineering, 2019, 19(5): 180-190. doi: 10.19818/j.cnki.1671-1637.2019.05.018 |
| [17] | YANG Lin-jian, ZHAO Xiang-mo, HE Bing-hua, WEI Qiu-yue, AN Yi-sheng. An ant colony optimization algorithm of stochastic user equilibrium traffic assignment problem[J]. Journal of Traffic and Transportation Engineering, 2018, 18(3): 189-198. doi: 10.19818/j.cnki.1671-1637.2018.03.019 |
| [18] | WU Chao-zhong, WU Hao-ran, LYU Neng-chao. Review of control switch and safety of human-computer driving intelligent vehicle[J]. Journal of Traffic and Transportation Engineering, 2018, 18(6): 131-141. doi: 10.19818/j.cnki.1671-1637.2018.06.014 |
| [19] | Zhang Fu-ren. Intelligent selection system of adaptive optimal route based on GIS and grey-fuzzy decision theory[J]. Journal of Traffic and Transportation Engineering, 2007, 7(5): 106-111. |
| [20] | PAN Ming-yang, YAN Fei, XIE Hai-yan. Intelligent traffic simulation based on agent and cellular automata[J]. Journal of Traffic and Transportation Engineering, 2006, 6(2): 70-74. |
| 1. | 牛延强，王燕涛，乔引庄. 智能网联环境下的汽车共享平台设计. 汽车文摘. 2025(01): 47-51 .  | |
| 2. | 王秉路，靳杨，张磊，郑乐，周天飞. 基于多传感器融合的协同感知方法. 雷达学报. 2024(01): 87-96 . | |
| 3. | 刘伟，肖七瑞，陈新海，饶畅，张宇，王博思. 基于时间自动机的无信号交叉口车路协同系统建模与验证. 系统仿真学报. 2024(07): 1682-1698 . | |
| 4. | 张振，王顺，赵腾. 基于车联网的高速公路车路协同应用研究. 汽车知识. 2024(03): 248-250 . | |
| 5. | 杨奕，顾青，孟宇，方华珍，杨俊. 基于强化学习的高优先级车辆通行决策. 无人系统技术. 2024(06): 93-102 . | |
| 6. | 朱克忠. 智能网联环境下城市道路交通协同控制关键技术研究及应用. 电子制作. 2024(24): 32-35 . | |
| 7. | 谢宪毅，赵鑫，金立生，郭柏苍，李克强. 融合深度强化学习与滚动时域优化的智能车辆轨迹跟踪控制. 交通运输工程学报. 2024(06): 259-272 .  本站查看 | |
| 8. | 杜富瑞，潘伟，尹贻辉，李佳梦. 三山岛金矿井下电机车自动驾驶系统设计. 齐鲁工业大学学报. 2023(04): 1-6 . | |
| 9. | 路庆昌，秦汉，刘鹏，崔欣. 基于MFD的快速路新型混合交通流特性分析. 东南大学学报(自然科学版). 2023(05): 905-914 . |
Figures(62) / Tables(8)
Copyright《Journal of Traffic and Transportation Engineering》编辑部陕ICP备05001904号-1
Address ：Editorial Department of Journal of Traffic and Transportation Engineering, Chang 'an University, Middle Section of South Second Ring Road, Xi 'an, Shaanxi（710064） Tel：029-82334388 Email：jygc@chd.edu.cn
All visit:1717833Today's visit:112
Supported by:
Beijing Renhe Information Technology Co. Ltd
SHANGGUAN Wei, LI Xin, CHAI Lin-guo, CAO Yue, CHEN Jing-jing, PANG Hao-jie, RUI Tao. Research review on simulation and test of mixed traffic swarm in vehicle-infrastructure cooperative environment[J]. Journal of Traffic and Transportation Engineering, 2022, 22(3): 19-40. doi: 10.19818/j.cnki.1671-1637.2022.03.002
| 仿真阶段 | 仿真手段 | 仿真对象 | 典型测试方法/架构 | 特征 |
| 萌芽期 | 虚拟仿真 | 交通个体/ 宏观交通流 | 黑盒测试/ 元胞自动机 | 低智化个体化 |
| 起步期 | 视景一体化仿真 | 小规模群体 | 高层体系架构 | 智能化分布式 |
| 发展期 | 虚实交互 | 大规模群体 | 硬件在环/ 实车在环 | 群智化规模化 |
DownLoad:
CSV
| 控制模式 | 场景名称 | 通信 | 分类 |
| 单车控制 | 绿波车速引导 | V2I | 效率 |
| 长直路段车路通讯 | V2I | 服务 | |
| 限速预警 | V2I | 安全 | |
| 多车协同 | 前向碰撞预警 | V2V | 安全 |
| 盲区预警/变道预警 | V2V | 安全 | |
| 协作式变道 | V2V | 效率 |
DownLoad:
CSV
| 类型 | 模型输入 | 反馈形式 | 模型输出 |
| HDV模型 | 车头间距、本车/前车速度 | 非线性 | 加速度 |
| AV模型 | 车头间距、本车/前车速度 | 线性 | 加速度 |
| CV模型 | 队列车辆位置、速度、加速度、时延 | 线性 | 加速度变化量 |
| CAV模型 | 车头间距、前车速度、本车速度/加速度 | 线性 | 加速度/速度 |
DownLoad:
CSV
| 场景功能分类 | 场景功能特征 | 仿真构建需求 |
| 协作式交叉口通行 | 信号控制交叉口车辆常规通行 | 安全/效率 |
| 信号控制交叉口处理异常停车 | 安全/效率 | |
| 交叉口信号自适应控制 | 效率 | |
| 交叉口动态车道管理 | 效率/管理 | |
| 无信号控制交叉口组织通信异构车辆常规通行 | 安全/效率 | |
| 协作式城市快速路匝道控制 | 城市快速路单匝道控制 | 安全/效率 |
| 城市快速路主线多匝道协同控制 | 安全/效率 | |
| 城市快速路与高速公路结合部控制 | 安全/效率 | |
| 协作式城市路段控制 | 城市常规路段组织车辆编队 | 安全/效率 |
| 常规路段组织紧急车辆优先通行 | 安全/效率 |
DownLoad:
CSV
|
|
DownLoad:
CSV
|
|
DownLoad:
CSV
|
|
DownLoad:
CSV
|
|
DownLoad:
CSV
| 仿真阶段 | 仿真手段 | 仿真对象 | 典型测试方法/架构 | 特征 |
| 萌芽期 | 虚拟仿真 | 交通个体/ 宏观交通流 | 黑盒测试/ 元胞自动机 | 低智化个体化 |
| 起步期 | 视景一体化仿真 | 小规模群体 | 高层体系架构 | 智能化分布式 |
| 发展期 | 虚实交互 | 大规模群体 | 硬件在环/ 实车在环 | 群智化规模化 |
| 控制模式 | 场景名称 | 通信 | 分类 |
| 单车控制 | 绿波车速引导 | V2I | 效率 |
| 长直路段车路通讯 | V2I | 服务 | |
| 限速预警 | V2I | 安全 | |
| 多车协同 | 前向碰撞预警 | V2V | 安全 |
| 盲区预警/变道预警 | V2V | 安全 | |
| 协作式变道 | V2V | 效率 |
| 类型 | 模型输入 | 反馈形式 | 模型输出 |
| HDV模型 | 车头间距、本车/前车速度 | 非线性 | 加速度 |
| AV模型 | 车头间距、本车/前车速度 | 线性 | 加速度 |
| CV模型 | 队列车辆位置、速度、加速度、时延 | 线性 | 加速度变化量 |
| CAV模型 | 车头间距、前车速度、本车速度/加速度 | 线性 | 加速度/速度 |
| 场景功能分类 | 场景功能特征 | 仿真构建需求 |
| 协作式交叉口通行 | 信号控制交叉口车辆常规通行 | 安全/效率 |
| 信号控制交叉口处理异常停车 | 安全/效率 | |
| 交叉口信号自适应控制 | 效率 | |
| 交叉口动态车道管理 | 效率/管理 | |
| 无信号控制交叉口组织通信异构车辆常规通行 | 安全/效率 | |
| 协作式城市快速路匝道控制 | 城市快速路单匝道控制 | 安全/效率 |
| 城市快速路主线多匝道协同控制 | 安全/效率 | |
| 城市快速路与高速公路结合部控制 | 安全/效率 | |
| 协作式城市路段控制 | 城市常规路段组织车辆编队 | 安全/效率 |
| 常规路段组织紧急车辆优先通行 | 安全/效率 |
|
|
|
|
|
|
|
|
DownLoad:
DownLoad:
