1. Introduction
Technological advancements are occurring at an unprecedented rate, profoundly impacting every aspect of daily life. The automotive sector is one such area and, particularly, progress in fields such as artificial intelligence (AI), machine learning (ML), sensor technologies, and autonomous systems has paved the way for the development of AVs, which can lead to profound changes in the transportation systems. With these vehicles, drivers have become acquainted with various technologies such as adaptive cruise control (ACC), lane keeping, and autonomous driving. These technological advancements offer opportunities to achieve goals such as reducing the negative impacts of driver behavior in traffic, utilizing infrastructure more efficiently, and transitioning to sustainable energy sources. Advancements in sensor technologies, the increasing complexity of AI algorithms, and the growth in computational power are creating a milestone with the potential to enhance the safety and efficiency of AVs, thereby reducing traffic accidents [
1,
2]. Given that human error is responsible for over 90% of traffic accidents, the significance of these developments becomes even more apparent [
3,
4].
AVs’ reaction times are expected to be significantly shorter than those of human drivers due to their reliance on communication technologies [
5]. Consequently, these vehicles will be capable of maintaining shorter time headways, enabling them to follow each other more closely and safely. This will increase road capacity and substantially reduce traffic accidents [
6]. However, until all vehicles on the roads become fully autonomous, both traditional human-driven vehicles and AVs will coexist within the same traffic flow [
7]. While AVs are predicted to fully dominate traffic as early as 2050 [
8], some researchers suggest that this scenario might extend until 2070 [
9]. Therefore, it is anticipated that human-driven vehicles and AVs will coexist in mixed traffic conditions for an extended period.
Human drivers inherently possess the potential to exhibit unpredictable behaviors while driving. These behaviors may arise from a variety of factors, including aggressive driving, lack of sleep, alcohol use, inexperienced driving, and old age. The data collected by AVs regarding road structure and environmental conditions allow them to respond in real time. Furthermore, the decisions made by AVs based on specific algorithms, combined with the unpredictable behaviors of human drivers in mixed traffic conditions, create a high potential for adverse situations affecting traffic performance, safety, and sustainability. Given that AV technology is still in the developmental stage and that different levels of automation exist, the complexity of this interaction is further increased.
The penetration rate of AVs, i.e., the ratio of AVs to the total number of vehicles, is a critical factor shaping this interaction. As the penetration rate increases, the quantity and nature of interactions between autonomous and conventional vehicles change. This shift has the potential to lead to unpredictable outcomes in mixed traffic conditions [
10,
11]. For instance, scenarios in which human errors contribute to uncertainties in interactions among non-AV drivers may constrain the potential benefits of integrating AVs and adversely impact traffic flow. Moreover, an anticipated increase in the penetration rate of AVs is likely to correlate with a rise in travel demand. An increased presence of AVs might change human travel preferences, thereby influencing the capacity of transportation systems. Fundamentally, variations in penetration rates modify the interaction dynamics between vehicles, thereby affecting the performance of transportation systems. When a particular penetration rate threshold is exceeded in mixed traffic conditions, it is expected that sustainability indices might exhibit negative results. Consequently, the importance of transportation management systems and decision-making processes informed by penetration rates will increase. Within this framework, investigating the implications of changes in penetration rates and vehicle interactions on sustainability indices has been recognized as a gap in the literature and a critical area requiring further exploration.
Intelligent Transportation Systems (ITS) in modern cities are built upon the essential roles of transportation planning and traffic network management. Traffic network management includes models that control vehicle traffic to maintain acceptable levels of traffic flow. One of the primary objectives of this study is to examine and enhance system changes by employing network management models on main arterial flows. These improvements aim to achieve a homogeneous traffic flow, prevent congestion, and mitigate or delay queue formation and shock waves. In this study, investigations have been conducted on ToC, which describe the control transition processes of drivers and AVs, and DL applications, which are specific lanes allocated for certain types of vehicles or purposes, within an open-source urban mobility environment. This research aims to analyze the effects of ToC and DL applications on traffic flow under different critical penetration rates.
2. Literature Review
The integration of AVs into transportation systems has attracted considerable academic attention, leading to an extensive body of literature exploring their potential impacts. A key focus of these studies is how network management systems can reshape traffic dynamics, enhance safety, and improve the overall efficiency of transportation networks. Additionally, considerable attention has been given to the interaction between AVs and human-driven vehicles, particularly in mixed traffic conditions where diverse behaviors and decision-making processes coexist.
Studies in the literature have investigated the anticipated impacts of AVs on mixed traffic conditions across various penetration levels. Shladover et al. [
12] examined the impact of penetration variability of vehicles equipped with ACC and co-operative adaptive cruise control (CACC) systems on highway capacity. They observed that CACC can significantly increase capacity in scenarios with a penetration rate of 40% and above. It was stated that, upon reaching this penetration level, an environment is created where vehicles can be safely followed at shorter gaps due to their high dynamic response capabilities. It was demonstrated that the maximum lane capacity is around 4000 vehicles per hour when all vehicles are equipped with CACC. Arnaout and Arnaout [
13] demonstrated that, in low-volume traffic scenarios, varying penetration rates have no significant impact. However, in high-volume scenarios above 40%, significant improvements in traffic performance were observed. Nevertheless, it was noted that it might not be possible to reach high penetration levels in the near future and, therefore, the implementation of special lanes such as High-Occupancy Vehicle (HOV) lanes should be examined to effectively use vehicles with CACC at low penetration levels.
Talebpour and Mahmassani [
14] examined the efficiency and fundamental diagrams of traffic flow, as well as the impact of AVs and Connected and Autonomous Vehicles (CAVs) in a simulation environment. They investigated two scenarios where AVs were used optionally and with lane restrictions. They concluded that the performance of AVs increased in the optional use scenario, while overall performance decreased in the restricted lane section on a two-lane road section and a four-lane highway section. Additionally, it was determined that capacity values yielded significantly positive results at penetration rates above 20–30% for CAVs. Similarly, other academic studies have also shown that road capacity increases with the rise in CAV penetration rate and identified a 20–30% penetration rate as a critical threshold, beyond which road capacity gradually increases [
15,
16].
DLs for AVs and CAVs emerge as a crucial traffic management strategy, with the potential to both enhance efficiency and introduce new challenges in mixed traffic flows. This strategy contributes to lane-level homogeneity in traffic flow by allowing AVs to be separated from mixed traffic. Providing DLs for CAVs enables them to reveal their potential more rapidly, even within heterogeneous traffic flows with low penetration rates. Ye and Yamamoto [
17] identified CAV penetration and CAV performance as significant factors influencing the effectiveness of DLs. They also demonstrated that, at high penetration rates, DLs positively impact overall efficiency under conditions of high travel demand. Research suggests that, with higher market penetration, DLs decrease total travel time by providing a faster and more reliable travel alternative. Pourgholamali et al. [
18] examined the sustainable distribution of AVs in terms of social, economic, and environmental factors within a two-level optimization problem. Within their proposed framework, they formulated the locations, number, and pricing policies of DLs for AVs by considering sustainability indices. Literature findings broadly imply that sustainable planning of AV-DLs can contribute to making urban transportation systems both more efficient and equitable.
The controlled transition from autonomous driving to manual driving has been examined within the scope of the TransAID project, considering scenarios where AVs must be deactivated due to various reasons (such as system failures, unclear or missing lane markings, complex traffic situations, etc.), and has been integrated into the urban mobility environment SUMO. These controlled transitions are expected to impact traffic operations due to irregular vehicle behaviors and potential Minimum Risk Maneuvers (MRM) [
19]. This impact is anticipated to be exacerbated in environments with high vehicle interactions resulting from mixed traffic conditions. Although future AVs are designed to predict the intentions and imminent actions of other road users without connectivity [
20], developers of driverless technology foresee that it will not be possible to fully exempt the human driver from responsibilities for an extended period due to the limitations of vehicle automation in handling edge cases. Therefore, examining the effects of ToC on road traffic is of significant importance. Liu et al. [
21] conducted a comprehensive analysis of how various takeover times affect traffic flow stability in mixed traffic environments involving manual driving vehicles, fully automated driving vehicles (FADVs), and conditionally automated driving vehicles (CADVs). The study found that different takeover times significantly impact traffic flow stability. Specifically, a moderate takeover time of seven seconds allows drivers to adequately observe surrounding traffic conditions, thereby mitigating the adverse effects of takeover transitions and enhancing traffic flow safety. The research also demonstrated that increasing the total penetration rates of CADVs and FADVs or CADVs alone expands the traffic flow stability area. The stability improvements increase with higher penetration rates of both vehicle types. Noulis [
22] stated that, if the driver fails to regain control within a specified time, the ToC device initiates an MRM to ensure safety. This study highlighted the potential of DRL algorithms to improve traffic management in mixed autonomy scenarios by delaying the need for human intervention and optimizing traffic flow. This review aims to synthesize current findings and identify gaps in the literature, particularly regarding the implications of AV penetration rates and the efficacy of ToC and DL applications in enhancing traffic performance and sustainability. The novelty of this study lies in the detailed examination of ToC applications under specific traffic conditions, evaluating their impacts both locally and on an urban scale, with a particular focus on the presence of DLs. Additionally, the research explores the implications of different AV penetration rates on the effectiveness of ToC and DL strategies.
3. Background
3.1. Establishing the Study Area in an Urban Mobility Environment
In this study, the transportation network of İzmir province was established in an open-source urban mobility environment. İzmir city covers an area of 11,892 square kilometers and consists of a total of 30 districts, with 11 of them being central. The city is one of the urban centers accessible via road, rail, air, and sea transportation modes. In transportation planning, the study area is divided into subdivisions and the collected data are analyzed within these subdivisions. Since neighborhoods are the smallest administrative units in İzmir Province, the study area was analyzed based on Traffic Assignment Zones (TAZ). Additionally, within the boundaries of İzmir, certain zones were merged or divided considering the similarities in land use characteristics and socioeconomic status. Universities, organized industrial zones, airports, ports, shopping centers, and similar locations were designated as special attraction centers.
“OpenStreetMap” (OSM) serves as a map service that is easily and freely accessible to the public. The platform provides data on locations, roadways, transportation systems, and various points and areas of movement commonly used in daily life worldwide. In the simulations generated within the SUMO environment, data on lane width, the number of lanes, pedestrian crossings, highways, and numerous other real-world road geometric elements were obtained using OSM. TAZs and special zones, composed of neighborhoods covering the entire district of İzmir, were established in the urban mobility environment. İzmir urban transportation network was transferred to the SUMO environment through OSM, resulting in 68,782 nodes and 186,438 link data. A total of 807 TAZs were defined in the İzmir city center, and the demand matrix was prepared for use.
Figure 1 shows the TAZs created for the İzmir city transportation network, with those for the Buca district provided as an example.
In the context of organizing data related to İzmir’s transportation network, the travel demands from the Origin/Destination (O/D) matrix used in the 2015 İzmir Transportation Master Plan (UPİ2030) were first processed in the MATLAB (R2024a) environment [
23]. Based on the automation levels of AVs, open-source codes were developed according to the O/D data and prepared for use in the SUMO configuration (sumocfg) environment.
3.2. Modeling of Autonomous Vehicles
Modeling the impact of AVs on urban areas requires parameter values that reflect different levels of automation [
24,
25]. Vehicles fully controlled by the driver are classified as Level 0 (no automation), whereas those operating entirely without driver input are designated as Level 5 (full automation). Each level clearly defines the vehicle’s capabilities and the required degree of driver intervention. To provide a clearer understanding of these levels,
Table 1 below presents the definitions for the six levels of automation as defined by the Society of Automotive Engineers (SAE) International [
26]. Classifying AVs based on their automation levels not only provides a technical classification but also significantly influences the adoption of this technology, the development of legal regulations, and the assurance of safety [
27,
28].
Several calibration parameters are typically employed to model the differences between AVs and human-driven vehicles. These include the “tau” parameter, representing the minimum headway; the “sigma” parameter, accounting for driver imperfections; and the “speed factor”, pertaining to adherence to speed limits. For accurate modeling of AVs within designated transportation networks, it is necessary to define appropriate driving models for each vehicle category and to modify existing models accordingly. Within the scope of this research, studies on AVs from the literature were examined to identify common points and, in this context, SUMO car-following and lane-changing models were developed. AV parameters were determined based on academic studies in the literature, considering various future conditions and levels of automation. Commonly used AV parameters in the literature are presented in
Table 2.
The reason for the change in parameter values according to the levels of automation is that, as technology advances, it represents greater safety and efficiency in AVs. It can be assumed that the “sigma” parameter, representing driver error in IDM parameters, will significantly decrease as the level of automation increases [
29,
30,
31]. Additionally, higher levels of automation are expected to improve the consistency and predictability of vehicle behavior, thereby reducing the variability associated with human driving. As automation progresses, reliance on human judgment diminishes, leading to more uniform driving patterns and enhanced traffic flow. This evolution in parameter values is crucial not only for ensuring the accuracy of simulations but also for reflecting the true potential of AVs in real-world scenarios. Therefore, careful calibration of these parameters is essential for modeling the expected improvements in safety, efficiency, and overall system performance as AV technology advances.
3.3. Simulations in Urban Mobility Environment
The detailed evaluation of traditional and AVs in mixed traffic conditions based on the Sustainable Urban Mobility Indicators (SUMI) was conducted within the scope of the TÜBITAK project numbered 221M310. Within the project, over 300 simulations were conducted at various penetration levels of AVs to identify critical thresholds. In addition to modeling with multi-criteria decision-making (MCDM) techniques and decision support systems (DSS), AV penetration rates at which the system begins to be adversely affected in terms of sustainability and where preventive measures should be taken were identified. Numerous combinations, ranging from a scenario where all vehicles in the transport network are Level 0 to those where all vehicles are Level 5 automation, were examined in the urban mobility environment. As a result, the penetration rates established in the SCN-WI and SCN-WII scenarios generated the lowest sustainability indicator outcomes. SCN-PI and SCN-PII were identified as the scenarios where the system begins to yield negative sustainability outcomes, indicating a need for preventive measures.
Table 3 outlines the penetration rates for the four defined scenarios.
All simulations conducted on the Sioux Falls test network as part of the project were also performed on the urban transportation network of İzmir. The results obtained from the test network are similar to those from İzmir’s transportation network, which were based on data from UPİ2030. The consistency of the results highlights the reliability of the simulation approach, validating that the model accurately captures real-world dynamics in urban settings such as İzmir. Critical scenarios and penetration conditions produced identical results across the transportation network, reinforcing the validity of the findings. The study’s focus on İzmir ensures that the conclusions drawn are directly relevant to the city’s unique transportation challenges, offering practical solutions that can be implemented within the local context.
3.4. Transition of Control Mechanism
AVs may encounter scenarios that require the temporary deactivation of autonomous driving due to system errors, unclear or missing lane markings, or mixed traffic conditions. In these circumstances, facilitating an uninterrupted transition from autonomous to manual control is critical for protecting both driver safety and traffic stability. As part of the TransAID project, this situation was analyzed, and the urban mobility environment was adapted for use in SUMO simulations. These controlled transitions are anticipated to influence traffic operations due to irregular vehicle behaviors and the potential need for MRMs [
19,
32]. This situation is anticipated to become even more pronounced in environments with high vehicle interactions due to mixed traffic conditions. Although future AVs are being designed to predict the intentions and upcoming actions of other road users without direct communication [
20], developers of self-driving technology acknowledge that it will not be possible for a long time to completely exempt the human driver from his/her responsibilities, as vehicle automation cannot cope with boundary situations. Therefore, examining the impact of ToC on road traffic remains critically important.
The ability of vehicles to switch between manual and autonomous driving modes necessitates the integration of a mechanism into the modeling process that supports these transitions. For example, the capability to dynamically select different parameter sets for following or lane-changing models defined for vehicle types or to switch from one model to another is a fundamental requirement for such a model. To accurately represent control transitions, it is also essential to include elements that define the takeover process. In SUMO, a comprehensive model has been integrated into a vehicle device (ToC device) to depict these control transitions. This device not only manages the transitions between automated and manual driving models but also controls the degradation in driving performance during the ToC preparation phase and while manual control is maintained. The ToC operational state diagram is illustrated in
Figure 2.
This approach aims to minimize negative interactions arising from various vehicle parameters. Specifically, scenarios involving dynamic control in congested links and the activation of the ToC device for a controlled transition have been analyzed. These scenarios have been thoroughly analyzed to understand the impact of such transitions on overall traffic flow and various sustainability outcomes. The analysis particularly considered factors such as greenhouse gas emissions, emission levels, and the potential for reducing congestion-related delays. By evaluating these metrics, this study aims to provide insights into how controlled transitions can contribute to more sustainable urban mobility solutions.
Figure 3 presents the detailed pseudo-code for the network management system, developed using the TraCI interface, which allows for real-time communication and control within the simulation environment.
In this application, which is designed to activate the ToC device under specific traffic conditions, a systematic and dynamic monitoring approach is implemented for the entire İzmir urban transportation network. The network’s connection elements are monitored at five-minute intervals, allowing for real-time assessment of traffic conditions across different segments. The simulation begins by setting up a counter to track the simulation steps, with each step representing an incremental moment in the traffic simulation. During each interval, the current state of the network is evaluated by gathering key traffic metrics, such as occupancy rates, average speed, and the number of lanes on each edge (road segment) within the network. These metrics provide critical insights into traffic flow and congestion levels across the network.
A crucial part of the monitoring process involves identifying specific edges that exhibit potential congestion. An edge is flagged for further attention if it meets the following criteria: an occupancy rate of 75% or higher, an average speed of 6.94 m/s (25 km/h) or lower, and a length exceeding 200 m. These thresholds are selected to identify areas with high traffic density and slow-moving vehicles, indicating potential bottlenecks or congestion hotspots. Once such edges are identified, the simulation constructs a detailed report for each, including its occupancy rate, average speed, and length. These data are used to make informed decisions regarding traffic management interventions. Specifically, for vehicles classified as Level 4 or Level 5 AVs traveling on these congested edges, a request is automatically sent to activate the ToC device within a set period of five seconds. This request prompts the vehicle to prepare for a transition from automated to manual control, depending on the traffic situation. The activation of the ToC device in these AVs allows for more responsive and adaptive behavior in challenging traffic scenarios, potentially reducing the risk of accidents or further congestion. This process also facilitates smoother transitions in mixed traffic environments, where both autonomous and manual-driven vehicles coexist.
3.5. Dedicated Lane Application
DLs are road lanes reserved for specific classes of vehicle or modes of transport. These lanes facilitate the more efficient and sustainable integration of vehicles that share the same road with conventional vehicles but operate under different speed, safety, or automation standards. In large metropolitan areas where traffic congestion is a major problem, DLs are frequently employed on road sections that meet specific criteria. Particularly, freeways with at least three lanes are preferred for these lane separation applications. These applications are implemented on roads that experience severe congestion during morning and evening peak hours [
33,
34]. Upon examining the urban transportation network in İzmir, it is observed that these criteria are met on the high-standard road section extending from Balçova Viaduct to the Northern Aegean Motorway, known within the transportation network as the “İzmir Ring Road”. In this context, a DL application for the left lane has been designated on the İzmir Ring Road within the urban mobility environment.
In the network.xml file created for the DL, filtered under “İzmir Ring Road”, the allow = “vClass” variable, defined as index = “2” for three-lane sections and index = “3” for four-lane sections, has been changed to allow = “private” for vehicles classified as Level 4 and Level 5 (
Figure 4). This change ensures that DLs are exclusively used by private vehicles and those at specific levels of automation. This implementation is considered a strategic approach to reducing traffic congestion and enhancing mobility in urban areas.
Figure 5 illustrates examples of DL applications designed for left lanes within the urban mobility environment on road segments highlighted in gray. These gray areas permit specific vehicle classes to travel with priority and without interruption, especially under heavy traffic conditions in major metropolitan areas. The segments highlighted in gray are critically important for enabling vehicles with high levels of automation to pass through dense traffic quickly and safely.
AVs traveling on DLs incorporate detailed calculations for various driving scenarios, such as lane changes, acceleration from a stop, and accounting for the leader vehicle’s acceleration, all aimed at reducing jerks. The microscopic acceleration profiles of vehicles on these lanes are replicated through calculation steps. Specifically, a mechanism has been defined to prevent collisions by reducing deceleration when gaps in distance decrease momentarily after entering the DL. The equation presented in the car-following model considers the leader vehicle’s acceleration,
an(
t), to calculate the Constant Acceleration Heuristic (CAH),
aCAH(
t), as shown in Equation (1) [
35].
In Equation (1),
θ represents the Heaviside step function. In Equation (2), the acceleration
is calculated using a new “coolness” parameter,
, which ranges between 0 and 1. This parameter describes how “coolly” (calmly) the driver reacts when gaps decrease.
Through the models used, it is possible to temporarily accept shorter gaps in terms of distance, employing smaller safe gaps, particularly noticeable during lane changes. Additionally, the models include adjustments aimed at minimizing jerks in various driving scenarios, such as lane changes and acceleration after a stop. The car-following model exhibited on the DL operates optimally within time intervals ranging from 0.05 to 0.5 s [
36].
4. Results and Discussion
Different combinations of ToC and DL applications were systematically implemented under peak hour traffic conditions in the urban transportation network of İzmir. The analyses primarily focused on two distinct and critical perspectives to ensure a comprehensive understanding of their impacts. The first perspective involved a detailed examination of data from 199 specific link elements along the İzmir ring road, where the DL application had been strategically integrated into the existing urban transportation network. Data extraction was performed using a Python API (version 3.11.5), enabling precise and accurate collection of relevant traffic information. These 199 link elements were evaluated on a local scale using a set of carefully selected performance criteria. These criteria include key metrics such as density, occupancy rates, time loss, travel time, and waiting time, all of which are essential indicators of traffic flow efficiency and overall transportation performance within the network.
The second perspective expanded the analysis by focusing on the urban-scale implications of simulations that incorporated various combinations of preventive measures. These simulations aimed to assess the comprehensive impact of ToC and DL applications on İzmir’s entire transportation network. The evaluations paid particular attention to results that directly influence various sustainability indices. These indices included critical factors such as greenhouse gas emissions, closely linked to environmental sustainability; air pollutant emissions, which affect public health; energy efficiency, reflecting the economic sustainability of transportation systems; delay times, indicative of the network’s operational efficiency; and overall traffic performance, which measures the effectiveness of the transportation network in managing urban traffic volumes during peak hours.
The local and urban-wide impacts of ToC and DL implementations were examined, providing a comprehensive assessment of their potential to enhance traffic management, reduce congestion, and contribute to the sustainable development of urban transportation systems. The results from the initial local evaluation are presented in
Table 4.
Table 4 examines the operational modes under four distinct conditions. The first row presents the results for the scenario where only the DL application is implemented within the transportation network, while the second row, labeled “non-DL”, serves as the baseline scenario without any preventive measures like ToC or DL. The third row lists the scenarios where both DL and ToC are implemented concurrently, and the last row shows those where only ToC is applied without DL. The improvements resulting from these preventive measures, as compared in
Table 4, are detailed in
Figure 6.
Figure 6a illustrates the improvement percentages achieved through the implementation of DL. The data reveal that all scenarios involving DL consistently show higher improvement rates across all comparison metrics compared to non-DL scenarios. Notably, in the SCN-PI and SCN-PII scenarios, improvements of at least 15.79% in travel time and up to 53.39% in waiting time have been observed. These results demonstrate that DLs promote a more homogeneous distribution of vehicles, thereby reducing bottlenecks and enhancing the overall efficiency of traffic flow. This finding is particularly significant as AV penetration rates increase, where effectively managing mixed traffic conditions becomes crucial for maintaining optimal traffic performance.
Upon adding the DL precaution alongside the ToC application, as shown in
Figure 6b, the SCN-WI and SCN-WII scenarios—previously exhibiting the poorest sustainability outcomes according to penetration rates—achieved better results in density, occupancy, and time loss values compared to the improvements in
Figure 6a. The application of ToC led to Level 4–5 AVs transitioning to manual control in the SCN-WI and SCN-WII scenarios, which improved mixed traffic conditions. However, this also negatively impacted travel time and waiting time due to the restricted use of the left lane.
Figure 6c shows the improvements achieved through the ToC application without DL implementation. In the SCN-PI and SCN-PII scenarios, the improvements vary, ranging from positive to negative, with a maximum of 2.35%. This limited improvement is likely due to the congestion conditions required to activate the ToC device not being met among vehicles using the İzmir Ring Road, resulting in outcomes similar to the reference situation. In contrast, in the SCN-WI and SCN-WII scenarios, the ToC device was activated, leading to an approximately 10% improvement in density, occupancy, and time loss metrics. Additionally, for these scenarios, waiting time—representing the duration vehicles remain stationary—improved by 3.3% to 7.7% due to the ToC application. As indicated in
Table 3, the penetration rates of Level 4–5 vehicles in the SCN-WI and SCN-WII scenarios are 85% and 90%, respectively. The lack of improvement in travel times in these scenarios may be attributed to the ToC device negating the advantages of AVs, such as higher speeds and shorter following distances.
Figure 6d presents the improvements observed when comparing the “DL” scenarios with the “ToC with DL” scenarios. In the SCN-PI and SCN-PII scenarios, there was no significant improvement in density, occupancy, and time loss metrics, similar to what was observed in
Figure 6c, due to the absence of congestion and the subsequent nonactivation of the ToC device. However, in the SCN-WI and SCN-WII scenarios, the combined implementation of DL and ToC measures led to improvements of 9–18% in density, occupancy, and time loss values, while travel time and waiting time outcomes were negatively affected. This can be attributed to the inability to fully utilize the performance advantages of AVs due to restrictions on the use of the left lane.
Following the findings obtained from the first perspective, the effects of preventive measure combinations within the scope of the second perspective were examined on an urban-wide scale. These evaluations analyzed the broad impacts of various ToC and DL implementations across İzmir’s entire transportation network. The results of this assessment of urban-wide effects are presented in
Table 5.
Table 5 similarly contains various combinations of ToC and DL implementations. In contrast, the columns present results that directly affect sustainability indices on an urban scale. The improvement outcomes from nine different evaluation criteria, which impact environmental, social, and economic sustainability indices, are thoroughly compared in
Figure 7.
As illustrated in
Figure 7a, the DL application has achieved significant urban-level enhancements across all comparison measures when compared to the reference scenario. Although the DL’s strong local influence diminishes somewhat as it extends into the urban area due to the heavy traffic on the İzmir Ring Road during peak hours, its effectiveness in improving traffic flow and mitigating congestion remains significant at the urban scale. The improvements are consistently positive, with particularly notable increases observed in the SCN-PI and SCN-PII scenarios compared to others. The enhancements achieved by DL are noteworthy, even at lower traffic density. In these scenarios, improvements ranging from 9% to 18% were noted in all variables except speed and energy consumption. In the SCN-WI and SCN-WII scenarios, the integration of AVs with high levels of automation and the use of DL led to a 12–13% improvement in energy consumption results. Moreover, in all other scenarios, reductions in CO
2 and air-polluting emissions (NO
x and PM
x) were observed, ranging from 7.70% to 17.76%. These findings indicate that DL not only improves traffic performance but also significantly contributes to environmental sustainability.
Figure 7b shows the improvements resulting from DL implementation in the presence of ToC. The outcomes are consistent with the findings discussed in
Figure 7a for the SCN-PI and SCN-PII scenarios. Since the activation of the ToC device is dependent on congestion levels and considering that traffic volume is low and mixed traffic conditions are not severe in these scenarios, the improvements observed are similar to those provided by the DL implementation alone. However, in the SCN-WI and SCN-WII scenarios, the activation of the ToC device has led to a reduction in improvement values for certain metrics such as speed, waiting time, and total travel time, with some even showing negative results. This could similarly be attributed to the restricted use of the left lane.
Figure 7c illustrates the improvements achieved through the application of ToC alone, without DL implementation. Among all the scenarios, the most notable improvement was observed in SCN-WI and SCN-WII, with approximately a 10% reduction in waiting time. In other scenarios, the improvements varied from positive to negative, ranging from a maximum of 5.38% to a minimum of −6.47%. This graph indicates that the ToC application generally has a limited number of positive effects on an urban scale. The other graphs in
Figure 7 also show that DL implementation leads to more effective improvements.
Figure 7d shows the improvements achieved when ToC is added in the presence of DL. It can be said that the results are similar to or even more negative than those in
Figure 7c. In the SCN-WI and SCN-WII scenarios, the activation of the ToC device and the presence of DL caused the previously positive total travel time results seen in
Figure 7c to decline to around −6 to 7%. This could be attributed to the presence of Level 4–5 vehicles at high penetration rates. The advantages of AVs, such as higher speeds and shorter following distances, were negated when the ToC device was activated. In both cases, whether DL is implemented or not, ToC does not provide sufficient improvements at the urban scale and only produces local effects.
5. Conclusions
This study conducts an in-depth examination of the effects of AVs on transportation systems and their interaction with traffic management systems. The simulations and analyses conducted highlight the impact of various ToC and DL scenarios, particularly in mixed traffic environments, on traffic performance and sustainability. The results demonstrate that DL implementations improve traffic flow, reduce congestion, and support sustainable urban transportation systems on both local and urban-wide scales. Nevertheless, the integration of ToC and DL was found to be more compatible in specific scenarios (e.g., SCN-PI and SCN-PII), underlining the precautions that should be taken. Conversely, more limited benefits were observed in scenarios such as SCN-WI and SCN-WII.
For the effective application of ToC measures in road segments with DL implementations, it is essential to achieve high penetration rates of Level 4–5 AVs and to manage congestion conditions. Particularly, the design of implementations on these roads should be approached with caution, as metrics such as travel time and waiting time might lead to undesirable results in certain situations. In general, the findings emphasize the importance of strategically applying ToC and DL measures, especially in urban areas with severe congestion. These insights can guide future urban transportation planning and suggest that a combination of advanced traffic management strategies be employed to mitigate congestion and enhance sustainability outcomes. Overall, this study highlights the need to optimize traffic management strategies as AVs become more common and emphasizes the importance of comprehensive planning to facilitate the effective integration of these vehicles with traditional ones.
However, it is important to acknowledge certain limitations of this research. The simulations were based on specific assumptions and excluded dynamic factors such as weather conditions, pedestrian behaviors, and the presence of heavy vehicles, which may affect the generalizability of the results. The examination of heavy vehicles’ impact in open-source urban mobility environments has been addressed only to a limited extent in the academic literature. Specifically, in residential areas where the distribution of heavy vehicles exceeds certain levels, the microscopic or mesoscopic effects of these vehicles on autonomous systems will be investigated in future work. Additionally, this study focused on the urban mobility environment of İzmir, which may limit the applicability of the findings to other cities with different transportation infrastructures and regulations. Each city possesses unique characteristics that can influence the outcomes of ToC and DL applications. Future studies are recommended to investigate various penetration rates and vehicle interaction dynamics in greater detail, incorporate a broader range of real-world variables, and develop new traffic management strategies that adapt to the evolving dynamics of AV technologies.
Furthermore, the autonomous vehicle models and control strategies employed in this study are based on specific assumptions related to the existing literature and the simulation environment. As AV technology continues to evolve, future AVs may incorporate more sophisticated sensors, AI algorithms, and decision-making processes, introducing uncertainties regarding how accurately the current models will reflect real-world applications. Addressing these technological advancements in future studies will enhance the relevance and applicability of the results. Future research should aim to address these limitations by incorporating a wider range of factors and exploring diverse urban contexts. A more comprehensive examination of the impacts of technological advancements in AVs will further enhance the relevance and applicability of the obtained results, ensuring that traffic management strategies remain effective in the face of rapidly evolving transportation landscapes.