Abstract
Evaluating the economic benefits of traffic optimization from connected and autonomous vehicles (CAVs) and relevant traffic organization methods is significant, which will help to put forward suggestions for policymakers to promote the application of CAVs. The impacts and related benefits from CAVs with level 2 automation (L2 CAVs) on traffic efficiency and energy consumption of expressways are analyzed in this paper. Average travel time and actual road capacity are on behalf of traffic efficiency while average electric energy consumption is used to compute traffic energy consumption. The corresponding traffic economic benefits consist of travel-time-saving benefits, road construction benefits, and energy-saving benefits. A benefit evaluation framework is newly proposed and microscopic traffic simulation software is applied as the experiment platform. Different market penetration rates of L2 CAVs and various traffic flow statuses are considered. Besides, dedicated lanes for CAVs are also involved in this research, which are regarded as a traffic organization method expected to promote the realization of CAV’s traffic benefits. It is found that L2 CAVs can save the average travel time and reduce average energy consumption for a single vehicle in most scenes. However, negative impacts on energy consumption are observed in several scenes due to the increase of actual road capacity. Positive economic benefits are obtained as soon as the traffic flow rate is out of saturation, which become increasingly higher as CAV’s market penetration rate turns larger. Additionally, amplification in traffic economic benefits appears only if CAV lanes are provided under proper conditions.
1. Introduction
Driven by the new generation of scientific and technological revolution, connected and autonomous vehicles (CAVs) have become a new developing trend of vehicle products [1]. CAVs are developed to assist or even replace human drivers to complete driving tasks and make travels safer, more efficient, and more comfortable than those by traditional human-driving vehicles (HVs) [2]. CAVs are not only equipped with different kinds of intelligent parts such as sensors, computing platforms, and wire control actuators but also “Vehicle to X (V2X)” communication modules enabling them to communicate with other vehicles (V2V), transportation infrastructures (V2I), networks (V2N), and pedestrians (V2P) [3, 4]. Therefore, they could realize autonomous driving through self-intelligence or V2X communication. China’s traffic efficiency needs to be improved urgently. The economic loss caused by traffic congestion accounts for 20% of the disposable income of the urban population, which is equivalent to 5–8% of China’s gross domestic product [5]. Traffic energy consumption and emissions account for a large proportion of China’s total energy consumption and emissions. In 2020, China’s traffic carbon emissions reached 930 million tons, making up 15% of the national terminal carbon emissions while road traffic carbon emissions account for 90% of them [6]. CAVs can keep a smaller time headway and a shorter reaction time than human drivers and access real-time traffic information such as congestion, weather, and travel services. As a result, it is expected to improve traffic efficiency and reduce traffic energy consumption.
At present, some scholars have conducted researches on traffic impacts from CAVs with a combination of different road types, traffic conditions, functions, and market penetration rate (simplified as “MPR”) of CAVs. Narayana et al. made a systematic review on intelligent vehicle microsimulation studies [7]. Nearly thirty studies were included, which are mainly about the emission impacts [8–11], traffic flow impacts [10–14], and safety impacts [15–19] of CAVs or autonomous vehicles (AVs). Do et al. also gave a literature review on simulation research studies about the adaptive cruise control (ACC) and the cooperative adaptive cruise control (CACC) systems [20]. Stogios et al. conducted a research on the impacts of AVs on traffic states and energy consumption of urban closed highway and open road networks through microscopic traffic simulation. It was announced that the impacts are not always positive, which depends on the traffic volume and driving styles [21]. Song et al. also took driving styles into account in their research and believed that the single-lane automatic driving system should be developed to be more aggressive to improve traffic efficiency and reduce carbon dioxide emissions [2]. Ma et al. analyzed the traffic velocity influences of AVs with different automatic levels on urban road networks by microscopic and macroscopic simulation. It was indicated that AVs with higher automatic levels are able to make traffic velocity maintain stability. But different automatic levels of AVs were only distinguished by several kinematic parameters, which do not meet the criteria for autonomous driving classification [22]. He et al. claimed according to their microscopic simulation on freeway networks that CAVs could not increase the actual road capacity and exert obvious impacts on traffic velocity when their MPR is lower than 10% [23]. Liu et al. claimed that the time headways of CAVs are generally shorter than those of human driving vehicles. They also found that CAVs could raise the actual road capacity by more than one time as their MPR reaching 100% but reduce it when their MPR is lower than 30% [24]. Davis confirmed that when the MPR of the ACC system reaches 50%, traffic volume could be increased by 20% at merging areas of expressways [25]. Arnaout et al. made a research on the influences of CACC on traffic flow characteristics of circle highways through microscopic simulation. It was found that the CACC system could reduce travel time, and the same effects of optimization on actual road capacity were observed with its MPR equal to 60%, 80%, and 100%, respectively [26]. There are also some research studies focusing on the impacts of exclusive lanes for CAVs or AVs. Zhang et al. believed that setting dedicated lanes for self-driving buses could significantly reduce pollutant emissions and energy consumption [27]. Carrone et al. claimed that the economic benefits of AV lanes mainly derive from AV’s ability to save travel time and improve throughput. The benefits will be more than double when the MPR of AVs increases from 90% to 100% [28]. Yu et al. declared that roadway capacity is improved by 84% by providing AV lanes [29]. However, these relevant research studies still have some disadvantages. First, there are few studies modeling CAVs with different automatic levels, but most of them pay attention to only one function of CAVs. Second, the road scenes modeled in these studies are generally idealistic without basis and the setting of CAV MPR is not systematic enough. Therefore, the authenticity and confidence of them remains to be improved. Third, these research studies mostly concentrate on traffic impact analysis and few of them evaluate from the perspective of economic benefits. In result, it is difficult for governments to carry out policies to promote CAV development and implementation based on these studies due to the lack of the practical reference value.
Vehicle automatic capability is divided into five levels (from level 1 to level 5, simplified as “L1 to L5”) by SAE [30]. The capability gradually increases with the level changing from 1 to 5 in order. At present, because L3 and L4 autonomous driving are limited by the problem of accident responsibility division and technical maturity, the capability of mass production passenger vehicle is still at L2. Long-term stay and continuous maturity of technology make the MPR of vehicles with L2 automation in China improve continuously. It has reached 30% in the first half of 2022 and is expected to approach 50% in 2025. Nevertheless, the installation rate of V2X communication modules is not as high as expectation [4]. Autonomous driving functions currently developed are almost based on expensive intelligent parts rather than Internet of vehicles, showing disadvantages in perception accuracy, scope, and development costs. As a result, this paper focuses on L2 CAVs and proposes an evaluation framework to study their traffic impacts on urban expressways with the help of microscopic traffic simulation in VISSIM. Not only are the traffic efficiency and energy consumption analyzed but also the related economic benefits are evaluated. Besides, it is expected to intuitively highlight the economic and social value of L2 CAVs and provide government with specific suggestions on the way to promote L2 CAV’s application.
In terms of the road scenes, first, the urban expressway provides fast and efficient traffic services for long-distance motor vehicle travel in the city. Although it has the lowest distance proportion among urban roads, it services a large amount of traffic volume, which is usually oversaturated during the peak periods and leads to congestion. Second, due to the relatively simpler scenes compared with other kinds of urban roads, it is more suitable for L2 CAVs to drive on. Functions of L2 CAVs are also mainly designed for closed roads such as highways and urban expressways [4]. Therefore, it is highly targeted to select the peak hours of urban expressways as the research road scene, which is expected to be significantly affected by L2 CAVs. Specifically, the expressways in Beijing during rush hours are chosen to be the road scene, which is treated as a case study. As an international metropolis and a super large city in China, Beijing is struggling against serious traffic congestion. The peak delay index of Beijing’s road network in 2019 is 1.86 [31], which means that residents have to spend 1.86 times as much as the time in flat hump periods to reach the destination in peak periods. In 2017, the annual congestion cost per capita in Beijing even reached 4013 yuan [32]. Beijing has regarded CAVs as one of the most significant way to improve traffic efficiency and established a demonstration area for CAVs to promote their development. In addition, it has been demonstrated that increasing the probability of CAV-CAV following state could avoid CAVs degenerate into AVs or human-driving vehicles (HVs), which could promote the realization of traffic benefits of CAVs [33]. Providing dedicated lanes, as a frequently-used traffic organization method, is able to separate designated kinds of vehicles from the others. So, it is expected that setting up CAV dedicated lanes (simplified as CAV lanes) under suitable situations could probably increase the probability of CAV-CAV following state to maximize CAV’s traffic benefits. Therefore, the impacts of CAV lanes are also analyzed.
The remainder of this paper is organized as follow. Chapter 2 first constructs the evaluation framework of CAV’s traffic benefits as the methodology and then describes the used models and methods including the vehicle models, energy consumption model, and traffic economic benefits model. Chapter 3 introduces the simulation settings and the results. Chapter 4 comes up with policy suggestions based on the results and makes a conclusion.
2. Methodology
2.1. Evaluation Framework of CAV’s Traffic Benefits
The evaluation framework of CAV’s traffic benefits is depicted in Figure 1. It mainly includes the evaluation model of impacts on traffic efficiency and traffic energy consumption and the benefit coupling model. As for the calculation process, first, L2 CAVs and their exclusive lanes are regarded as research objects to be input into the framework. Through the impact evaluation model, traffic efficiency and traffic energy consumption per vehicle per kilometer can be obtained. Second, the travel characteristics and road network structure information are input into the framework. Meanwhile, the variations of unit travel time, unit energy consumption, and actual road capacity are input to the framework. With help of the economic benefit coupling model, the total economic benefits for expressways from L2 CAVs and their dedicated lanes could be worked out.
As mentioned above, the framework is composed of two models. The first one is the impact evaluation model of traffic efficiency and traffic energy consumption. The road section geometry and the layout, different traffic flow rates, vehicle power types and basic characteristics, and different vehicle fleet composition should be set in advance to construct simulation environments. The second one is the benefit coupling model, in which the output of the first model is coupled with benefit calculation parameters, such as travel characteristics, road network structures, vehicle class distribution, and electricity prices.
2.2. Vehicle Models
The principal driving status for vehicles consists of longitudinal driving and lateral driving. Longitudinal driving encompasses free-driving, vehicle-following, and emergency-braking statuses while lateral driving is composed of lane-changing and turning statuses. Due to the differences between CAVs and HVs in the perception range, reaction time, and control logic, the driving behaviors of them at longitudinal and lateral statuses are quite different. Both the two statuses could be described by models with kinematic parameters. Consequently, the selection of scientific models and their calibration parameters that reflect the driving behavior variations between CAVs and HVs is of great importance to this research.
CAVs and HVs have nearly identical control logic at the free-driving status. They will accelerate to the desired velocity at their desired accelerations if their current velocities are lower than the desired velocity, and they will decelerate at the safe decelerations if their current velocities are higher than the desired velocity.
The driving behavior of vehicles in the vehicle-following or emergency-braking status depends on the conduct of their preceding vehicles. Thanks to the accurate and real-time perception information of their predecessors obtained from onboard sensors and vehicle-to-vehicle (V2V) communication, CAVs are able to output precise acceleration in each time step as vehicle control command. Different perception methods have different detection ranges. Considering the current representative mass-produced passenger vehicles, the front detection distance of autonomous vehicles is set to just 200 meters with the help of onboard sensor [34]. This kind of autonomous vehicles is simplified as AVs as mentioned before. V2V communication can offer longer perception distance than vehicle sensors. Assuming that all CAVs in this paper are equipped with LTE-V2V communication modules, their front detection range is determined as 500 meters [34]. L2 CAVs or AVs in this paper are equipped with longitudinal driving functions of connected adaptive cruise control (adaptive cruise control for AVs) and autonomous emergency braking. So, they are able to assist or even replace human drivers in those two statuses. The model “MIXIC” developed for ACC or CACC is used to model CAVs [35, 36]. The algorithm of the MIXIC model could be found in supplementary information, and its parameter values are listed in Table S1.
The longitudinal control strategy of CAVs is specifically shown in Figure 2. If the piolet vehicle of a CAV is not a CAV, it will degenerate to an AV. [m] is the front perception distance of the subject vehicle depending on the types of its preceding vehicle. [m] represents the current space headway between the subject vehicle and its predecessor. [m/s] and [m/s] are on behalf of the current velocity of the subject vehicle and the desired velocity. is usually set to the road velocity limit. [m/s2] is the acceleration output by the MIXIC model. [m/s2] is the eventual control demand of the subject vehicle.
The Wiedemann 99 model is used to simulate HVs’ vehicle-following and emergency-braking behaviors [37]. It is a psychophysiological driving model established to simulate the driving behaviors of human drivers with various driving styles such as aggressive, moderate, and cautious style [38]. The details of the Wiedemann 99 model can be found in supplementary information and its parameter configuration is shown in Table S2. Other parameters defining the vehicle longitudinal behavior of human drivers in VISSIM can also be found in supplementary information.
Lane change is one of the most common and complex lateral driving behaviors. In this paper, the rule-based lane-changing model incorporated in VISSIM is selected because of its ability to emulate the mandatory and free lane-changing behavior. In order to accurately analyze the impacts of L2 CAVs on longitudinal driving behavior and traffic flow characteristics, as exhibited in Table S3 in supplementary information, the model parameter configuration of CAVs and HVs are consistent, which is optimized based on previous research studies to ensure safety [2].
2.3. Energy Consumption Model
Electrification has become one of the main development directions of the automobile industry and is expected to become the vehicles’ leading power type in the future [39]. The MPR of China’s new energy vehicles (NEVs) has climbed from 7% in 2020 to 26.7% in July 2022, with battery electric vehicles (BEVs) accounting for 81%. Its growth rate has exceeded the predicted value given by the China society of automotive engineers [4]. BEVs have been proved to consume lower energy than traditional fuel vehicles and discharge nothing during their use stages. Under the background of China’s “double carbon target” and with the gradual maturity of BEV’s key technologies, BEV’s costs are expected to decrease so that its MPR is great likely to increase faster [5]. Therefore, it is assumed that all vehicles in simulation are BEVs.
The Virginia Tech Comprehensive Power-based EV Energy consumption model (VT-CPEM) is selected to estimate vehicles’ electric energy consumption [40]. The VT-CPEM is able to get easily implemented to microscopic traffic simulation because the instantaneous electric energy consumption of a vehicle can be worked out by its current velocity and acceleration. In addition, instantaneous braking energy regeneration of BEVs can also be considered in this model. What is more, compared with other regression models using instantaneous velocity and acceleration as input variables, the VT-CPEM applies vehicle dynamic models, contributing to its accuracy and flexible regional universality. The average electric energy consumption of a vehicle in one kilometer ( can be computed by dividing the total electric energy consumption ( [kWh]) by the total mileage [km] of all vehicles, as shown in the following equation: . is the integral of the sum of all vehicle instantaneous electric power to time. The way to compute is shown in supplementary information, and the parameter set of the VT-CPEM is shown in Table S4.
2.4. Economic Benefit Model
The total economic benefits ( [billion Yuan]) from L2 CAVs analyzed in this paper are made up of the economic benefits of traffic efficiency ( [billion Yuan]) and energy consumption ( [billion Yuan]). TEB consists of two parts, which are the economic benefits from saving travel time ( [billion Yuan]) and road construction costs ( [billion Yuan]). , which is caused by the changes in travel time and actual road capacity, is computed by equation (1).
[person/day] is the average volume of people traveling by car in downtown Beijing on weekdays [31]. represents the volume proportion of people traveling by private car and taxi in peak periods to the total volume of people traveling in Beijing on weekdays [31]. is on behalf of the actual traffic flow rate. is the actual traffic flow rate when the MPR of CAVs is equal to 0. is a selection function. If the independent variable is not smaller than 0, the function value will return 0. Otherwise, it will return 1. As for the subscripts, is the label of scenes with different MPRs of L2 CAVs (varying from 0 to 100% with the gradient of 25%) and special lane management schemes. represents the label of different traffic flow rates. Different combinations of and build a matrix of research cases, as shown in Table 1.
[km] represents the average commuting distance during rush hours on weekdays in Beijing [41]. is on behalf of the proportion of use intensity of urban expressways to that of all kinds of urban roads in Beijing during peak periods on weekdays. Use intensity means the total mileage derived from all vehicles in unit time. [yuan/(s·person)] represents the average personal time value of daily travel in Beijing [42]. [day] is on behalf of the number of weekdays in a year. [s/(km)] is the variation in average travel time per kilometer under various CAV MPRs and traffic flow rates.
Among these model parameters, could not be obtained directly since there is no relevant public statistics. So, it is necessary to resort to the method of traffic allocation, where the origin and destination pairs (ODs) are of great importance. ODs generally belong to governments or mobile communication operators. They are strictly limited to being disclosed as they are involved in national geographic information security, personal information security, and business secrets. To this end, the top 20 metro stations with the biggest volume of incoming passengers and the top 20 metro stations with the largest quantity of outcoming passengers during peak periods in Beijing are regarded as the travel origins and the destinations, respectively [31]. Consequently, 400 ODs are constructed. It is reasonable to apply them to estimate because no matter what kind of transportation modes people choose, there is always a large volume of departure and arrival demands around those metro stations. In addition, the distribution of these stations is highly consistent with that of work and residence areas in Beijing. Next, with the help of the “Gaode” map platform, routes between these OD pairs are planned under the driving mode expecting to spend the shortest time. If there are several routes with the same travel time, then the one with the shortest distance will be chosen. The route suggestions were queried during the peak period of a workday, which is 17 : 30–18 : 00 on October 10th (on Tuesday), 2022. Expressway mileage of a plan is screened from every path instruction and the use intensity of expressways of the plan is then given by dividing the expressway mileage by the total mileage. is eventually worked out as the average value of the arithmetic mean value of the expressway mileage ratio from each origin, which is weighted by the number of passengers entering each related original station.
[billion Yuan] can be worked out by equation (2). It is assumed that the improvement in actual road capacity is equivalent to reducing the width and even the number of lanes of expressways, thereby saving construction costs. [billion yuan/km] represents the average material costs per kilometer of urban expressways. They are nearly 80% of the total construction costs [43]. The costs of pipeline migration, vegetation transplantations, and compensation for demolition are excluded. is the saturated traffic flow rate when the MPR of CAVs is equal to 0. is utilized instead of the designed traffic flow rate in order to avoid exaggeration of . The calculation of refers to the construction costs of Beijing’s fifth ring road and sixth ring road and the proportion of material costs [43, 44]. [km] represents twice the total mileage of the urban expressways in Beijing. is on behalf of the number of years of expressways’ life.
As shown in equation (3), is composed by two parts: the changes of electric energy consumption per vehicle and the total alteration from variation of actual road capacity, which respectively effect the traffic energy consumption directly and indirectly.
[vehicle/time] is the desired maximum number of private vehicles traveling, which is equal to the number of private vehicles in Beijing. [time/vehicle] represents the number of traveling times of private vehicles during rush hours in Beijing [31]. [Yuan/kWh] is on behalf of Beijing’s average charging price of electric vehicles. [billion Yuan] and [billion Yuan] are, respectively, the variation and the absolute value of electric energy consumption per kilometer of a vehicle under different CAV MPRs and traffic flow rates. Parameters of equations (1) to (3) can be found in Table 2.
3. Simulations and Results
3.1. Simulation Experiments
Traffic simulation is extensively utilized in pertinent research studies. It shows better performance than the field test in flexibility, costs, and workload. In addition, field tests of CAVs and related traffic organization methods with large scale are usually not permitted because of legal restrictions and damage to normal traffic conditions. There are three kinds of traffic simulation methods totally, which are microscopic, mesoscopic, and macroscopic traffic simulation [38]. Microscopic traffic simulation is selected, through which the influence of intelligent functions for vehicle dynamics is able to be precisely described. So, CAVs with varying intelligence levels can be well differentiated. All simulation tests are conducted on a microscopic traffic simulation platform called VISSIM, which has been the most widely used microscopic traffic simulation platform since 2005 [7]. In VISSIM, both the lane-level road characteristics and the dynamic control parameters of vehicles can be accurately modeled, and various types of roads and vehicles can be combined flexibly to emulate complicated traffic scenes systematically [7].
The expressway in this research is modeled as a one-way road section with the length of 4 kilometers to ensure that the vehicles entering the road section could reach a stable condition. Based on Chinese standard for urban road construction, there are 3 lanes with the width of 3.75 meters on the road section and the speed limit is set to 100 km/h [45, 46]. Ramps are not taken into account in order to avoid traffic congestion derived from external factors such as merging and diversion. Therefore, it is expected to mine the maximum potential of the traffic influence of L2 CAVs on urban expressways. The traffic conditions are determined by three states of input traffic flow rates (simplified as “Qi”): unsaturated, saturated, and oversaturated states. And there are five conditions considered in all, which are 80%, 100%, 120%, 140%, and 160% of the saturated traffic flow rate (simplified as “Qs”). There are at most two kinds of vehicles driving on the road, which are L2 CAVs and HVs. The ratio of aggressive driving behavior to normal driving behavior to cautious driving behavior is set to 24.5%:19.7%:55.8% [47, 48]. CAV dedicated lanes are also analyzed as a special traffic organization. The dedicated lane is set on the rightmost lane and the rightmost two lanes services only for CAVs if there are two dedicated lanes to be set. CAVs give priority to driving in the exclusive lane but can also drive in the ordinary lanes. They can enter the dedicated lanes at any time on the premise of ensuring safety. However, HVs are not allowed to drive in them. Different simulation scenes are discriminated by the CAV MPR and the number of CAV lanes. Different scenes and the input traffic flow rate constitute different simulation cases, the detail of which could be found in Table 1. Simulation test for each study case continues for 1800 seconds to guarantee the full input of the traffic flow. Simulation resolution, random seed, random seed increment, and dynamic assignment volume increment are set to the default values, which are 10 time steps per simulation second, 42, 1, and 0, respectively. The maximum simulation speed is selected. Because the dynamic link library has to be accessed to obtain the decision logic of CAVs, only one core of the computer is allowed for using. Each simulation test is repeated for 5 times to avoid and eliminate the effect of simulation randomness.
3.2. Results
3.2.1. Impacts on Travel Time
The average travel time per kilometer in different scenes under different traffic conditions is depicted in Figure 3. Under the free flow condition without traffic conflict and external disturbance, as the traffic flow rate increases from the unsaturated state to the saturated state and finally reaches the oversaturated state, the average travel time per kilometer rises at first and then keeps stable in general. Without CAV lanes, the average travel time per kilometer under the same traffic flow rate decreases with the increase of the MPR, which can be reduced by 3.2% (scene B, Q = 120% Qs), 5.2% (scene D, Q = 120% Qs), 8.9% (scene G, Q = Qs), and 11.7% (scene J, Q = Qs) compared to the base scene (Scene A).
The value of the average travel time per kilometer of the base scene and the variation proportion of the average travel time per kilometer from the base scene are listed in Table 3. As for scene C and scene F at the state of the unsaturated traffic flow rate, the inappropriate number of CAV lanes takes up lots of road resources but receives a low utilization rate, contributing to increase in average travel time and negative impacts on traffic efficiency. However, the average travel time per kilometer in these two scenes keep stability at a high level under the conditions of saturated and oversaturated traffic flow rates. This is because the traffic condition within dedicated lanes could be optimized and performs as the free flow state though the whole traffic flow rate has been saturated or even oversaturated. In other words, the decrease of the overall average travel time of vehicles is at the expense of the increase of that of vehicles outside the exclusive lane. Nevertheless, the proper number of CAV lanes under certain scenes, especially after MPR exceeds 50%, is able to force CAVs to form a queue so that the reduction of travel time could be further amplified with lighter disturbance to and from vehicles outside the dedicated lanes. Like scene E, scene H, and scene I, more reduction is observed compared to the scenes with the same MPR but without the CAV lane such as scene D and scene G, respectively.
3.2.2. Impacts on Actual Road Capacity
Actual road capacity in different scenes under different traffic conditions is depicted in Figure 4. It is characterized by the quotient of the real and input traffic flow rate (simplifying the real traffic flow rate as Qr), i.e., Qr/Qi. Under the free flow condition without external disturbance, the actual road capacity exhibits a significant downward trend as the traffic flow rate changes from the unsaturated state to the oversaturated state. Without CAV lanes, actual road capacity under the same condition is constantly improved with the increase of MPR, which can be raised by 5.7% (scene B, Qi = Qs), 14.4% (scene D, Qi = 120% Qs), 26.5% (scene G, Qi = 160% Qs), and 31.8% (scene J, Qi = 160% Qs) compared to the base scene. In addition, the actual road capacity reduces slower and its starting point of attenuation moves back as the MPR increases, which means CAVs probably enhance the robustness of the traffic flow.
The value of actual road capacity of scene A and the variation proportion of actual road capacity from scene A are shown in Table 4. It is indicated that scene C and scene F lead to serious decrease in the actual road capacity, as marked by green rectangles in Figure 4. The former will cause the attenuation of the actual road capacity up to 12.59% while the latter will result in 27.5% reduction of that to the greatest extent. The negative effects could also be attributable to the waste of road resources. Approximately similar to scene D, setting one CAV lane based on scene D, i.e., scene E, could also improve the actual road capacity, with a maximum increase of 10.6% from that of the base scene. But the benefits are slightly less than that of scene D. After CAVs become the major component of the traffic flow, the CAV lane could achieve significant positive benefits. When CAV MPR is equal to 75%, setting one exclusive lane, i.e., scene H, could enlarge the actual road capacity by up to 27.6% (Qi = 120% Qs) compared to the base scene, and the maximum increment to scene G is 1.9% (Qi = 120% Qs). Setting two exclusive lanes could also obtain positive benefits with the actual road capacity increasing by up to 13.4% compared to the base scene, which is less than scene G and scene H owing to its disturbance to the traffic flow outside the dedicated lanes.
3.2.3. Impacts on Electric Energy Consumption
The average electric energy consumption per vehicle per kilometer in different scenes under different traffic conditions is depicted in Figure 5. Under the free flow condition without any disturbance, when the traffic flow rate is unsaturated, energy consumption becomes increasingly large with the increase of the CAV MPR. It is enlarged by up to 36.6% when the MPR reaches 100%. The positive benefits of travel time from CAVs are probably responsible for that. When the MPR is smaller than 50%, energy consumption rises at first and then remains stable as Qi increases. The larger the MPR is, the gentler the energy consumption increases but the higher degree the energy consumption is stable on, which even exceeds that of the base scene. The former phenomenon is due to the consistent driving style of CAVs and improvement of traffic flow homogeneity with the MPR increasing. The latter one is because of the optimization in traffic efficiency caused by CAVs. After the MPR surpasses 50%, nearly opposite trend appears with Qi rising. As for scene G, energy consumption keeps stable basically with small fluctuation. Reduction occurs as long as Qi is not smaller than Qs, which is up to 16.9% of energy consumption in the base scene (Qi = Qs). In terms of scene J, energy consumption decreases at first and eventually reaches a platform, the value of which is up to 19.0% smaller than that of the base scene.
The value of the average electric energy consumption per vehicle and its variation proportion from that of the base scene are shown in Table 5. When Qi is smaller than Qs, no matter what the CAV MPR is and how many CAV dedicated lanes are set, energy consumption is further enlarged compared to that without the dedicated lane. It is mainly because of the internal disturbance derived from exclusive lanes as Qi overpasses Qs. Providing dedicated lanes shows extremely significant energy saving benefits. Scene C could save up to 4.2% of energy consumed in scene A (Qi = 120% Qs). Scene E and scene F could separately reduce energy consumption by 9.9% (Qi = Qs) and 13.2% (Qi = 120% Qs) maximally compared to that of the base scene. They are completely contrary to the situations with the same CAV MPR but without dedicated lanes, i.e., scene B and scene D. When the CAV MPR reaches 75%, as for scene H, up to 9.8% of energy is saved compared to that of scene G, whose net energy consumption is even lower than that of scene J, i.e., the CAV MPR is equal to 100%. In terms of scene I, it shows a stable energy consuming performance quite similar to scene G; the variation proportion of which towards the base scene fluctuated between −10.9% and −13.5%.
In order to give a better explanation of the impacts of CAVs on energy consumption, insight about vehicle accelerations is given in Figure 6. It is well known that the magnitude and frequency of accelerations (or decelerations) directly affect vehicles’ electric energy consumption. As indicated in Figure 6(a), the average acceleration of the system is calculated by working out the arithmetic mean of all acceleration values of every vehicle at every simulation moment. The average acceleration in scene A under the oversaturated traffic flow rate is extremely larger than that under the unsaturated traffic flow rate. With the CAV MPR increasing, the average acceleration under the same traffic flow rate declines significantly. In other words, the increase of CAV MPR can significantly mitigate the influence from traffic flow rate variation on the average acceleration.
(a)
(b)
(c)
The frequencies of negative and positive accelerations are shown in Figures 6(b) and 6(c), respectively. BEVs are at the state of braking energy regeneration when their acceleration values are negative. As exhibited, the frequency of negative accelerations per vehicle per kilometer ascends apparently in general as the CAV MPR increases under the oversaturated traffic flow rate, which is good for energy saving. As for the frequency of positive acceleration per vehicle per kilometer, it rises at first and then decreases to that of scene A under the oversaturated traffic flow rate with MPR of CAVs growing from 0 to 100%. Combining Figures 6(a) and 6(c), it could be inferred that the positive acceleration value in scene J is relatively smaller than that in scene A even though their frequencies of positive accelerations are nearly consistent.
3.2.4. Economic Benefits
The economic benefits and their compositions are depicted in Figure 7. When the traffic flow rate is at the unsaturation state, negative are obtained even though is almost positive. And the absolute value of the negative turns larger as the CAV MPR increases. It is caused by the reduction of travel time, which contributes to more energy consumption, i.e., negative EEB covering up the positive . Positive emerges as long as Qi is not lower than Qs. The maximum is 1.18 billion yuan appearing in scene I when Qi reaches 160% of Qs, while the minimum value is 0.03 billion yuan in scene B when Qi is equal to 140% of Qs. The first rises and keeps stable eventually. At the oversaturation state, the rapid growth of emerges when the CAV MPR exceeds 50% without setting any exclusive lane. It is interesting to find that there are several scenes with negative EEB, which is quite different from the analysis before. The increase of actual road capacity could be accounted for that. Although average energy consumption in many scenes is reduced, the rise of traffic volume would expand total energy consumption, contributing to the offset of the energy saving benefits.
Setting a certain number of dedicated lanes in appropriate scenes when Qi is not lower than Qs, such as scene E, H, and I, helps to further amplify the traffic benefits from CAVs. Under different traffic flow rates, derived from providing one or two dedicated lanes as the CAV MPR reaches 75% is close to or even outweighs that with the CAV MPR reaching 100%. The in scene C and F, as indicated by the red dots in the Figure 7, are also prominent with the most contribution from . However, they may not be desirable solutions because their CEBs are negative, which means that the actual road capacities in these scenes decrease. In the short term, the lost traffic volume has to flow into other road sections so that disturbance or even congestion would probably arise and spread within the road networks. In the long term, the change in the travel mode structure is likely to emerge that vehicle trip volume may decrease.
The hourly traffic flow rate during the peak periods on working days in Beijing is about 120% of the saturated traffic flow rate. And the MPR of CAVs in Beijing has been already more than 30%. Thus, from L2 CAVs during peak periods on weekdays in Beijing is about 0.12 billion yuan, as indicated in Figure 8. It is estimated that the CAV MPR will reach 50% in 2025. Then, the is expected to be approximately 0.19 billion yuan and setting a CAV lane could probably increase the by nearly 3.0 times to 0.56 billion yuan when the MPR of CAVs is equal to 75%, providing one or two dedicated lanes can reap the same benefits as those when it reaches 100%. Scenes C and F are infeasible due to their negative CEBs, which are marked with the red dots in the Figure 8.
4. Conclusion
In this paper, impacts of L2 CAVs on traffic efficiency, energy consumption, and the relevant economic benefits from their influences were systematically studied with the consideration of various CAV’s MPRs and traffic flow rates. What is more, traffic impacts of exclusive lanes for CAVs, a traffic organization method expected to amplify the optimization effects from CAVs, were also analyzed. An evaluation framework of CAV’s traffic economic benefits was creatively put forward, which integrates the traffic impacts and economic benefits together. It is a framework with strong universality for different cities and even countries. The impacts on travel time, actual road capacity, and electric energy consumption were exhibited with the MPR of L2 CAVs rising from 0 to 100% at different traffic flow saturations. It is found that L2 CAVs could save average travel time and reduce average energy consumption in most scenes in terms of a single vehicle. But as for the overall benefits, negative benefits of energy consumption were observed due to the increase of actual road capacity, which is approximately equivalent to an increase in traffic volume. A case study was conducted about the traffic economic benefits from L2 CAVs for Beijing’s expressways during peak periods on weekdays. As for the traffic condition and CAV MPR of Beijing, L2 CAVs are expected to generate nearly 0.6 billion yuan of economic benefits if a CAV lane in about two or three years on Beijing’s expressways is set up.
Three suggestions to realize traffic benefits of L2 CAVs are proposed according to the research results as follows:(1)Governments should promote the implementation and application of CAVs through policy, financial, market, and other means. The increase of the MPR of CAVs is the precondition to release their traffic benefits. What is also significant is to carry out CAV’s traffic benefit assessment and vigorously publicize its benefits and advantages, which is expected to stimulate consumers’ enthusiasm to purchase CAVs. Thus, a positive cycle will be formed.(2)The combination of intelligent vehicles and appropriate means of traffic organization can further promote the realization of the traffic benefits from intelligent vehicles. As indicated in the previous sections, setting proper number of dedicated lanes for CAVs could amplify their traffic benefits apparently, which shows the same or even better effects compared to those derived from the increase of CAV MPR.(3)It is crucial to improve the installation rate of V2X modules on vehicles and accelerate the intelligent upgrading of transport infrastructures. On the one hand, it is well known that the difference of traffic volume in different periods of a day is extremely significant and the proportion of CAVs on a road section is not necessarily equal to their MPR. On the other hand, as claimed before, there are several scenes where providing CAV lanes show negative impacts on actual road capacity because of the relatively low MPR of CAVs. Therefore, CAV lanes do not need to be set for the whole day but only in suitable periods. The traditional physical isolation method of lanes may not be suitable for such flexible requirements. The intelligent transportation infrastructure can perceive the traffic flow status in real time and predict the time period when CAV lanes need to be provided. Then, lane assignment information is broadcasted to vehicles through V2I (vehicle to infrastructure) communication and information board on roadside. Vehicles that do not obey the rule will be fined.
There are still some points remained to be improved. First, in order to explore the maximum capacity of L2 CAVs for traffic optimization, no ramp was simulated in order to avoid traffic conflicts, which is relatively ideal. Second, all vehicles were assumed to be BEVs so that the development trend of the vehicle power type could be characterized. However, the MPR of BEVs is still lower than that of fuel vehicles though its growth rate rapidly increases. Third, L2 CAV’s impact on traffic safety is not included, but reducing accident rates is an important function of CAVs. So, in the following research studies, expressways with ramps will be modeled to introduce conflict zones and more advanced functions of CAVs may be involved such as ramp intelligent merging and diverging as well as lane change assistance. The prediction results of development of the vehicle power type will be introduced in the calculation of traffic energy consumption. Additionally, CAV’s impact on traffic safety and relevant economic benefits will also be analyzed.
Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Authors’ Contributions
H. S., F. Z., and Z. L. came up with the research topic. H. S. selected and modified the vehicle models and the energy consumption model. H. S. constructed the economic benefit evaluation model assisted by G. Z.. H. S. completed the simulation experiments and gave explanation about the results. F. Z. and Z. L. checked the results. H. S. wrote the main manuscript text and supplementary information. All authors reviewed the manuscript. Z. L. applied for the funding.
Acknowledgments
This research was supported by the National Natural Science Foundation of China (5227120041).
Supplementary Materials
The description of the MIXIC model, Wiedemann 99 model, VT-CPEM model, and Tables S1–S4 are given. (Supplementary Materials)