Connected vehicles (CVs) can exchange a variety of information including speed, acceleration, direction, and location instantly with nearby CVs and connected infrastructures. With real-time information, the forthcoming driving behavior could be predicted. Based on that, computers or human drivers could make critical driving decisions to minimize collision damages or avoid potential collisions [1
]. CVs, with dual roles of exchanging information and decision-making, could improve the road capacity by enhancing the running speed, and thus contribute to a safer, more efficient, and comfortable driving experience [2
Currently, studies on CVs mainly focus on three key aspects, environment perception, behavior decision, and motion control [3
Environmental perception technology mainly relies on sensors. As regard to the latest technologies, laser radar can receive 60% to 75% of the three-dimensional information of vehicles’ surrounding environment [4
], followed by camera’s visual information, for example CCD (Charge Coupled Device) camera millimeter wave radar’s direction and distance information, GPS’s position information, ultrasonic, infrared, and other photoelectric sensors’ information. Behavior decision systems, through which road and vehicle information can be obtained, help autonomous vehicles (AVs) drive more safely and reasonably. Motion control system is also used to control a vehicle itself [5
]. Kang and Wang et al. proposed the limitations of current simulation methods: There is no prediction of complex relations between vehicles, driving environments, and routes [6
]. They used a simulation model to build a series of methods on coordination, positioning, route planning, and evaluation for autonomous vehicles (AVs). Chen and Gong et al. pointed out that, in the complex and changeable dynamic environment, the acquisition of driving rules and decisions is still the key point restricting the development of AVs [7
]. “Rough set theory” is an important method to resolve the issue. The motion control of AVs, including steering control, vertical and horizontal control, path tracking control, etc., is often inseparable from environmental perception [8
]. Sense of environment allows vehicles to be controlled with fewer errors. Scholars often consider motion control along with sensors. Shan and Li et al. proposed an AV control strategy for path tracking, including curve fitting, curve discretization, and tracking controller design [9
]. The Bessel curve fitting method was used to fit the generated results to create paths smoothly. Emmanuel proposed a fuzzy logic control strategy for AVs that combines GIS positioning, image processing, and image recognition [10
]. Chen and He et al. proposed an innovative, hybrid path balancing model and established a virtual network to replace the original AV network. Each virtual link represents a group of paths connecting the entrance and the exit of the AV zone [11
As can be seen, most of these studies did not focus on the requirements of the integrated traffic parameters. At present, the traffic index system of conventional road design is still adopted without considering the appearance of AVs or CVs. Taking the particularity of a vehicle network into account, this study intends to carry out a quantitative research on the integrated key traffic parameters, and put forward optimized running speed by balancing multiple objectives such as road efficiency, safety, and economy.
Moreover, scholars have made sufficient researches and discussions on the impact of AV on the traffic congestion, traffic safety, traffic flow theory, objective detection, and legal policies. Chen et al. explored the management of a fleet of shared autonomous electric vehicles (SAEVs) using a regional, discrete-time, and agent-based model [12
]. The simulation examined the operation of SAEVs under various vehicle ranges and developed different scenarios in a city grid to analyze the impact of connected and autonomous vehicles (CAVs) on the infrastructure construction decisions. Fagnant et al. focused on the impact of CAVs on the traffic and environment. They established a city model that was divided into several grid regions, each of which were generated at a certain time interval. The variables such as the service area size, network congestion level, relocation strategy, and fleet size were analyzed successively. Thus, the impacts of CAVs on the environmental factors, such as urban travel efficiency, pollution emissions, and energy efficiency, were quantified [13
]. The strategy designed in the study for exchanging CVs’ information is advisable: Vehicles share information, such as positions and speed, to adjust or reposition their own speed or acceleration according to the driving state of surrounding vehicles, so as to improve the traffic efficiency and optimize the overall running speed.
In regards to the traffic or road design parameters under the CAV environment, Chen et al. estimated the effects of autonomous trucks’ lateral distribution within the lane with respect to the rutting depth and fatigue damage by finite element analysis under certain environmental conditions [14
]. Xie et al. combined VISSIM with the Car2X module (a built-in script in VISSIM that enables vehicles to exchange information) to accurately capture the acceleration, speed, and position of all vehicles in a certain area. The information was entered into an optimization model based on MATLAB, which then automatically decided the optimal control strategy for each vehicle and returned the simulation results [15
]. In this paper, researches on running speed under the environment of instant traffic network were conducted, which requires a similar evaluation platform to compare and analyze the road traffic conditions at different running speeds. Based on the simulation results, the optimal traffic management strategy could be selected. Ntousakis et al. designed an analytical trajectory planning method to assist in the merging of vehicles on highways. It solved the problem of passengers’ discomfort by minimizing the acceleration [16
]. In addition, an alternative solution to the time-varying linear quadratic regulator method was introduced, which applied the model predictive control (MPC) scheme to compensate for the possible interference in the trajectory of nearby vehicles. The results of the instant updated measurements were repeatedly applied to analyze the optimal solution until the merging process was finally completed [17
]. This method was widely used in the area of the mixture of CVs and non-CVs. Ye and Yamamoto developed a heterogeneous traffic flow model to study the impact of connected vehicles on traffic flow [18
]. Based on the recently proposed two-state safe-speed model, a two-lane cellular automaton model was developed, in which CAVs and conventional vehicles were incorporated into heterogeneous traffic flows [19
]. To sum up, the relevant models of CAV have been explored, but there is little discussion on the intergraded traffic parameters.
Scholars have explored various micro-simulation models for AVs. However, it is difficult to verify the reliability of these models. In addition, the simulation methods for CVs still require further analyses and verifications. VISSIM is an effective platform for simulation and secondary development. It is mainly utilized for studying on-ramp path control, intersection signal optimization, and CAVs’ cruise control. Therefore, in this paper the traffic parameters in the context of CVs were defined and imported into VISSIM to conduct the simulation.
In summary, scholars have analyzed the impact of CAV driving technology in various fields on traffic flow. However, studies on the impact of CV technology on the basic and comprehensive traffic parameters are still rare. In an attempt to gain insight into the impact of vehicle networking on vehicles, this research simulated the environment of vehicle networking by applying the speed optimization model to the road model established in VISSIM. Besides, the influence of different CV rates on the results and the optimal CV marketing rate were also analyzed.
In this paper VISSIM was utilized to build the road model for CVs simulation. The results of this study were based on simulation, but they provide a new orientation for studies of CVs: Statistical analysis before implementing CV experiments in real roads, which is more economical and can provide estimations for road design and traffic analysis, etc.
Results quantitatively showed how vehicle networking affects CVs’ running speed, travel time, and delay time. Although the results of this study depended on different kinds of software, they provide a reference for the study of safety and efficiency of CVs. When the model was verified, it was found that the running speed tends to be stable after several iterations. This is consistent with Xie’s findings [15
Although Fmincon solves nonlinear constraints, the objective function should be continuous. The function decided step size according to the given initial value, and could only obtain the local optimal solution. As was mentioned by Voelz, if the initial value given was an infeasible solution, it was difficult to converge to the optimal solution, and the final result might exceed the upper and lower limits in order to reach the target function value [24
]. It must be pointed out that the average speed optimization model in this research was not suitable for the crowded traffic flow. The traffic flow of the simulation model designed in this paper was basically unobstructed, with only little queuing of vehicles. Some concerns need to be clarified:
Most vehicles’ speed in the crowded traffic flow tended to be zero, which brought negative effects on the accuracy of speed analysis.
The vehicle distance was limited to 10 m in this model, but it was not true when the traffic was blocked. As a result, it is not conductive to analyzing queue length and vehicles’ passing time.
Under these concerns, the crowded traffic situation was not considered in this study. It should be noted that, till now, CAVs are mostly experimented in highways or areas with a good traffic condition, where the results of this study are applicable. As for the traffic flow in the crowded state, other speed optimization models need to be considered, such as ant colony optimization [25
], genetic algorithm [26
], and so on.
Based on the comparative analysis of running speed in the normal and the connected states, it can be concluded that the average running speed of the vehicle in the connoted state could increase by about 4 km/h compared to that in the normal state. Considering the influence of various other factors on the real road and the hardware and software limitations in the simulation process, running speed in the connected state is proposed: In the case of fully-connected driving conditions, the running speed of urban arterial roads with bidirectional diversions with a maximum speed limit of 70 km/h can increase by 4 km/h. However, other factors still need to be considered to analyze the running speed under actual vehicle networking conditions, such as the safety when the vehicle turns at a corner. A comprehensive study of the minimum radius of the road is required. Recently, there are many kinds of automatic control algorithms for CAVs, and errors in simulations are inevitable. Therefore, these issues need to be considered for the precise adjustment of running speed comprehensively. There was only one road model in this study, but this kind of road is very common in China. It is hoped that more samples can be supplemented in future studies. The road model was conducted according to Chinese roads, but the internal design parameters in VISSIM, such as the lane change and lateral behavior, were modified through a separate experiment through the VISSIM’s V2V (vehicle to vehicle) sample to ensure the credibility of the results. The driver responses of CVs were ignored, as it is assumed that the vehicle was under a totally connected state without human control. Above all, the method mentioned in this paper provides a simulation-based adjustment scheme for running speed, which can be used for preliminary estimation and reference before AV experiments.