Abstract
The positive effect of traffic-responsive signal control can be assured when real-time traffic data is reliable, but data reliability may be an issue that depends on the number of probe vehicles equipped with navigation devices or smartphones. However, there is a high chance of improving reliability with the recent deployment of connected vehicles (CVs) that use the vehicle-to-everything (V2X) communication data. Therefore, this paper proposes a traffic signal control strategy that utilizes V2X communication data obtained from CV operations, which is called the capacity waste reduction (CWR) strategy. In this strategy, vehicle queues on each road lane as an intersection approaches are initially estimated using V2X data. Then, the signal control algorithm determines the duration of the green signal for the currently applied phase based on the estimated vehicle queues. Furthermore, the strategy includes an algorithm for active priority signal control for the vehicles of bus rapid transit systems. The efficiency of the provided control strategy is tested with the VISSIM microsimulation program at different levels of the market penetration rate (MPR) of CVs. Based on the results of the experiment, the proposed strategy shows positive effects in both decreasing travel delay and increasing traffic flow even at the low levels of MPR of CVs. The results of the proposed strategy can be used as the base data for the development of smart intersection operations.
1. Introduction
Traffic congestion is currently one of the major urban issues owing to population growth and city-centered life patterns. To mitigate traffic congestion, the field of urban traffic control has been studied and developed in various ways over the past decades, and researchers have been developing strategies for traffic-responsive signal control. The signal control system functions by allocating priorities for various types of traffic flow on the roads for safety and mobility. Hence, traffic flow dynamics on urban roads where numerous intersections spatially exist are crucially influenced by a given traffic signal control system [1]. With this background, numerous strategies of traffic-responsive signal control have been studied and developed such as the Split Cycle Offset Optimization Technique (SCOOT) [2] and Sydney Coordinated Adaptive Traffic System (SCATS) [3], along with the adoption of intelligent transportation systems (ITS).
Previous studies have been conducted from a few different perspectives. Some studies have focused on managing traffic flow at a single intersection [4–6] using real-time data collected through automatic vehicle identification (AVI) technologies like loop detectors, video cameras, and global positioning system (GPS). Others have focused on strategies for road network-wide control solutions, with representative works such as the Optimized Policies for Adaptive Control (OPAC) [7], PRODYN [8], RHODES [9], and Traffic-responsive Urban Control (TUC) [10]. The common issue in the network-wide strategies is the highly complex optimization algorithms owing to multiple control agents according to the number of intersections within the networks. Thus, the recent trend has been toward conducting distributed approaches [11, 12] or machine learning techniques [13–15] for control optimization. Another type of approach is the perimeter control that manages the amount of transfer flows between urban regions in large-scale level. This approach has been intensively studied in the last decade, in particular [16–20].
Irrespective of the approach employed, traffic-responsive signal control can increase its efficiency only when reliable information on the actual traffic situation on the roads is collected, such as the length of vehicle queues for individual road lanes or the delay time of individual vehicles. The use of GPS on probe vehicles equipped with navigation devices or smartphones is the preferred way of collecting real-time traffic data on road sections these days owing to broad spatial coverage and low operational costs [21]. However, a major issue in using the GPS-based data is that information reliability is reduced when the number of probes is low compared with the entire traffic volume within a road section [21, 22]. Furthermore, it is difficult to obtain detailed lane-by-lane traffic information [21, 23]. Since not all vehicles use GPS-based services, it is necessary to find alternatives to improve the process of collecting more detailed real-time traffic data.
One of the possible solutions to resolve the aforementioned issue is to utilize the vehicle-to-everything (V2X) communication data, which is a part of the cooperative ITS (C-ITS) technology. C-ITS is operated by exchanging microscopic driving-related data from vehicle to vehicle or vehicle to infrastructure. Such datasets can be useful in deriving traffic control solutions because they include details such as real-time vehicle locations, accelerations, and speed of multiple vehicles in each road lane.
Several studies on using V2X communication data for traffic signal control have been conducted. Most of these studies focused on signal control at a single intersection, considering that it is currently the initial stage of deploying such communication technology. Lee and Chiu [24] showed the capability of V2X for a few smart city applications like emergency vehicle preemption and public transit signal priority, but this work was more into the functional tests than the analysis on the effect of signal control on the traffic flow near intersections. There are some other existing studies that analyze the effect on traffic flow, such as Goodall et al. [25] and Jafari et al. [26], but their works used the traffic information in the unit of road links, which did not fully utilize the capability of V2X data that can provide lane-by-lane information. The examples of using lane-by-lane traffic information were shown by Lee and Park [27] and Xu et al. [28]. Their work showed that the control algorithms for connected vehicles (CVs) using V2X communication can improve the traffic flow. However, their methods were designed under the assumption of a 100% market penetration rate (MPR) of CVs, which may not be suited yet for the mixed traffic operations these days where legacy vehicles, autonomous vehicles, and CVs exist. Considering mixed traffic operations, Feng et al. [29] and Han et al. [30] proposed V2X data-based traffic signal control strategies and analyzed their effect by different MPR environments. Their strategies showed improved performance in reducing the travel delay of vehicles. The common feature of the strategies is that they used the signal phase allocation algorithms, which determine the signal phase sequences at given intervals by changing the order of phases to be applied. However, in general, the signal phase sequences are fixed as an almost indispensable constraint in traffic signal control because they can confuse drivers and cause over-waiting for vehicles owing to the randomly-given sequences if such a constraint is ignored [31]. Hence, it is necessary to consider a feasible approach rather than the mathematical optimizations that lack consideration of real-world constraints.
This study aims to provide a traffic signal control strategy that utilizes V2X communication data obtained from CV operations, which is called the capacity waste reduction (CWR) strategy. In this strategy, vehicle queues on each road lane as an intersection approaches are initially estimated using V2X data. Then, the signal control algorithm determines the duration of the green signal for the currently applied phase based on the estimated vehicle queues. Furthermore, the strategy includes an algorithm for active priority signal control for the vehicles of bus rapid transit (BRT) systems. The provided control strategy is to be tested using the VISSIM microsimulation program [32], and the test will focus on the improvement of traffic congestion at different MPRs of CVs.
The remainder of this paper is organized as follows: The second section describes the framework and control algorithms of the CWR strategy; the third section describes the settings for testing the proposed control strategy, including traffic scenarios and indices for evaluation; the fourth section presents the results of the tests on the effects of both traffic improvement and active signal priority for BRT; and the last section concludes this paper with considerations for further works.
2. Materials and Methods
2.1. Framework for Signal Control Based on V2X Communication Data
The overall structure of the signal control strategy using V2X communication data is shown in Figure 1. This strategy is divided into (1) collection of lane-by-lane traffic information and (2) signal control strategy based on lane-by-lane traffic information. For this configuration, a strategy to reduce road capacity waste that can be caused by intersection signal control will be used, which is called the “capacity waste reduction (CWR)” signal control strategy. Individual modules are described below.(i)Vehicle coordinate extraction: The location coordinates of individual vehicles are extracted from V2X communication message information.(ii)Vehicle location extraction: The location information of vehicles running on individual lanes is extracted by matching the extracted vehicle coordinates with the lane information in HDmap.(iii)Vehicle arrival time estimation for each phase: The vehicle arrival time to the intersection is estimated based on the location information of vehicles on each lane.(iv)Intersection signal control: The appropriate signal control input is determined in accordance with the state of the vehicle queue by calculating the remaining green signal length of the current signal phase and comparing it with the vehicle arrival time.(v)Active signal priority control for BRT (additional function): BRT signal priority is additionally operated to improve the efficiency of BRT vehicle traffic.
This study focuses on proposing a detailed control method in terms of the implementation of functions of traffic information-based signal control and conducting various analyses to evaluate it. The focus of the analysis is on the intersection signal control methods including “vehicle arrival time estimation for each phase” and “intersection signal control” functions using V2X communication data. Furthermore, analysis of the “active signal priority control for BRT” function is performed to assess the expandability of the method toward various forms of traffic control solutions.
2.2. Review of the Related Data
The signal control strategy proposed in this study utilizes the information that is collected from individual vehicles with V2X communication capability near intersections. Such information is provided in certain formats, such as basic safety message (BSM) and probe vehicle data (PVD) sets, which were established by the society of automotive engineers (SAE) in terms of SAE J2735 [33].
In this study, it is assumed that among the corresponding messages, the real-time coordinates (latitude, longitude) of autonomous cooperative driving vehicles, the vehicle speed of PVD, and vehicle ID (probe ID), which are common parts in the PVD and BSM sets, are used as the main information. Furthermore, the signal phase and timing data may be used to achieve real-time signal control. In addition, the identification number of intersection (intersection ID), information on which signal phase is being executed (event state), and remaining signal time of the current signal phase (timing) are used. The basic direction is to calculate the signal control algorithm and collect data in second units. The data format used in this study is summarized in Table 1.
2.3. Intersection Signal Control for CWR
2.3.1. Basic Concept
The basic concept of the CWR control strategy is to reduce the waste of road capacity that can be caused by inefficient intersection signal control. For example, at an intersection where a main road meets with a subroad, if green signals are given more than necessary for vehicles on the subroad with significantly lower traffic, as shown on the left side in Figure 2, it is likely that green signals are applied even after all the vehicles on the subroad have passed, resulting in inefficient use of road capacity.
Minimizing the occurrence of these cases is the basic concept of the CWR control strategy. CWR applies a real-time control method that predicts the intersection arrival times of vehicle queues on individual lanes, stops the corresponding green signal when the queues decrease and disappear, and then gives the green signal in the next signal phase. In this way, road capacity waste can be reduced by minimizing unnecessary signal waiting at intersections, and an increase in overall road traffic can be promoted, as shown on the right side of Figure 2. The CWR strategy can be considered an actuated control that dynamically applies the minimum green signal time according to the traffic situation to minimize unnecessary signal waiting at independent intersections. However, there is a difference in the method of collecting traffic volume and queue information based on V2X communication. Furthermore, the detailed decision-making method is different from the existing general actuated control because the current signal phase, which is applied based on predefined rules, is immediately stopped according to the real-time situation, instead of calculating and applying the minimum green signal time in advance when the signal phase is changed.
2.3.2. Lane-by-Lane Information
In this study, it is assumed that the detailed traffic information on each lane of the road is collected through information matching between the data of V2X communication vehicles and the data of the high-definition map (HDmap), which is a precise electronic map. In the process of extracting the vehicle position information on each lane of the road, the location coordinates of individual vehicles are first extracted from the V2X communication information including vehicle speed and driving direction. Subsequently, the length, shape, and type of the corresponding lane and the longitude and latitude data of the starting and ending points are extracted from the link property information of each lane on the road of the HDmap. The detected location coordinates of the extracted vehicle are compared with the lane information of the HDmap and matched with the nearest lane link to allocate the lane on which the vehicle is currently running. Based on this, the location information of vehicles running on individual lanes is extracted. This study assumed that the real-time location information of vehicles on each lane of the road is collected through this process and proposed the CWR signal control strategy based on it.
2.3.3. Vehicle Arrival Estimation Algorithm
V2X communication vehicles are mixed with general vehicles on the road. Among them, the queues of individual lanes corresponding to the current green signal phase are estimated based on the data collected from V2X communication vehicles.
Considering the dynamic behaviors of vehicles on the road, there are rear-end vehicles that continuously approach the intersection. Therefore, we need a criterion to separate the rear-end vehicles and select the rearmost part of the queue at the intersection when estimating the vehicle arrival time at the intersection. This criterion is necessary because even if the current maximum back of the queue (MBQ) is removed and the current signal phase is immediately stopped according to the concept of the CWR signal control, there can be additional rear-end vehicles that approach the intersection. Depending on the case, it can be more efficient for intersection operations to process these rear-end vehicles first. Hence, the method of immediately stopping the current signal phase unconditionally with the removal of the MBQ can cause adverse effects.
Therefore, to prevent such effects, this study proposes the concept of Estimation Boundary Upon Remaining Green (EBURG), which is the red line on the road as shown in Figure 3, based on the remaining green signal length of the current signal phase. Even though there are additional approaching vehicles in the rear, if these vehicles are outside the current range of the green signal, they cannot pass the intersection at the current green signal; hence, there will be no negative effect of immediately stopping the current signal phase. In contrast, if approaching rear-end vehicles are within the range of the current green signal, the current signal phase is maintained considering the approach of these vehicles. In this study, the queue estimation range is set based on the EBURG, then the MBQ vehicles are selected, and finally the vehicle arrival time at the intersection corresponding to the current signal phase is estimated.
The CWR signal control method proposed in this study is performed based on the estimated vehicle arrival time described previously. The related signal control variables are described in Table 2. The phase system of signal intersections can be composed of various combinations of vehicle movement behaviors such as going straight, turning left, and turning right for each traffic direction. Although one movement is assigned to each individual phase like a single-ring or dual-ring phase system [34], in actual signal operation, multiple movements are simultaneously assigned to individual phases in most cases. One representative example is the straight forward and left turn simultaneous signals. In actual signal operation, whether to select straight-forward and left-turn simultaneous signals or separate signals can be decided depending on the number of lanes of the approach, intersection shape, and traffic volume. For reference, the method proposed in this study is presented in such a manner that it is possible to implement regardless of the applied phase system.
The arrival time of every individual vehicle related to the current signal phase is calculated based on the current location and speed of all the collected autonomous cooperative driving vehicles and is presented as follows:
Note that, using the data in Table 1, of each vehicle is calculated every second. Even if a deceleration or acceleration behavior occurs during the intersection approach, the changed speed and location upon such behavior are applied to the calculation. Then, of each vehicle is updated every second based on the dynamic behavior.
To select the MBQ, , which is the length of time (current green) that has been applied from the time when the green signal of the current signal phase was applied until the present in the real-time signal control information system, is collected. The equation for calculating the length of the remaining green signal time from the present by comparing the collected information with the maximum green signal time of the corresponding phase is as follows:
The range of EBURG in Figure 3 is determined based on the remaining green signal time. Based on this, it is necessary to distinguish only the vehicles within the corresponding range among the intersection arrival times of all vehicles that have been collected before as the queue estimation range of the current signal phase. To this end, the arrival time set of only the vehicles within the corresponding range is formed separately as follows:
The maximum value in the arrival time set is selected as the MBQ vehicle corresponding to the current signal phase, and the equation for estimating the intersection vehicle arrival time of the current signal phase based on the MBQ vehicle is as follows:
The aforementioned intersection vehicle arrival time estimation process was developed as an algorithm, and is illustrated in Figure 4.
2.3.4. Intersection Signal Control Algorithm
CWR applies a real-time control method that estimates the intersection arrival time of the vehicle queue, immediately stops the current signal phase when the queue decreases and disappears, and then assigns a green signal to the next signal phase. The time when the vehicle arrival time is estimated is set as the time when the corresponding phase is applied. However, equation (1) cannot be established if the V2X communication-based vehicle position data are not collected depending on the road situation, and the situation of the intersection approach road cannot be estimated. In this case, if the situation of the intersection approach road cannot be known clearly, the signal phase being applied is maintained, as shown in Figure 5.
When the vehicle arrival time is collected, the existence or absence of a queue can be estimated. When there is a queue, the value of is different from ; thus, the vehicle arrival time will continue to have a value larger than zero by equation (1). In this case, the signal phase is maintained only until the maximum green signal time. In contrast, if the queue decreased and disappeared within the maximum green signal time, this means that the MBQ vehicle arrived at the intersection. In this case, the value of becomes the same as , the vehicle arrival time becomes zero by equation (1), and it is estimated that there is no queue. Hence, the current signal phase is immediately stopped. In this case, the green signal control input for each case is determined as follows:
2.4. Active Signal Priority for BRT
2.4.1. Estimation Algorithm for Arrival Time of BRT Vehicles
Assuming that BRT vehicles are operated based on V2X communication, this study proposes a BRT signal priority, an additional feature of the CWR strategy. Signal priorities are classified into passive priority and active priority [35]. In this study, active priority is applied.
The BRT signal priority application method proposed in this study is based on the intersection arrival time of BRT vehicles. The applied BRT vehicle arrival time estimation method for the green signal phase is similar to the intersection V2X communication vehicle arrival time estimation method presented previously. The related variables are described in Table 3.
The current location of every collected BRT vehicle is defined as , and the current speed as . Based on these two data, the arrival time of every individual BRT vehicle related to the corresponding signal phase is calculated as follows: where denotes the dwell time for passengers of a BRT vehicle at a station. In this study, this variable is inputted as 20 s if there is a station near an intersection, and 0 s otherwise.
Note that of each BRT vehicle is calculated every second. Hence, the deceleration or acceleration behaviors can be reflected in the calculation. 20 s is selected for the constant value of based on the work of [36] which provides that the average dwell time in general situations is between 10 s and 20 s. We selected the maximum value of the given range.
When two or more BRT vehicles approach, an arrival time set of vehicles is formed, and the vehicle with the maximum value among the BRT approach vehicles is selected as the MBQ vehicle. The equation for estimating the intersection BRT vehicle arrival time of the corresponding phase based on the intersection arrival time of the MBQ vehicle is as follows:
2.4.2. Algorithm of Signal Priority Control for BRT Vehicles
When the queue in a general lane decreases and disappears, the green signal of the corresponding signal phase can be immediately stopped accordingly. However, when a BRT vehicle approaches the intersection, a method of deciding the priority signal applicability is necessary, and if it is applicable, a method of deciding the scope of the priority signal’s application is also necessary. The conditions for BRT signal priority decisions are described in the succeeding paragraphs.
As shown in Table 4, the first condition for a signal priority decision is the remaining green signal time of the current signal phase. As shown in Case 1, even if a BRT vehicle is approaching the intersection and the remaining green signal time is zero, it is impossible to apply signal priority; therefore, it is not applied.
The second condition for the signal priority decision is based on a comparison of the remaining green signal time of the current phase and the BRT vehicle arrival time . As shown in Case 2, if a BRT vehicle is approaching the intersection, and the current remaining green signal time is greater than or equal to the BRT arrival time , it is not necessary to apply signal priority; thus, it is not applied. Conversely, if the remaining green signal time is smaller than the current BRT arrival time ), it is necessary to apply signal priority, but a decision needs to be made about whether or not to apply signal priority.
To make a decision on signal priority application, this study proposes the concepts of the required time for priority, , and the maximum allowed time for priority . indicates the difference between the BRT vehicle arrival time and the remaining green signal time and is calculated as follows:
Whether or not to apply signal priority can be decided by comparing the calculated with , which can be determined by the road situation.
As shown in Case 3, if a BRT vehicle is approaching the intersection, and if is smaller than , the signal priority for BRT vehicle is applied by adding the signal length as much as to the current remaining green signal time. Conversely, as shown in Case 4, if a BRT vehicle is approaching the intersection, and if is larger than , the signal priority for BRT vehicle is not applied by carrying out the current remaining green signal as it is.
The aforementioned priority signal application method for BRT vehicles at intersections was developed into an algorithm and is shown in Figure 6.
3. Test Settings
3.1. Test Site
In this study, an environment in which signal control effects in various scenarios can be analyzed was constructed by applying a new signal control strategy and algorithm based on VISSIM, an existing commercial traffic simulation program.
A network was configured in the same manner as the actual road environment based on the BRT axis of Sejong City Living Zone 1, as shown in Figure 7. To classify and analyze intersections according to their type, the areas for effect analysis were configured as follows:(i)Intersection type 1: General intersection (Seongguem intersection, located in front of the Ministry of Personnel Management in Sejong City)(ii)Intersection type 2: Intersection including the BRT route (Intersection located between stations on the north and south of the government complex through which the BRT route passes)(iii)Intersection type 3: Intersection including the BRT route and stations (Intersection located near the south station of the government complex)
Note that, as shown by the blue lines in Figure 7, only the road sections directly connected to the three intersections are included for evaluation at the network level. There are a total of 32 road sections in each direction within the network, and the average length of the road sections is approximately 250 meters.
3.2. Traffic Scenario Settings
In the testing road network, all 32 road sections have two lanes for general vehicles. There are also two separate BRT lanes crossing the network (one from the north to south direction and one from the south to north direction). The speed limit for all road lanes is set to 50 km/h just like at the actual site.
For the traffic setting in the simulation of this study, the annual average actual traffic volume for a peak time of two hours in 2020 was applied. The 15 min traffic (point/total) based on vehicle detection sensors provided by Sejong City Traffic Information was converted to 30 min units. The traffic volumes reflected in the corresponding section in 30 min units are 3136, 3462, 4476, and 3934. The traffic is set to be generated at the boundaries of the network in the simulation. The details of the traffic volume input are provided in Table 5.
The signal control information of the three types of intersections selected in this study was provided by the Korea Road Traffic Authority (KoROAD) that manages the Sejong City intersection operation database. The received signal control data were applied by the default signal control method for comparison before and after CWR application in the simulation of this study. This default method is operated by the Time of Day (TOD) method based on traffic pattern analysis. Thus, this study assumed that the default signal control method is the closest to the optimum method in a nonreal-time control environment and selected the default method as the control for comparison with the real-time control method proposed in this study.
In the simulation of this study, the ratio of vehicles equipped with an onboard unit capable of V2X wireless communication in total traffic on the simulation is defined as the MPR of V2X communication vehicles. To analyze the intersection signal control operation efficiency at various conditions in terms of V2X-based information collection, the MPR environment (ratio of V2X communication vehicles to total traffic) was set to 0%, 25%, 50%, 75%, and 100%.
In this study, the traffic impact according to various tolerances was evaluated by setting the as the “maximum allowed time for priority control” discussed previously in relation to the CWR method. was set to 0 s, 20 s, 40 s, and 60 s in this study.
For the allocation intervals of BRT vehicles, a range of 5 to 10 min was applied, which is the actual allocation interval of the B2 route (Banseok Station to Osong Station) in operation by the Sejong City Transportation Corporation. The spatial scope of BRT in the simulation was approximately 2.2 km from the South Station of Sejong Government Complex to Sejong Chungnam University Hospital, which corresponds to Living Zone 1, and the number of BRT vehicles applied to this simulation was 64 vehicles in total (32 toward Osong Station and 32 toward Banseok Station).
3.3. Indices for Evaluation
The indices for evaluating the CWR-based intersection signal control strategy selected in this study are discussed in the succeeding paragraphs. The time series indices, which are not one-time indices such as intersection flow and vehicle delay, were set to be calculated in 5 min units for detailed analysis of the changes owing to the nature of this study, in which real-time signal control changes can occur frequently.
The first index is flow, for which the sum of all vehicles that passed an intersection from each approach of the intersection for 5 min is calculated as 1 hour unit flow rate and is presented in vehicles per hour (vph). The second index is the cumulative vehicle hours traveled (VHT). The cumulative VHT is the amount of VHT accumulated over all road sections applied to the simulation. This index is used for the evaluation of traffic congestion of all sections and is presented in veh․hr unit. The third index is the average vehicle delay, for which the delay of every vehicle passing through an intersection from each approach of the intersection is calculated as a 5 min average in seconds directly from the simulation. The fourth index is BRT vehicle travel time, which is the travel time of every BRT vehicle in every section applied in this simulation and is for comparison with before and after BRT signal priority application of CWR. This index is presented in seconds and is directly calculated from the simulation.
The last index is CWR Efficiency. This index has been suggested considering the characteristic of the control method of this study that immediately stops the current signal phase. Before the CWR Efficiency index, the concept of saved capacity (SC) is presented first. SC is the amount by which the waste of traffic capacity is decreased because the current signal length is not applied to the maximum range and is immediately changed to the next signal phase by the CWR algorithm. It is presented in vph and is calculated by the difference between the existing traffic capacity of the individual signal phase and the traffic capacity after application of CWR.where : signal phase of intersection , : existing cycle length of intersection , : maximum green of signal phase , : saturation flow rate of j lane corresponding to signal phase , : applied sequence number of signal phase (e.g., p-th applied signal phase k among simulations), : total number of applications of signal phase k, : the length of green signal of the p-th signal phase k changed by CWR, : existing traffic capacity of signal phase k at intersection , : traffic capacity saved by CWR in p-th signal phase k at intersection y, and : traffic capacity saved by CWR in all signal phases at intersection y (SC)
CWR efficiency is the degree of flow that actually improved compared to the SC owing to the application of CWR. It is unitless and is calculated as follows:where : throughput at intersection y before CWR application, : throughput at intersection y after CWR application, : traffic capacity saved by CWR in all signal phases at intersection y, and : efficiency of the traffic capacity saving by CWR.
Because this index has been proposed in consideration of the characteristic of the control method proposed in this study that immediately stops the current signal phase, it can be applied only to control methods similar to the proposed method.
4. Test Results
4.1. Effect of Traffic Improvement
4.1.1. Effect by Intersection Types
As summarized in Table 6, in general intersections (type 1), a certain level of the effect of the CWR signal control strategy can be obtained in both aspects of flow and delay at 25% MPR conditions. Cumulative flow is increased by 10.1%, and the average vehicle delay is reduced by 14.8% at this MPR condition. More apparent effects can be observed at MPR conditions greater than or equal to 50%. Cumulative flow is increased by nearly 28%, and the average vehicle delay is reduced by nearly 50% at this MPR condition. The differences in the effect of this strategy in the 50–100% MPR range are insignificant.
These results suggest that a positive effect can be observed even if the MPR is only 25%, and the effect can be even increased further by increasing the MPR to more than 50%. As mentioned in the introductory section, we postulated that irrespective of the signal control approach, the efficiency of control can be increased when information is more reliable. The results of this study show that the postulate is true. Furthermore, because the performance difference is insignificant from MPR of 50% to 100%, it can be assumed that the real-time traffic information collected when MPR is at least 50% can be a proxy of the actual traffic situation.
The number of signal phase changes increased by CWR application was 112 to around 200. This implies that the CWR strategy made early signal changes to the next signal phase more frequently by estimating the vehicle queue approaching the intersection and reducing the length of unnecessary green signal time. At 25%–50% MPR, phase changes occurred more frequently (more than 230 times) because the vehicle queue is estimated based on incomplete information compared with that of a higher MPR condition. At 75%–100% MPR, unnecessary signal phase changes were reduced because the vehicle queues were more accurately estimated based on more collected information.
The SC and CWR efficiency indices were calculated to quantitatively verify this phenomenon. SC indicates the degree of traffic capacity that is reduced whenever a signal is changed earlier. Furthermore, CWR efficiency is the increased rate of total flow at the intersection compared with the unnecessary capacity saved by the CWR strategy. In other words, this index shows the significance of changing the traffic capacity in each traffic direction at the intersection. It should be noted that if the value of this index is less than one, using the CWR strategy is not ideal because the increased rate of total flow is less than the SC. On the other hand, if the value of the index is greater than one, there is a positive effect and the CWR strategy can be applied. The index calculation produced high results: the efficiency is 7.7 at 25% MPR and higher than 20 at 50% or higher MPR. These results suggest that the CWR strategy generally has substantial effects at general intersections during peak hours and particularly high effects at 50% or higher MPRs.
Table 7 summarizes that the intersection of a BRT route (type 2) has quantitative differences from general intersections owing to the difference in existing traffic volume, but the effect of increase or decrease patterns is similar. At 25% or lower MPR conditions, a certain level of effect of the CWR signal control strategy can be observed. Cumulative flow is increased by 18.5% percent, and the average vehicle delay is reduced by nearly 47.0% at this MPR condition. Much larger effects can be seen at 50% or higher MPRs. Cumulative flow is increased by more than 40% percent, and average vehicle delay is reduced by more than 60% at this MPR condition. However, the differences in effects of this strategy in the 50–100% MPR range are insignificant. Furthermore, the CWR efficiency index values are 4.8 at 25% MPR and nearly 10 at an MPR equal to or greater than 50%, which imply positive effects of applying the CWR strategy. Although there are only quantitative differences due to the given traffic volume, the effectiveness patterns in type 2 intersection resemble the patterns of type 1 intersection.
Table 8 summarizes that although the intersection that includes BRT station (type 3) has quantitative differences from other intersections owing to differences in existing flows, the effect increase and decrease patterns are similar. At 25% or lower MPR conditions, a certain level of effect of the CWR signal control strategy can be observed. Cumulative flow increased by 17.1% percent, and the average vehicle delay was reduced by nearly 36.3% at this MPR condition. Much larger effects are expected at 50% or higher MPRs. Cumulative flow is increased by more than 20% percent, and the average vehicle delay is reduced by more than 50% at this MPR condition. However, as with other intersections, the differences in effects of this strategy in the 50–100% MPR range are insignificant. CWR Efficiency values are low in this intersection because the existing traffic volume is relatively smaller than that of other intersections. Still, the values are greater than or equal to one, which still shows a positive effect. Although there are only quantitative differences due to the given traffic volume, the effectiveness patterns in this type 3 intersection resemble those of the other two intersections. These results suggest that the proposed control strategy of this study is consistent regardless of the type of intersection.
4.1.2. Effect on the Entire Network
Figures 8(a) and 8(b) show the changes of cumulative VHT and average vehicle delay of the entire network, respectively. Table 9 summarizes the numerical results of the effects of CWR intersection signal control for the entire road network. As shown in these figures and table, compared with the case of 0% MPR to which conventional signal control was applied, the cumulative VHT decreased in all the cases where the CWR signal control strategy was applied (MPR 25–100%). In the case of 25% MPR, the degree of congestion decreased by 21.6%. In the case of 50% MPR, the degree of congestion decreased by 28.7%. Other cases showed similar decrease rates of around 40%. In terms of vehicle delay, the CWR signal control strategy generally decline the vehicle delay compared to the conventional signal control. The average vehicle delay decreased by 31.6% at 25% MPR, 55.8% at 50% MPR, and 60% or higher at other conditions.
(a)
(b)
The number of signal phase changes generally decreased by CWR application was 384 to around 600. This implies that the CWR strategy made early signal changes to the next signal phase more frequently by reducing the length of unnecessary green signal time. As mentioned above, if the value of CWR Efficiency is smaller than one, it means that applying the CWR method itself is meaningless. In this study, the calculated CWR efficiency was 3.6 at 25% MPR and 9.0 or higher at other MPR conditions. These values suggest that the proposed CWR signal control strategy will be effective for traffic improvement when V2X communication data are used and even when the MPR is 25%.
These results are consistent with those of the individual intersections, and this is demonstrated by the outcome that the positive effect of the proposed control strategy increases as MPR increases and the effect reaches the maximum performance when MPR is greater than 50%. These results demonstrate the significance of the reliability of real-time traffic information. Hence, it is shown that the CWR control strategy, which utilizes V2X communication data for increasing the reliability of traffic information, is a promising solution for improving traffic management at intersections.
4.2. Effect of Active Signal Priority for BRT
4.2.1. Travel Time of BRT Vehicles
Table 10 shows the travel time improvements of BRT vehicles when the signal priority is applied based on the CWR strategy. As shown in this table, at 25% MPR, the average travel time increased, causing an adverse effect in every case where (the maximum allowed time for priority control) was 5 to 15 s. However, the average travel time decreased by 1.1–4.3% in all the other conditions. Moreover, the higher the MPR, the greater the travel time reduction effect. This means that the effect of signal priority increases in general when the collection rate of vehicle travel information is high, thus increasing the accuracy of information. This suggests that the proposed CWR-based signal control strategy has promising performance.
However, the quantitative values of improvement in the BRT travel time are rather low in general. Furthermore, the travel time increased at 25% or lower MPRs where the collection rate of vehicle travel information was low. Hence, the BRT signal priority algorithm proposed in this study has room for improvement in follow-up research so that the effect can be further increased under all conditions including low MPRs.
Unlike the comparison by MPR condition, the effect of the increase and decrease according to the change in the maximum allowed time for priority control did not show a clear pattern. At 50% MPR condition, the travel time reduction effect was the largest when the was 10 s. At 100% MPR condition, the travel time reduction effect was the largest when the was 5 s. These indicate that the effect of BRT signal priority may not be much sensitive to the maximum allowed time for priority control or the level of travel information collection for surrounding vehicles. This phenomenon also requires in-depth examination in future research.
4.2.2. Impact on General Road Lanes
Figure 9 shows the average vehicle delay of general roads where BRT signal priority is applied based on the CRW strategy. When we examine the average vehicle delay on general roads for each MPR condition by the BRT signal priority application in this figure, the delay pattern on general roads was similar to the pattern before the BRT signal priority application in every case. In some conditions, there was a negative effect of slightly increasing the delay, but the increase was less than 2.0%. This indicates that the effect of the BRT signal priority on general roads is not substantial. On the contrary, at the 25% MPR condition, the decrease rate of delay on general roads is slightly larger than in other conditions. This result is presumed to be because some general vehicles traveling on the road corresponding to the added length of green signal time benefited from receiving signal priority. The effect of the BRT signal priority on general roads is not substantial because the proposed CWR signal control strategy is designed to first make a decision on the signal control based on the traffic information of general roads before making a decision on the BRT signal priority. Therefore, it is inferred that the BRT signal priority in the CWR-based strategy in this study can have a positive effect in terms of the operation of the entire roads because it does not have a significant impact on general roads while reducing the travel time of BRT vehicles.
(a)
(b)
(c)
(d)
However, as discussed in Section 4.2.1, there was no clear difference in the pattern of delay changes on general roads according to the maximum allowed time for priority control. Furthermore, no clear pattern differences were observed in the MPR condition of V2X communication vehicles. This shows that within the scope of the experimental environment in this study, the effect of the BRT signal priority is not sensitive to the maximum allowed time for priority control or the level of travel information collection of surrounding vehicles. This phenomenon requires more in-depth research by setting up another experimental environment in future studies.
5. Conclusion
This paper proposed an intersection signal control strategy using V2X communication data and analyzed its effects based on a simulation according to the level of the MPR of V2X communication vehicles. The CWR signal control strategy proposed in this study applies a real-time control method that predicts the intersection arrival time of vehicle queues on individual lanes using V2X communication data and immediately stops the green signal of the current signal phase when the queue decreases and disappears, and then applies the green signal in the next signal phase. The BRT vehicle priority signal feature is additionally applied to this strategy.
On the aspect of traffic improvements regarding intersection flow and average vehicle delay, the general intersection, BRT route intersection, and BRT station intersection showed different quantitative values of the effects of CWR signal control owing to the existing traffic flow levels. However, the effect increase and decrease pattern commonly showed similar results regardless of the intersection type. The CWR signal control strategy showed a certain level of effect at 25% or lower MPRs and much larger effects at 50% or higher MPRs, but the effects were similar at 50–100% MPRs. The results at individual intersections were reflected in the evaluation of the effect of all the target areas of evaluation, and the highest improvement effect on the traffic of the entire road network could be obtained at 50% or higher MPR conditions. This suggests that the signal control strategy proposed in this study can have a positive effect when the MRP of the autonomous cooperative driving vehicles is at least 25%, and a larger effect can be achieved if the MPR is higher than 50%. These findings show the promising performance of the CWR signal control strategy proposed in this study. Furthermore, it can be estimated from these findings that even if the MPR of autonomous cooperative driving vehicles is only around 50%, real-time road traffic information collected based on them can represent actual road conditions. This is a subject for research that is worth investigating in the future.
The application of the BRT signal priority decreased the travel time of BRT vehicles in most conditions. Moreover, the travel time increase effect increased with the MPR. This implies that the effect of signal priority increases in general as the accuracy of information increases owing to the high collection rate of vehicle travel information. Furthermore, it was found that when the BRT signal priority was applied, the effect on general roads was not substantial. This suggests that the proposed CWR signal control strategy shows reasonable performance in the application of the BRT signal priority.
In this study, however, there was no clear difference in the pattern of vehicle delay changes on general roads according to the maximum allowed time for priority control, and no clear difference in the pattern was observed regarding the MPR conditions of V2X communication vehicles. This shows that within the scope of the experimental environment in this study, the effect of the BRT signal priority is not much sensitive to the maximum allowed time for priority control or the level of travel information collection of surrounding vehicles. This phenomenon requires more in-depth analysis by constructing a different experimental environment in future research. In addition, the optimization problem for the maximum allowed time for priority control, which was proposed in this study, is a subject that is worth investigating in future research.
The CWR signal control proposed in this study showed favorable performance through various analyses, but it still has several limitations. In this study, the location information of vehicles that cannot communicate was omitted because only V2X communication data were used to estimate the vehicle queues on individual lanes on the road. Therefore, the estimation accuracy for the actual length of vehicle queues is inevitably low in an environment with a low MPR for V2X communication vehicles. The results of this study also showed relatively small effects of signal control when the MRP condition was low. Hence, attempts should be made in future research to utilize various statistical methods to increase the accuracy of queue estimation only with collected communication information, even in an environment with less than 100% MPR.
Furthermore, the CWR signal control algorithm proposed in this study sets the maximum green signal time of individual signal phases as a constant, assumed at 60 s for this experiment. However, this can be set at various values for realistic operation aspects of signal control including the environment of the surrounding road infrastructure. In the future, the maximum green signal time can be set as a variable and an optimization problem for this variable can be handled. The dwell time of passengers at the station in the BRT signal priority also needs to be processed as a variable instead of a constant. This involves making predictions based on the number of passengers waiting at a station. This requires not only V2X communication but also information collection and a linkage system for the situation of individual stations in the future, which is a topic worth investigating in follow-up research.
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 there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
This work was supported by the Technology Innovation Program (20018101, development on automated driving with perceptual prediction based on T-Car/vehicle parts to intelligent control/system integration for assessment) funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea).