Evaluation of a Dynamic Lane Assignment on Urban Freeways by Using a Microscopic Traffic Flow Simulation

Abstract

At a merging section of freeways, the mainline traffic flow conflicts with the traffic flow from an onramp, which often causes traffic congestion. To mitigate this conflict, it is important to dynamically manage the lane utilization on the mainline at bottleneck points. Recently, connected and automated vehicles (CAVs) are expected to be widely used on freeways. Guiding CAVs to change lane to avoid conflicts with merging vehicles may alleviate traffic congestion. In this paper, we propose a novel dynamic lane guidance system that can dynamically manage the lane changes in the merging sections with urban freeway ramps. In the proposed system, CAVs are guided to change lane at the lanes upstream of bottlenecks according to traffic conditions, which are detected by the following two methods: 1) cross-sectional vehicle speed information obtained from traffic detectors and 2) vehicle speed information obtained from driving trajectories based on probe data. Then, the effectiveness of the proposed dynamic lane guidance system is evaluated by a microscopic traffic simulator, Vissim, which is calibrated to reproduce the observed traffic states based on full vehicle trajectory data collected at the section near the Tsukamoto merge area on the Hanshin Expressway Route 11 Ikeda Line in Japan. As a result, the obtained results showed that dynamic lane guidance is effective in reducing overall travel time, thus demonstrating the usefulness of dynamic lane guidance. It was also found that traffic control based on partial vehicle information, such as probe data, is sufficiently effective.

Keywords

operations freeway management freeway traffic control

Traffic congestion on highways is serious and causes many economic, social, and environmental losses every year ( 1 ). Traffic congestion is not only a problem in urban areas but a nationwide problem. As a result, it is vital to eliminate or alleviate traffic congestion. In general, congestion is caused when traffic demand exceeds traffic capacity ( 2 , 3 ). To address these problems, measures such as lane expansion have been used to increase traffic capacity; however, these measures are considered difficult because of the huge budgets and time required. Therefore, some dynamic traffic management processes concerned with traffic congestion on freeways have been studied, evaluated, and analyzed to facilitate traffic flow in response to traffic conditions, which are everchanging. Such processes include lane-use optimization and variable speed control ( 4 , 5 ). In a traffic flow analysis on conventional lane-use management, based on data obtained from traffic detectors, Shiomi et al. stated that giving lane change instructions to identified vehicles can be expected to improve speed at bottlenecks ( 6 ). Park et al. proposed a traffic management strategy to mitigate the complications between the main freeway lanes and the merging lanes by issuing a lane change advisory upstream of the merge point to reduce the slowdown of the main lanes and the merging vehicles ( 7 ). Hattori et al. found that, in urban freeway ramp merging sections, increasing the utilization of the inner lane upstream of bottlenecks during critical flow conditions reduces the overall travel time ( 8 ). Sharma et al. designed a feedback/feedforward control law based on a linear–quadratic regulator to improve traffic flow by guiding lane control for individual vehicles on a section including a merge ( 9 ). Then, they evaluated the effectiveness of the proposed law through a microscopic traffic flow simulation. The results showed that the proposed lane change support system can suppress the shock waves (SWs) of traffic flow and significantly reduce traffic congestion. Amini et al. used the connected and automated vehicles (CAVs) technology to construct a trajectory optimization system for freeway fold-in merging sections and evaluated the system’s effectiveness through a traffic flow simulation ( 10 ). The results showed that the system enables vehicles to move as planned, and the system was effective in increasing traffic capacity and reducing travel time. Hu et al. proposed an algorithm to optimize both lane flow distribution adjustment and merging behavior control upstream of a merge using CAV technology to mitigate traffic congestion at highway merging sections ( 11 ).

Considering the future development of CAVs, it is technically possible to encourage individual vehicles to change their driving behaviors in response to traffic conditions. It is technically possible to avoid or alleviate traffic congestion by intervening in traffic conditions. There are two possible methods for acquiring information about traffic conditions. The first one is fixed-point observation using traffic detectors managed by the infrastructure side, and traffic volumes and vehicle speeds can be acquired through this method. The second method entails observation using data, such as probe data, that can acquire the driving history and behavior history of individual vehicles at any time. In a study on traffic management using probe data, Khairul et al. estimated traffic flows by combining traffic speed information obtained from probe vehicles and fundamental diagrams and found that flow estimation is feasible ( 12 ). ETC2.0 probe data includes the driving and behavior history outputs of the probe information utilization system, which was constructed by the National Institute of Land and Infrastructure Management and is used to prevent traffic accidents and improve the efficiency of transportation and logistics. In implementing these lane guidance systems, the difference in the control effect depending on the method of acquiring information about traffic conditions is important for the operators who operate traffic systems and provide services, as it is related to the cost of installing various equipment.

Thus, under certain traffic conditions, lane guidance is expected to improve speeds at bottlenecks and shorten section travel times. In the past, microscopic traffic flow simulations have been used for analysis and verification; however, there is no evidence that lane guidance control is a phenomenon. The evaluation of data acquisition methods for traffic conditions is important because the system operation must be optimized in a cost-effective manner. Therefore, it is necessary to verify dynamic traffic management under mixed CAV conditions. In introducing these lane guidance systems, the difference in the control effects depending on the method of acquiring traffic conditions is related to the cost of installing the equipment, which is important for the operators who operate traffic systems and provide services. In addition, many studies on dynamic traffic management have been conducted using simulations.

In this study, based on the results of Hattori et al. who showed that increasing the inner lane utilization rate upstream of bottlenecks improves congestion at bottlenecks under certain traffic conditions, a dynamic lane management system is proposed to improve traffic conditions by guiding the lanes upstream of bottlenecks ( 8 ). The effectiveness of the proposed dynamic lane management system was discussed and then evaluated using microscopic traffic flow simulations. Specifically, the simulations reproduced current conditions based on vehicle trajectory data collected at the section near the Tsukamoto merge area on the Hanshin Expressway Route 11 Ikeda Line in Japan. The traffic conditions were obtained using two methods: 1) cross-sectional vehicle speed information obtained from traffic detectors and 2) vehicle speed information obtained from driving trajectories based on probe data.

Methods

In this section, the target section to be simulated and data collected on the section for the simulation calibration are described. Then, an overview of the simulation and lane management system is presented.

Data

In this study, we used the Zen Traffic Data (ZTD) of the Hanshin Expressway Company, Ltd for the area near the Tsukamoto merging section (approximately 2 km) on the uphill section of the Route 11 Ikeda Line of the Hanshin Expressway ( 13 ). Multiple cameras are installed on the road lighting poles along the entire Hanshin Expressway, and the vehicle trajectory data of all vehicles traveling along the installation section is converted into vehicle trajectory data at 0.1 s intervals by image sensing using deep learning ( 14 ). An overview of the traffic dataset and data items included in the vehicle trajectory data is shown in Tables 1 and 2. Five sets of 1 h data were provided for the target section, and Data 1 (7:00–8:00), the busiest time of the day, was used in this study.

Table 1.

Data Components of Zen Traffic Data (ZTD)

Data type	Data outline	File format
Vehicle trajectory database	Vehicle trajectory data, vehicle attribute information, road surface alignment information	csv
Vehicle trajectory video	Video file with dots representing the location of each vehicle on the map	mp4
Vehicle detector pulse data	Pulse data of the target interval	csv

Table 2.

Data Layout

Data layout information	Contents
Common information	Vehicle ID
Common information	Time
Main unit information	Velocity
	Lane nunber
	Latitude
	Longitude
	Kilo-post

Target Section

The route map, the lane layout, and the traffic states of the target section of ZTD are illustrated in Figure 1, a to d . The target section (Route 11 Ikeda Line) is an important route connecting Osaka Airport, the Meishin Expressway, the Chugoku Expressway, and other national trunk roads in Osaka City, and it has a complex road environment with S-curves, sags, a Tsukamoto merging section, and a two-lane section with very heavy traffic. The 4.2 kilo-post (KP) is the beginning of the lane-change prohibited zone, the 3.8 KP is the merged zone, and the 3.5 KP is the end of the lane-change prohibited zone. As seen in Figure 1, c and d , which are time-space diagrams of vehicle trajectories colored by their driving speed, the traffic congestion occurred within the target section, and SWs are generated around of the 3.8 KP merging point in both outer and inner lanes.

Figure 1.

Layout of the target section and current states of traffic congestion: (a) route map of the section, (b) lane layout, (c) observed vehicle trajectories in the outer lane, and (d) observed vehicle trajectories in the inner lane.

Development of the Simulation Model

In this study, the microscopic traffic flow simulator Vissim, which was manufactured by PTV, was used to simulate the target section. Vissim is a microscopic traffic flow simulator developed by modeling the operation of urban and public transportation, and it is used to simulate highways and intersection signals.

For highway simulations, it is possible to change the characteristics of driving behaviors by using a vehicle model with freely configurable parameters and vehicle types. In addition to basic attributes, such as desired speed, acceleration, and deceleration behavior, it is also possible to calibrate car-following behavior and lane-change behavior. There is a function called component object model (COM) interface that allows Vissim to be controlled by external scripts (e.g., Python). In this study, we set up a scenario in which a vehicle, assumed to be a CAV, is guided from the inner lane to the outer lane in a specific traffic condition, thus utilizing the COM interface.

Model Setup

In this study, various parameters related to the desired speed, car-following, and lane-changing behaviors were obtained and adjusted using the ZTD dataset to construct a model that reproduces the observed traffic states. In the ZTD dataset, the vehicles that change lanes even in the lane-change-prohibited section from 4.2 KP to 3.5 KP were observed, which it is not possible to reproduce using the normal settings of Vissim. Thus, some vehicle types which would change lane at any location were individually set to reproduce the observed vehicle behaviors. The details of vehicle type settings are described later. In the simulation, the total hourly incoming traffic volume was fixed at 3,200 vehicles per hour, regardless of the random seed, although there was some random variation in the time headways at the entrance of the target section.

Validation of the Simulations

In this study, simulations were run with 12 different random seeds and compared with the observed values for three cross-sections (4.1 KP, 3.8 KP, and 3.5 KP) obtained from ZTD. As an index showing the accuracy of the simulation, the GEH statistic of the traffic volume on the outer lane, inner lane, and both lanes were used. The GEH is expressed by Equation 1 as follows:

GEH = \sqrt{\frac{2 {(M - C)}^{2}}{(M + C)}}

(1)

where

M = model-estimated volume, and

C = field count.

The GEH statistic is a kind of a standardization method of the error values and can be applied to the wide range of traffic volume so that it is often used to validate the reproducibility of current conditions in microscopic traffic flow simulations ( 15 – 17 ). If it is less than 5.0, the model-estimated volume fits field counts, while if it is greater than 10.0, there is a problem with the simulation ( 18 ). Kojima et al. used model-estimated volume and field count cross-sectional traffic volumes on a two-lane section of a freeway to determine reproducibility ( 17 ). Therefore, in this study, the 5 min traffic volume of outer lane, inner lane, and two-lane traffic were used as indices for each of the three cross sections, that is, 4.1 KP, 3.8 KP, and 3.5 KP, respectively.

As a result, we confirmed that SW occurred from 3.8 KP and that no SW other than the bottleneck occurred. The correlation coefficients between the 1 h average GEH statistic and the inner lane utilization at each measurement location are shown in Table 3. The GEH values for each random seed and each cross-section were less than 5.0, indicating good reproducibility of the current situation, and the correlation coefficient for the inner lane flow ratio was high. The other 10 random number seeds were found to be applicable to future research, as they showed good reproduction of the current situation, and the correlation coefficient for the inner lane flow ratio was high. Figure 2, a and b , shows the KV relationship, traffic density, and inner lane flow ratio at 4.1 KP for each of the 10 random seeds. Overall, it was confirmed that the KV relationship and inner lane flow ratio can be generally reproduced.

Table 3.

Reproduction Results

Point	4.1 KP				3.8 KP				3.5 KP				Shock wave
Random seed	Outer lane GEH	Inner lane GEH	Total GEH	Inner lane correlation coefficients	Outer lane GEH	Inner lane GEH	Total GEH	Inner lane correlation coefficients	Outer lane GEH	Inner lane GEH	Total GEH	Inner lane correlation coefficients
R54	0.59	0.80	0.31	0.71	0.49	0.72	0.24	0.78	0.71	0.77	0.25	0.71	○
R55	0.72	0.86	0.23	0.83	0.62	0.63	0.21	0.89	0.62	0.53	0.18	0.81	×
R56	0.73	0.64	0.19	0.93	0.74	0.57	0.17	0.95	0.86	0.61	0.25	0.79	○
R57	0.54	0.81	0.20	0.83	0.54	0.61	0.23	0.86	0.48	0.54	0.21	0.78	○
R58	0.83	0.72	0.24	0.83	0.63	0.58	0.14	0.93	0.65	0.58	0.17	0.78	○
R59	0.56	0.87	0.17	0.89	0.46	0.78	0.15	0.91	0.80	0.73	0.27	0.73	○
R60	0.85	0.65	0.24	0.85	0.68	0.74	0.19	0.87	0.76	0.66	0.20	0.84	×
R61	0.67	0.60	0.25	0.88	0.58	0.52	0.21	0.90	0.80	0.60	0.25	0.78	○
R62	0.59	0.80	0.27	0.82	0.54	0.66	0.19	0.92	0.64	0.61	0.19	0.82	○
R66	0.76	0.76	0.24	0.84	0.60	0.65	0.21	0.86	0.63	0.75	0.25	0.76	○
R69	0.65	0.63	0.25	0.84	0.50	0.52	0.21	0.87	0.68	0.51	0.28	0.70	○
R73	0.65	0.75	0.24	0.82	0.66	0.59	0.24	0.81	0.88	0.53	0.26	0.72	○

Note: KP = kilo-post; ○ = Shock wave are occurred from 3.8KP; × = Shock wave are not occurred from 3.8KP.

Figure 2.

Reproduction results (4.1 KP): (a) k–v curve and (b) k–inner lane flow ratio relationship.

Table 4 summarizes the number of lane changes by direction in the target section both for the observation (mentioned as ZTD in the table) and simulations with the different random seeds. It can be seen that the number of lane changes in the lane-change-prohibited section is generally consistent with the ZTD, while the number of lane changes in the no-prohibited section tends to be slightly overestimated. Overall, we can conclude that the simulation model well represents the lane-change behaviors in the target section.

Table 4.

Number of Lane Changes in the Target Section

	Lane-change-prohibited section		No-prohibited section
Random seed	Outer→inner	Inner→outer	Outer→inner	Inner→outer
Zen Traffic Data	75	21	319	211
R54	63	22	351	236
R57	70	35	353	230
R69	80	43	359	238
R56	78	32	384	267
R58	66	28	353	236
R59	59	24	346	238
R61	63	27	361	239
R62	61	23	383	246
R66	70	25	350	226
R73	74	16	328	241

Evaluation of the Proposed Lane Management System

Hattori et al., in a statistical analysis of ZTD, revealed that, at urban freeway ramp merging sections, an increase in the inner lane utilization at the upstream point of bottlenecks during traffic conditions of 40–65 km/h has the effect of reducing the overall travel time ( 7 ). Therefore, we propose a traffic flow management strategy that dynamically guides lane changes in response to traffic conditions near urban freeway merging points. Specifically, we propose a lane guidance system that directs CAVs at the upstream section of measurement points to change lanes from outer lane to inner lane when the travel speed at a measurement point reaches a predetermined speed threshold value. A schematic showing the proposed solutions is shown in Figure 3, a and b . Here, two methods for measuring the traffic states were assumed; one is by using a traffic detector and the other is by travel time collected by CAV probe data; the differences between them will be investigated in the following section.

Figure 3.

A schematic of the proposed dynamic lane change management: (a) traffic detector data and (b) probe data.

Lane Management System Setup

This section describes: 1) assumptions about CAVs, 2) the speed range of the lane guidance system, 3) the vehicles to be controlled, and 4) the control section.

Assumptions for CAVs

The CAVs assumed in this study travel the same as the other human-driven vehicles, except in the controlled section, while, if the pre-fixed conditions of average speed at a measurement point or travel time in a measurement section are satisfied, they receive an instruction to change lanes from a roadside unit in the controlled area, and change the lane if there is a enough space in an adjacent lane. CAVs drive independently from the other CAVs, and they do not cooperate with each other.

Control Implementation Speed Region

A statistical analysis of the Hanshin Expressway ZTD revealed that the overall travel time decreases with the increase in the inner lane flow ratio at the upstream point of bottlenecks during traffic conditions of 40–65 km/h. Based on these findings, a control speed range of 40–60 km/h was set for the simulation. The data was compiled using 3.8 KP two-lane average vehicle speed measurements based on traffic detector data, which is permanently available on freeways, and the average vehicle speeds within a certain time were obtained from vehicle travel histories (CAVs in this case), such as ETC2.0 probe data. The flowcharts of lane guidance systems control based on traffic conditions are shown in Figure 4, a and b .

Figure 4.

Flowcharts of the traffic detector data and probe data (4.1 KP): (a) traffic detector data and (b) probe data.

Vehicles to Be Controlled

The vehicles that respond to the lane guidance system are inputs. In actual traffic flows, a vehicle is assumed to be capable of optimal driving based on information, such as a CAV. In the simulation, the lane change types were set as shown in Table 5; Type-D vehicles were assumed to be the vehicles to be controlled, and the mixing ratio of the vehicles to be controlled (1%–10%) was set in the vehicle composition.

Table 5.

Vehicle Composition

	Type-A	Type-B	Type-C	Type-D
Lane-change-prohibited section	×	×	○	×
Other	×	○	○	○

Note: ○ = setting; × = no setting.

Control Section

A control interval was set to encourage lane changes in the control speed range for the vehicles responding to the lane guidance system. To adjust the lane flow ratio at 4.1 KP, lane guidance had to be performed upstream of the lane-change-prohibited section. Therefore, the control section to encourage lane changes was set to 162.145 m upstream of the beginning of the lane-change-prohibited section (4.2 KP). As a specific setting in the simulation, the Vissim road has a setting called “blocked vehicle classes.” On a two-lane road, any vehicle on the outer lane can change lanes on a link if any vehicle on outer lane is set to “blocked vehicle classes” for the travel lane. Whenever any vehicle on the outer lane changes lanes on a two-lane road, it changes lanes to the inner lane if possible. In this study, when a vehicle is within the speed range of the control implementation on the control section, the vehicle class of the vehicle to be controlled is set in Classes, and the vehicle to be controlled is encouraged to change lanes to the inner lane. The control method utilizes Vissim’s COM interface.

Results

Simulations were performed by changing the mixing ratio of the vehicles to be controlled from 1% to 10%. The results of the per-vehicle lost time for each of the 10 random seed simulations in Table 3 are shown in Figure 5, a and b . Figure 5 shows the relationship between the penetration rate of CAVs and the loss time of the case, using traffic detectors and probe data, respectively. The loss time is defined as the difference between the simulated average travel time and reference travel time (time required for free travel) for the vehicles with average travel speeds of 60 km/h or less from 4.1 KP to 3.5 KP. For the reference travel time, the average travel time for the vehicles with an average travel speed of 60 km/h or higher from 4.1 KP to 3.5 KP was used. First, by focusing on the outer lane, the loss time tended to decrease with the increase in the mixing rate. In the inner lane, it can be confirmed that the loss time was almost unchanged. Next, the relationship between 5 min traffic volume and average speed at 3.5 KP is shown in Figure 6, a to d . It can be found that both cases of traffic detector and the probe data show an increase in speed in the outer lane and an increase in the maximum traffic volume in the inner lane, regardless of the data collection methods (see Figure 6, a to d ). The travel time per vehicle per lane in the 4.1 KP to 3.5 KP interval, the travel time of the vehicle, and the number of lane guidance trips are shown in Figure 7, a to d . The right axis corresponds to the histogram, which shows the number of lane changes by control.

Figure 5.

Relationship between the lane flow ratio and loss time: (a) outer lane and (b) inner lane.Note: CAV = connected and automated vehicle.

Figure 6.

Relationship between 5 min traffic volume and average speed: (a) outer lane (traffic detector), (b) inner lane (traffic detector), (c) outer lane (probe) and (d) inner lane (probe).

Figure 7.

Relationship between the travel time and number of lane changes (R58): (a) outer lane (with control using traffic detector), (b) inner lane (with control using traffic detector), (c) outer lane (with control using probe) and (d) inner lane (with control using probe).

This confirms the travel time decreased in the outer lane, while, in the inner lane, there was no significant change in the travel time by the proposed lane management system, indicating that the overall effect of reducing the travel time for both data collection methods. To see the detailed impact of the lane management system on the traffic flow, the time-space diagram of vehicle trajectories in the outer lane without and with control of both cases using traffic detectors and probe data are shown in Figure 8, in which the pink dots indicate the position and time of the lane change, and each trajectory is colored by the speed of a vehicle. It is clearly seen that in the cases of with control, the generations of the SWs are controlled the stop and go waves are cut-off by both measurement methods, a traffic detector and probe data.

Figure 8.

Time space diagram: (a) without control, (b) with control using traffic detector, and (c) with control using probe data.

Conclusions

As the number of vehicles responding to the lane guidance system increased, the loss time for the vehicles on the outer lane decreased. However, there was no significant increase in the loss time on the inner lane. From the above results, it can be confirmed that the loss time was reduced by the lane guidance system. The relationship between the travel time and number of lane changes shows that the travel time decreased significantly after lane guidance. This suggests that lane guidance upstream in response to traffic conditions downstream can prevent the occurrence of traffic congestion. Moreover, the trajectory diagrams show that the SW propagation was shortened. Overall, the above-mentioned results indicate that upstream lane guidance based on the traffic conditions near the heads of bottlenecks is effective in reducing or eliminating congestion.

The results of the two methods for acquiring information about traffic conditions showed no significant differences in the data collection methods, indicating that the limited data obtained from the trajectory of a specific section of CAVs was sufficient to improve or alleviate traffic congestion. Overall, these results provide important insights into traffic management from an economic perspective.

Overall, it can be concluded that, depending on traffic conditions, the proposed lane management system near the upstream of bottlenecks can reduce travel time for entire sections. It was also found that traffic control based on partial vehicle information, such as probe data, is sufficiently effective. However, since the effectiveness of lane guidance varies depending on on many factors, more detailed analyses are needed. Specifically, a sensitivity analysis could be carried out of the predetermined speed zone value, the location and length of the control section where CAVs are directed to change the lane, and the section and point where traffic state is monitored. In addition, since the section covered in this study is a special environment including a lane-change-prohibited section, it is necessary to conduct evaluations and analyses in other sections to confirm the generality of the results. Based on these results, we plan to conduct a detailed analysis and change the target section in future works.

Footnotes

Acknowledgements

The vehicle trajectory data used in this study were provided by Hanshin Expressway Co. Ltd. The authors would like to thank Enago () for the English language review.

Author Contributions

The authors confirm contributions to this paper as follows: study conception and design: Y. Shiomi, Y. Hattori; data collection: Y. Hattori; analysis and interpretation of results: Y. Shiomi, Y. Hattori; draft manuscript preparation: Y. Hattori. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: JSPS Grant-in-Aid for Scientific Research 19H02268 and 20KK0334.

ORCID iDs

Yuya Hattori

Yasuhiro Shiomi

Data Accessibility Statement

The data used to support the findings of this study are available from the corresponding author on request.

References

Ministry of Land, Infrastructure, Transport and Tourism in Japan. Ministry of Land, Infrastructure, Transport and Tourism Productivity Revolution Project. https://www.mlit.go.jp/common/001123977.pdf. Accessed July22. 2022.

Koshi

Kuwahara

Akahane

A Study on Congestion Phenomena at Tunnels and Sags on Moterways. Journal of JSCE, Vol. 18, No. 458, 1993, pp.65–71.

Kurihata

Hioki

Measure for Improvement of Traffic Capacity at Bottlenecks on Motorways. Journal of JSCE, Vol. 12, 1995, pp.731–738. (in Japanese).

Xing

Muramatsu

Harayama

Balance Lane Use with VMS to Mitigate Motorway Traffic Congestion. International Journal of Intelligent Transportation Systems Research, Vol. 12, No. 1, 2014, pp.26–35.

Shiomi

Dynamic Speed Management on Motorways. International Association of Traffic and Safety Science Review, Vol. 45, No. 3, 2021, pp.172–181. (in Japanese).

Shiomi

Controlling Lane Traffic Flow for Managing Uncertainty in Traffic Breakdown. Transportation Research Board 95th Annual Meeting Compendium of Papers, report number 16-6706, Washington, D.C., 2016.

Park

Sarma

Smith

Development and Evaluation of Enhanced IntelliDrive-Enabled Lane Changing Advisory Algorithm to Address Freeway Merge Conflict. Transport Research Record: Journal of the Transportation Research Board, 2011. 2243: 146–157.

Yuya

Yasuhiro

On the Relationship Between Lane Flow Ratio and Travel Time on the Merging Section of an Urban Freeway. JSTE Journal of Traffic Engineering, Vol. 8, No. 2, 2022, pp. A159–A168.

Sharma

Papamichail

Nadi

van Lint

Tavasszy

Snelder

A Multi-Class Lane-Changing Advisory System for Freeway Merging Sections. IFAC-PapersOnline, Vol. 54, No. 2, 2021, pp. 93–98.

10.

Amini

Omidvar

Elefteriadou

Optimizing Operations at Freeway Weaves with Connected and Automated Vehicles. Transportation Research Part C: Emerging Technologies, Vol. 126, 2021, p. 103072.

11.

Sun

Trajectory Optimization of Connected and Autonomous Vehicles at a Multilane Freeway Merging Area. Transportation Research Part C: Emerging Technologies, Vol. 101, 2019, pp. 111–125.

12.

Khairul

Filomon

Mecit

Estimating Traffic Flow Rate on Freeways from Probe Vehicle Data and Fundamental Diagram. Proc., IEEE 18th International Conference on Intelligent Transportation Systems, Gran Canaria, Spain, IEEE, NY, 2015, 2921–2926. http://dx.doi.org/10.1109/ITSC.2015.468.

13.

Hanshin Expressway Co. Ltd., Features of Zen Traffic Data. https://zen-traffic-data.net/english/. Accessed November 18, 2022.

14.

Seo

Tago

Shinkai

Nakanishi

Tanabe

Ushirogochi

Kanamori

, et al. Evaluation of Large Scale Complete Vehicle Trajectories Dataset on Two Kilometers Highway Segment for One Hour Duration: Zen Traffic Data. Proc., International Symposium on Transportation Data and Modelling (ISTDM2021), Washington, D.C., 2021.

15.

Ciuffo

Punzo

Torrieri

Comparison of Simulation-Based and Model-Based Calibrations of Traffic-Flow Microsimulation Models. Transport Research Record: Journal of the Transportation Research Board, 2008. 2088: 36–44.

16.

Fan

Calibration of Microscopic Traffic Simulation Models Using Metaheuristic Algorithms. International Journal of Transportation Science and Technology, Vol. 6, 2017, pp.63–77.

17.

Kojima

Muromachi

The Effectiveness of Traffic Control Measures with Real-Time Crash Prediction Model for Urban Expressway Safety. JSTE Journal of Traffic Engineering, Vol. 4, No. 1, 2018, pp. B52–B56. (in Japanese).

18.

Holm

Tomich

Sloboden

Lowranc

Traffic Analysis Toolbox Volume iv: Guidelines for Applying CORSIM Microsimulation Modeling Software. Report No. FHWA-HOP-07-079. Federal Highway Administration, Washington, D.C., 2007.