This article addresses the lack of a reliable, flexible simulation tool for evaluating the impacts of CAV platoons on existing traffic systems. The framework must adapt to evolving technologies and contexts. To capture both vehicle behavior and network effects, it employs two simulation levels:

1) a lower-level module that models individual CAVs and platoons from a vehicle-centric perspective, and

2) a higher-level traffic module that examines network-level implications under varying demand, control, and infrastructure designs.

1) What is the functional architecture of a connected and autonomous/automated vehicle, and which technologies are essential for platoon-capable CAVs?

2) How should vehicles be formed into platoons, how do platoons operate effectively in connected/automated environments, and what is an appropriate functional architecture for CAV platooning?

3) How are platoon-capable CAVs and CAV platoons modeled at the microscopic (vehicle) level, and how should mixed traffic with platoons be simulated?

4) Once platooned CAVs are introduced, how should their network-level impacts be evaluated?

Solving transportation problems requires attention to temporal and spatial dimensions. Roadway plans and designs often span a decade or more, and the rapid evolution of CAV technologies necessitates anticipating future operating conditions. Governments and standards bodies worldwide are actively shaping policies and technical guidance for real-world deployment. Within this context, vehicle platooning, a form of cooperative driving, has drawn significant interest. Developing a tool to explore its impacts is both timely and important for surface-transportation planning, design, and operations.

1) Survey CAV technologies and applications, and formulate a functional architecture for both standalone CAVs and CAV platooning.

2) Develop vehicular models for (a) human-driven vehicles, (b) non-connected automated vehicles, and (c) connected and automated vehicles with platoon functions.

3) Model platoon-capable CAV operations in a microscopic environment.

4) Build traffic-level simulation scenarios with varying market shares of platoon-capable CAVs and differing roadway designs.

5) Integrate the vehicle-level models, traffic simulations, and test scenarios to evaluate CAV-platoon impacts under diverse conditions; verify the framework with sample tests.

6) Validate the framework through case studies on real infrastructures and traffic conditions, assessing outcomes across different CAV-platoon shares.

This section has three parts. First, it presents the functional architecture of connected and automated vehicles (CAVs) adopted in this study. Next, it reviews implemented and emerging technologies for connected driving, automated driving, and vehicle platooning. Finally, it defines and explains the components that make up the architecture.

Functional Architecture of a Connected and Automated Driving System

Drawing on current implementations and ongoing developments, Figure 1 summarizes the functional architecture used in this study. A viable CAV must operate safely with limited human intervention, moving occupants and goods from origin to destination safely, correctly and reliably. To achieve these goals, the architecture is organized into three functional layers:

Perception

Acquires and interprets information about the vehicle and its surroundings.

Figure 1. Functional Architecture of a CAV.

Core capabilities include object/event detection and tracking, localization and mapping, and state estimation. Sensor fusion and connectivity (e.g., V2X) extend situational awareness beyond line-of-sight.

Planning

Determines what the vehicle should do to complete the trip safely, comfortably, and efficiently. This spans route planning, behavior/decision planning (e.g., yielding, lane selection, platoon join/leave), and motion planning that generates feasible trajectories under traffic rules, policies, and constraints.

Control

Executes the planned actions accurately by generating steering, throttle, and braking commands. This includes longitudinal and lateral controllers, trajectory tracking, and fault-tolerant mechanisms to maintain performance under disturbances or component degradations.

The remainder of this section reviews the enabling technologies for connected driving, automated driving, and platooning, and then details how these components interact within the architecture.

Implemented technologies include widely available driver-assistance features on the market. The Society of Automotive Engineers (SAE) defines vehicles that activate a single automated function as Level 1 driving automation. Table 1 shows

Table 1. Driver assistant technologies.

the levels of driving automation per the SAE J3016 standard. An integrated system with multiple driver-assistance functions operating simultaneously is considered Level 2.

The National Highway Traffic Safety Administration (NHTSA) groups driver-assistance technologies into four functional categories: collision warning, collision intervention, driving-control assistance, and other systems. Table 2 lists commonly implemented technologies within these categories. Features in Table 2 are often combined to deliver broader assistance.

Table 2. Summarizes the setting items of the calibrated model.

For example, Forward Collision Warning (FCW) can be integrated with Automatic Emergency Braking (AEB) to reduce the risk of rear-end collisions or mitigate severity when a crash is unavoidable. When a potential forward collision is detected, FCW alerts the driver to take corrective action; if the driver does not respond in time, AEB applies the brakes automatically. Another example is the pairing of Rear Cross-Traffic Warning (RCTW) with Rear Automatic Braking (RAB), which work together to help avoid collisions with overlooked objects while reversing—mirroring the FCW+AEB interaction. During steady cruising, combinations of Adaptive Cruise Control and Lane-Keeping Assistance/Lane-Centering Assistance can improve comfort for both driver and passengers.

In 2011, Schünemann developed a runtime infrastructure for coupling communication and traffic simulators and conducting sophisticated V2X simulations. The infrastructure is called V2X Simulation Runtime Infrastructure abbreviated as VSimRTI [1]. The infrastructure was designed to manage a federation of simulators. A Federate consists of a simulator, a federate ambassador, and a VSimRTI ambassador. The simulator could be coupled with VSimRTI, which includes SUMO, VISSIM, JiST/SWANS, OMNeT++, etc. VSimRTI has been upgraded to Eclipse MOSAIC.

Segata et al. [2] developed and verified a simulation tool for automated platooning in mixed highway scenarios with the combination of a communication simulation platform and a traffic simulation platform, OMNeT++ and SUMO, in 2012. The proposed platooning protocol in the study was CACC-based. The results showed that with 30% penetration of platoon-enabled vehicles, over 80% of platoons have a size of less than three. The recorded largest size of a platoon during the simulation was 8. In 2014, Segata et al. published the works done based on this simulation tool and named the completed tool PLEXE.

In 2013, Zhao and Sun developed a simulation framework that explores the vehicle platooning and car-following behavior under the CV environment. A six-vehicle CACC platoon was simulated on a microscopic traffic simulation platform VISSIM. Examined platoon maneuvers include forming, adjusting, splitting, dismissing, and joining. They concluded that the lane capacity is positively correlated with the market penetration of CACC-based platoons. On the contrary, platoon size impacts the capacity subtly. The minimum desired headway for ACC vehicles was set to 1.4 seconds as well as 0.5 seconds for CACC vehicles [3].

Jia et al. presented a disturbance-adaptive (DA) design for VANET-based vehicle platoons. The design aimed at improving the platoon’s stability of communication and operation. Within a DA platoon, four roles were played by vehicles. The first is the Leader. Like the role in a common platoon system, a Leader makes decisions on platoon-level operations (formatting, splitting, merging, broadcasting the existence of the platoon, etc.). The second role is the Tail. As the name indicates, the last vehicle in a platoon plays this role, and oversees communication with the leader of next platoon. The third role is called Relay. Relay vehicles perform data-forwarding in a multi-hop VANET environment. The last role is Member, who simply follows the plan decided by the Leader. Based on such theory, authors developed a driving strategy for Leaders of DA-platoon. The driving strategy adopts a sliding mode controller and determines the desired inter-platoon gap by gathering information from the Tail of the preceding platoon. The principles guaranteed the gap to be small enough for not breaking the connectivity with adjacent platoons and large enough for avoiding collisions. The principles also helped with determining the platoon size. Then, simulations were conducted for algorithm validation [4].

In 2014, a distributed framework was developed for coordinating HDVs for platooning. The core of this framework is a virtual controller system that coordinates HDVs at each vertex (e.g., each major intersection) for platooning. The objective is to maximize the earnings (energy savings minus costs). A case study was conducted based on the German Autobahn network. 0 to 7000 HDVs were initialized in the network. The results showed that as the number of HDV increases more fuel saving can be achieved [5].

Artery, a simulation framework based on OMNeT++, was introduced in 2015 as an extension of Veins. Recently, an extension called Artery-C to adapt C-V2X communication approaches was released. An overview of Artery-C was also published. The pros of Artery-C include 5G-selected features, open-source licensing, and capabilities for modeling facilities [6].

In 2015, Santini et al. presented a consensus-based approach for vehicle platooning under IVC environments. The consensus algorithm was implemented to achieve equal inter-vehicle gaps within a platoon. Validation simulations were conducted. The results showed that the approach performed better than the classic CACC approach by improving the stability and convergence of the platoons. Later, in another publication, Santini et al. presented their work on validating the consensus-based platooning approach when vehicle communication topology changes due to platoon maneuvers. Simulations were developed and run upon PLEXE. The results proved the reliability of the approach, even while a platoon with heterogeneous vehicles is maneuvering. Tested maneuvers include join-at-tail, leave-at-tail, join-at-middle, and leave-at-middle. Since the communication topology is a typical leader and predecessor following topology, vehicles in a platoon communicate with the platoon leader and the two adjacent vehicles (one ahead and the one after) [2].

In 2015, Li et al. reviewed relevant studies and presented a four-component framework of the vehicle platoon system. The four components of the framework are 1) Node dynamics, 2) information flow topology, 3) distributed controller, and 4) formation geometry. Node dynamics describes the behavior of every platooned vehicle and all others involved. Information flow topology means how vehicles exchange information with each other. Feedback controllers handle feedback control with neighboring information. Formation geometry indicates the desired inter-vehicle gap within a platoon [7].

Deng published a simulation framework for modeling and analyzing heavy-duty vehicle (HDV) platoons. The simulation framework was built upon the commercial simulation platform, VISSIM. Within the framework, a fuel consumption model was embedded to estimate the influences of HDV platoon on fuel saving. Input data for fuel consumption estimation is obtained by recording the state of each vehicle. The truck’s state data includes motion state (lane, longitudinal position, speed, acc-/deceleration), platooning information (in-platoon position, inter-vehicle distance), and temporal data (time instance). Three cases were studied to figure out the influences of HDV platoon and platoon formation maneuvers. The results showed that, with HDV platoons, even if the aggregate highway velocity dropped, the traffic flow rate increased on a two-lane highway section. Otherwise, under medium/high traffic scenarios, the interest in reducing platoon formation time conflicted with the arrival time of the HDVs.

Ribeiro et al. proposed a platooning management protocol. In the study, the protocol covers the procedures of platoon creating, merging, and dissolving operations as well as vehicle joining and leaving maneuvers. Simulation and testing were conducted on a combination of SUMO, V2X Simulation Runtime Infrastructure (VSimRTI), and ns-3. Two-truck platoons were introduced into the simulation network. The results showed that, with the protocol,

a) vehicle joining the platoon from rear has been completed in 69 seconds(s) on average, or in 114.2 seconds when joining from the adjacent lane;

b) on average, a vehicle leaving the platoon took 23.3 s to change to the adjacent lane, or 1 s if it’s the last vehicle in the platoon and leaves by increasing gap to its preceding vehicle;

c) adjusting the gap took 12.6 s for a joining maneuver or 17.4 s for a leaving maneuver on average;

d) merging two platoons was similar to the joining maneuver, but with 62.7 s adjusting on average;

e) dissolving maneuver took 3.8 s on average.

Additionally, message latency was limited to 100 milliseconds (ms), which meets the required maximum delay. The capacity of a lane was proven to be better with platoons.

In Bang and Ahn’s 2017 publication, a platooning strategy for CAVs based on the spring-mass-damper system is described. With the system, the longitudinal platoon control of CAVs by controlling the spring constant and damping coefficient is presented. The maximum acceleration/deceleration, mass, and length were used to determine the controlling parameters. Moreover, different relations between the spring constant and traffic flow were considered when developing simulation scenarios. The results proved that, with the critical damping coefficient, the maximum efficiency of completing platoon formation could be achieved when the maximum relation between flow and spring constant was selected.

In 2017, Jain et al. published their work on developing a prediction-based framework for vehicle platooning. An MPC-based control algorithm, which solely relied on V2V communication, was developed and implemented for vehicle platoon control. Both simulations and experiments were conducted to test the algorithm. Simulations were performed on the Dominion framework developed by German Aerospace Center. A two-vehicle platoon with both 5G-V2X and 802.11p communications was used for the experiments. Both simulation and experiment results showed the excellent potential of controlling a vehicle platoon with pure V2V communication. The authors also concluded that 5G performed better than 802.11p in providing V2V/V2I communication because of its larger data rate and communication range.

In 2017, Liu et al. published their research on the platoon system engineering process that considers safety and cybersecurity issues. The engineering process has four steps—1) defining the safety goal, 2) defining the attack model, 3) deriving the security goal, and 4) deriving functional security requirements. Due to the tight coupling among vehicles in a platoon, the impact of a cyber-attack against the platoon system could lead to decreasing stability, platoon dissolution, and even collision. Functional security requirements of a platoon system included but were not limited to detecting false messages, ensuring the timeliness of messages and responses, and keeping messages intact from attackers. The general approaches to developing a platooning system with the proper capability to maintain safety and security include optimizing the gap between vehicles with security consideration, enabling cyber-attack detection, and deploying fail-safe mechanisms to eliminate harm when attacks are encountered. In the paper, a proactive platooning approach was presented. The approach calculated the optimal acceleration difference threshold based on the desired acceleration under CACC and pure ACC situations. Moreover, the desired gap was determined based on the desired acceleration. Simulations with PLEXE and MATLAB validated the algorithm [8].

In 2018, Ramezani et al. developed a simulation model for exploring the influences of CACC-based truck platooning operations on traffic. Aimsun Next Micro-SDK was used for developing the simulation platform. The aspects of trucks, including desired acceleration speed, were computed explicitly and implemented in the simulation. A case study was conducted based on a 15-mile urban section of the I-710 Northbound an Interstate highway in Southern California. The results showed that truck CACC platoon could increase the speeds of cars by reducing congestion when penetration rates reach a high level (over 80%). However, in on-ramp areas, since truck platoons used the rightmost rule, the merging traffic had to wait longer when penetration rates were low. Generally, when penetration rates of CACC trucks reached 100%, the benefits included easing congestion propagation and increasing the average speed of traffic at uncongested areas. Otherwise, similar benefits were not found with CACC car platoons [9].

In a publication by Ibrahim et al., a co-simulation framework developed by the authors for vehicle platooning was presented. The framework consists of ns-3, SUMO, and MATLAB. Ns-3 simulated the packet broadcast of vehicles. SUMO simulated the traffic. Control algorithms were developed in MATLAB and replaced the algorithms that were given in SUMO. MATLAB also performed as the interface between SUMO and ns-3. The developed algorithms focused on longitudinal acceleration control. Model predictive control and state-feedback control algorithms were implemented for the upper- and lower-layer control. The upper-layer controller determined acceleration or deceleration regarding the gap to the preceding vehicle. The lower-layer controller worked to eliminate errors in acceleration/deceleration. Tests were conducted for framework validation. Simulations based on realistic highway scenarios were conducted as well. The results showed that both the severity and frequency of platoon speed fluctuations increased when packet losses grew.

Vieira et al. developed a realistic simulation framework for vehicular platooning based on the Robotic Operating System (ROS) framework and published a paper in 2019. ROS is a popular framework for designing robotics applications. An integration of Gazebo (ROS robotic simulator) and OMNeT++ was presented. ROS publish/subscribe mechanisms played a critical role in data delivery and simulator synchronization. In a later Vieira et al. publication, after more work had been done, this realistic simulation framework was named COPADRIVe.

In 2019, Gerrits et al. developed a study exploring the influences of opportunistic truck platooning matchmaking algorithms. One is First-Viable Match (FVM), and the other is Best-Match (BM). The FVM takes waiting time into account as the cost. Once a match can lead to a positive earning (subtracting savings over costs), the match is selected. The BM selects the match with the highest earnings within a searching area. Properties of the truck (hourly wage, urgency, brand, destination, matching locations) were added as factors into the simulation model. The results showed that BM performed better than FVM on saving. The wage savings are significant when platooning trucks.

Sethuraman et al. developed a simulation to evaluate the impacts of bus platoons on traffic. The simulation scenario was developed based on a 16-kilometer (KM) section of an urban roadway in Singapore with two major signalized intersections. The simulation was run on the VISSIM platform. Both the quality of services of the bus and the performance of the traffic were assessed. Numerical analysis showed that, generally, the simulation results showed that bus platooning increased the operational speed of buses and other cars and, as a result, the overall delay was reduced for both types of vehicles. Moreover, output data indicated similarity with platooning trucks since platooning buses reduced the aerodynamic drag and fuel consumption. Correlation between the number of buses in the platoon showed a positive coefficient with energy savings.

In 2019, Hoef et al. published their predictive framework for dynamic HDV platoon coordination. The presented framework aimed to coordinate the in-route formation of platoons. The core of the framework is a platoon coordinator. A layered control system architecture for coordinated platooning was also presented. The layers are the service layer, strategic layer, tactical layer, and operational layer. On the first layer, transport tasks are managed. On the second layer, strategies, such as coordinating platoons, are performed. For the remaining two layers, platoon management systems and vehicle control systems are deployed.

In 2019, Hyun et al. published a paper that overviews a statistical verification framework for a platooning system of systems (SoS). The framework, called StarPlateS, consists of three modules, scenario generation module, simulation module, and verification module. The first module performed platoon configuration generation and scenario generation. The simulation module was handed over to a SUMO/OMNeT++ integrated extension, Vehicular Network Open Simulator (VENTOS). The verification module checked the achievement rates of goals with the statistical model checking (SMC) algorithms. The two checked properties of the SoS were the throughput within a specific time horizon and the rejection rate of operation.

Since 2006, Sommer et al. have worked on a model library for OMNeT++. It’s named Veins. A 2019 overview summarized the developments of the library. Now, Veins supports simulations not only relating to the IEEE 802.11p family but also LTE and Visible Light Communication (VLC). Veins does not manage road vehicle simulations. However, by bidirectional coupling with SUMO through the Traffic Control Interface (TraCI), users may customize vehicular mobility models on demand. Models on the communication layer are created by Veins in the OMNeT++ simulator to represent vehicles. Then, by establishing mapping with the mobility models in SUMO, the cyber-physical system of a connected-vehicle environment is simulated.

Quadri et al. published work on a MEC-based vehicle platoon control framework for vehicle platooning in 2020. MEC is the abbreviation of multi-access edge communication. Being different from some distributed controllers, a MEC-based controller offers a centralized approach to platoon control. Simulations relying on SUMO and Python-based applications were conducted for two types of scenarios—sinusoidal and real-trace vehicle movement patterns. According to the results, with a MEC-based controller, inter-vehicular spacing can be shortened to 5 meters or less. However, since the round-trip time (RTT) of cloud computing could hardly be achieved below 150 ms on average, deploying the controller onto the cloud was not suitable for such centralized control.

In 2021, Hidayatullah and Juang published their study on the centralized and distributed control framework under homogeneous and heterogeneous platoons. In the paper, they used PreScan and MATLAB/Simulink to establish simulations to investigate the string stability of both centralized and distributed control frameworks comprehensively. Features that include vehicle dynamics, sensing, and V2X communication are added to the simulations. The performance index integral square error (ISE) was used to evaluate two frameworks. The results showed that, with 0.05 latency and 30% packet loss probability, a distributed framework achieved slightly less ISE of mean spacing error than a centralized framework.

In 2021, Miekautsch et al. published their study on a situation-dependent communication topology for platooning heterogeneous vehicles. In that publication, platoon systems were studied via the four-component framework. A heterogeneous vehicle platoon indicates that a platoon was formatted by vehicles with various configurations (power-train time constant, max acc-/deceleration, max speed). Two situations were simulated, emergency braking and count-in. For the emergency braking situations, the author presented a flexible reversed Leader-predecessor-follower (LPF) communication topology. With such topology, the vehicle’s configuration and the order within its platoon were considered. By deploying this topology, a platooned vehicle with a higher maximum deceleration rate can brake harder compared to where the traditional LPF topology was deployed, and the vehicle was limited to use the lowest deceleration rate of all vehicles within its platoon.

In 2021, a publication by Xu et al. revealed an open-source simulation tool for cooperative driving automation (CDA). It’s called OpenCDA. The key features of OpenCDA were summarized as IFMBC, integration, full-stack platform, modularity, benchmark, connectivity, and cooperation. OpenCDA selected CARLA and SUMO as the simulator for traffic. The hierarchical architecture of the tool was demonstrated. The top layer was the “PlatoonManager” class. Then is the “VehicleManager” class. The bottom layer includes “PerceptionManager”, “LocalizationManager”, “BehaviorManager”, “ControlManager”, and “V2XManager”. The internal operational sequence of the OpenCDA modules is illustrated in the OpenCDA logic flow shown in Figure 2.

The simulation framework developed in this research integrates four key platforms including GIS, MATLAB/Simulink, SUMO, and OMNeT++, into a co-simulation environment to model vehicle-level dynamics, network-level traffic behavior, and real-time V2X communications of Connected and Autonomous Vehicle (CAV) platoons. This integrated structure allows for the detailed exploration of both microscopic and macroscopic effects of platooning, bridging the gap between control-theory-based vehicle modeling and transportation-system-level

Figure 2. OpenCDA logic flow.

traffic evaluation.

This suite of platforms was selected because each provides distinct and complementary modeling capabilities. GIS ensures spatial accuracy and realistic roadway geometry; MATLAB/Simulink enables detailed modeling of vehicle dynamics and cooperative control algorithms; SUMO provides scalable microscopic traffic simulation; and OMNeT++ supports realistic modeling of V2X communication performance. Together, these tools offer an adaptable and comprehensive environment for studying CAV platoon behavior.

2.2.1. Architectural Overview

The framework is built upon the Veins-PLEXE architecture, a widely used simulation interface for coupling traffic and communication simulators. In this structure, each vehicle in the SUMO traffic simulator is mirrored by a corresponding communication node (module) in OMNeT++, following the APP-MAC-PHY hierarchy compliant with IEEE 802.11p and IEEE 1609.4 standards. These layers collectively simulate the Wireless Access in Vehicular Environments (WAVE) and ETSI ITS-G5 communication stacks, enabling the emulation of Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) messaging within platoons. The interaction among SUMO, OMNeT ++, and the TraCl interface in our integrated simulation setup is illustrated in Figure 3.

Figure 3. Architecture of Integrated Simulation based on PLEXE.

The integration is managed through the Traffic Control Interface (TraCI), a TCP-based client-server protocol that synchronizes state data between SUMO and other platforms such as MATLAB or OMNeT++. This ensures that changes in vehicle states (e.g., speed, acceleration, position) in SUMO are instantaneously reflected in OMNeT++ communication nodes and MATLAB-based controllers, forming a cyber-physical feedback loop.

2.2.2. Vehicle-Level Modeling

At the microscopic scale, MATLAB/Simulink is employed to simulate vehicle dynamics, platoon control, and autonomous driving algorithms. Three distinct control models were implemented to represent different vehicle types:

Cruise Control (CC) for Human-Driven Vehicles (HDVs),Adaptive Cruise Control (ACC) for partially autonomous vehicles, andCooperative Adaptive Cruise Control (CACC) for connected platoon-enabled CAVs.

Each control model is designed around the perception-planning-control architecture of an Autonomous Driving System (ADS). The longitudinal controller determines acceleration using spacing errors, relative speeds, and communication inputs, while lateral control manages lane-keeping and merging behaviors.

Communication latency and perception delay are also introduced in the Simulink environment, typically set between 1 - 10 ms for V2V/V2X communication and 5 - 20 ms for perception delay, modeled as Gaussian distributions with parameters μ and σ calibrated through field studies.

The platoon management logic includes state transitions such as “join platoon”, “leave platoon”, and “split/merge platoon”. For each vehicle, the control system dynamically switches between ACC and CACC modes depending on whether communication data is available from preceding vehicles. This approach supports realistic simulation of disturbances such as temporary communication losses, and the resulting adaptive control responses.

2.2.3. Traffic-Level Simulation

The macroscopic traffic simulation is executed in SUMO (Simulation of Urban Mobility), chosen for its scalability, open-source flexibility, and lightweight computational requirements. SUMO allows multi-lane, multi-vehicle microscopic traffic modeling while accommodating heterogeneous vehicle types. Using GIS-based roadway geometry, the northbound I-95 corridor from the Delaware House Travel Plaza to the Christiana Interchange was reconstructed with high positional accuracy. Real-world parameters such as lane configurations, speed limits, and traffic demand profiles were directly imported from DelDOT datasets and spatial layers.

In this simulation layer, PLEXE extends SUMO to represent CAV platoon behaviors including:

Platoon formation and dissolution,Inter-vehicle communication topology updates (Leader-Predecessor-Follower),Speed synchronization,Cooperative lane changing and merging, andString stability assessment.

The platoon’s operational logic is event-driven, controlled by thresholds in relative distance, speed, and communication delay, ensuring that vehicles can join or leave platoons dynamically based on traffic context and communication reliability.

2.2.4. Communication Network Simulation

OMNeT++ operates as the dedicated communication simulator. Each vehicle in SUMO is instantiated as a mobile communication node in OMNeT++, connected through the Veins framework. These nodes exchange Cooperative Awareness Messages (CAMs) and Decentralized Environmental Notification Messages (DENMs) using the IEEE 802.11p protocol. The Leader–Predecessor–Follower topology is used, where the platoon leader transmits acceleration and speed data downstream, while each follower relays acknowledgment messages upstream.

Key communication metrics such as latency, packet delivery ratio (PDR), and network throughput, are continuously logged to assess communication quality. In the developed framework, average delays remained within 5 - 8 ms, and PDR exceeded 97% under nominal load, confirming stable connectivity for cooperative control. Additionally, multi-hop relaying was modeled for extended platoons, enabling stable information flow beyond direct line-of-sight.

2.2.5. Integration and Synchronization

The co-simulation achieves synchronization through three primary interfaces:

1) SUMO-OMNeT++ (Veins/PLEXE): Exchanges vehicle states and message events in real time through TraCI.

2) MATLAB-SUMO: Transfers dynamic control outputs (e.g., acceleration commands) and retrieves aggregated performance indicators such as travel time, flow, and density.

3) GIS–SUMO: Provides geometric and spatial accuracy for roadway networks, demand assignment, and coordinate referencing.

The simulation timestep was fixed at 0.1 seconds to ensure smooth coupling across all platforms. Synchronization accuracy was validated by comparing time-stamped vehicle trajectories, confirming sub-millisecond temporal deviations between SUMO and OMNeT++ layers.

2.2.6. Framework Capabilities

The resulting simulation environment enables:

Detailed evaluation of string stability and inter-vehicle spacing performance.Network-wide assessment of mobility and emission impacts under varying platoon penetration levels.Testing of dedicated lane policies, mixed-traffic interactions, and communication delay effects.And analysis of platoon coordination failures due to network disruptions.

Overall, the proposed framework functions as a comprehensive multi-resolution platform, capable of translating low-level control dynamics into observable large-scale traffic phenomena, thus offering a powerful tool for assessing the real-world implications of CAV platooning on major highway systems such as I-95. The complete workflow of the integrated simulation framework used in this study is elaborately shown in Figure 4.

Figure 4. Logic flow of the instance of the proposed simulation framework in this study.

The testing and validation of the simulation framework were carried out in two main stages: calibration of vehicle dynamics and control parameters, followed by validation against real-world data collected from the I-95 corridor in Delaware.

2.3.1. Calibration Process

Calibration focused on ensuring that the longitudinal and lateral control laws reflected realistic driving dynamics. Three categories of vehicles were represented:

Human-Driven Vehicles (HDVs) using CC control,Autonomous Vehicles (AVs) using ACC control, andPlatoon-Enabled CAVs (P-CAVs) using CACC control.

Key parameters such as platoon size (2, 5, 8 vehicles), ACC time headway (1 - 2 s), and CACC constant spacing (2 - 20 m) were systematically tested. During calibration runs, MATLAB/Simulink simulated low-speed and lane-changing scenarios to assess string stability and speed synchronization between vehicles. Two measures of effectiveness (MOEs) were employed:

Integral of Gap Error, quantifying spacing deviations over time. Speed Synchronization Duration, measuring how long a following vehicle takes to match its leader’s speed after perturbations. Results indicated that smaller headways (5 m) improved stability without compromising safety, reducing speed synchronization time by up to 35% compared to larger gaps. The effect of communication or perception delays was limited, with less than 15% degradation in MOE performance, confirming the robustness of the model. Control parameters were finalized as: CACC damping ratio = 1, bandwidth = 0.2 Hz, and leader-weight factor = 0.7, ensuring optimal response across different platoon compositions.

2.3.2. Validation on I-95 Corridor

The validated scenario was modeled on the northbound I-95/Delaware Turnpike, extending from the Delaware House Travel Plaza to the Christiana Interchange. This 6-mile section was selected for its representative geometric design and heavy commuter traffic. Using field data obtained through GIS-based datasets, DelDOT sensors, and manual counts, baseline traffic volumes, densities, and average speeds were extracted. These data served as the benchmark for comparison with simulated results.

Simulation outputs, including speed profiles, flow rates, and density distributions, were statistically compared to field observations using two sample t-tests. The resulting p-value ≈ 0.07, evaluated using a significance level of α = 0.05, indicated no statistically significant difference between the two datasets, confirming the simulation’s ability to reproduce real-world conditions. Further verification confirmed that platoon maneuvers such as join-at-tail, leave-at-middle, and split/merge operations remained stable under high traffic demand, even with message delays of up to 10 ms.

In addition to behavioral validation, network-level communication reliability was evaluated using OMNeT++ metrics such as packet delivery ratio, latency, and inter-platoon connectivity duration. The average communication delay remained within 5 ms, ensuring that control commands were executed within acceptable real-time bounds. The consistency between observed and simulated macroscopic variables including average travel time and flow-density relationships, confirmed the high fidelity of the integrated model.

Through this calibration and validation process, the framework proved capable of accurately reproducing both vehicle-level dynamics and corridor-level traffic performance. It therefore provides a reliable analytical environment for exploring future deployment strategies of CAV platoons, including dedicated-lane designs, control algorithm optimization, and mixed-traffic interactions under realistic conditions on I-95.

This scenario simulated existing infrastructure without dedicated lanes, under traffic volumes projected for 2026, 2031, and 2036. The analysis revealed that as CAV platoon penetration increased, both average travel speed and throughput improved. Specifically, with 10% CAV platoon penetration, average speeds increased marginally by 0.1% - 0.2%; at 30%, improvements reached 0.9% - 1.6%; and at 50%, gains rose to approximately 3% - 4%. Moreover, the duration of sub-desired speed travel, an indicator of congestion, decreased by about 20% when platoon shares ranged between 30% and 50%. These findings highlight that even moderate platoon integration can enhance mobility under high-volume traffic.

The second scenario investigated both added and converted dedicated lane configurations for platoon-exclusive use on the same I-95 segment. Design volumes representing 2036 were employed to reflect future congestion levels. When dedicated lanes were introduced, network efficiency improved further at medium to high platoon penetration rates (≥20%). The simulations demonstrated that converting one existing lane for platoon use initially caused slight slowdowns for low penetration levels (<15%), due to underutilization. However, when penetration exceeded 25% - 30%, total corridor throughput increased, and average travel times decreased significantly. Scenarios with added lanes (rather than converted ones) yielded the best overall system performance.

Environmental metrics also showed benefits: fuel consumption and CO₂ emissions declined as platoon penetration grew, attributed to smoother accelerations, reduced headways, and fewer stop-and-go events. At 50% penetration, CO₂ emissions decreased by roughly 6% - 8% relative to the baseline, indicating both traffic and environmental advantages from coordinated platoon deployment.

A detailed analysis of the emission profiles was performed for selected vehicles under different penetration rates of CAV platoons. The emission curves showed that CAV platoons generally produced smoother acceleration and deceleration patterns, resulting in reduced CO₂ emissions among regular vehicles as platoon penetration increased. However, CAV platoon vehicles themselves exhibited higher instantaneous emissions at higher speed limits due to more frequent acceleration adjustments required to maintain string stability. The emission trends over time indicated that when the penetration rate reached 30% - 50%, the overall system-level CO₂ emissions decreased by approximately 6% - 8% compared to the baseline scenario.

The simulation results showed that when speed limits increased for the CAV platoon dedicated lane(s), CAV platoon flow could achieve better travel time than the regular lanes. Moreover, converting a lane to dedicated with low percentages (settings A and B) of CAV platoon traffic increased the regular lanes’ burden (lowered the capacity). But as the percentage increased, the average travel speed of all types of traffic increased. When adding a dedicated lane instead of converting one, settings F, G, and H showed the benefits for both types of traffic. However, when the percentage went higher, such a pattern solely improved the regular traffic, but situations of CAV platoon traffic remained similar for both adding and converting. As a result, improvement for both types of traffic have been achieved. As the percentage of CAV traffic increased, the improvement increased gradually.