Chapter 2. Evaluation of Analysis, Modeling, and Simulation Tools for Connected Vehicle-Enabled Road Weather Management Strategies

Assessment Framework

This section presents the assessment framework developed to evaluate existing simulation and modeling tools from the standpoint of connected vehicles (CV) and road weather management applications. Following an overview of existing methodological approaches, functional requirements are identified for the analysis, modeling, and simulation (AMS) tools with regard to these capabilities.

Road weather and CV factors have only been partially addressed in previous studies and in existing tools. Road weather impacts have mostly been captured in mesoscopic tools for which typical detector data can be used for calibration. Road weather impacts on CV performance, including communication capabilities, have not been addressed. The opportunities made available by CVs for better prediction and management are currently under development by researchers. The AMS testbed developed for Chicago is the only one that has sought to integrate road weather and CV factors.

Background on Existing Platforms and Related Features

Model platforms that integrate various components are required to capture interactions related to CVs. Platforms in this context are primarily conceptual analytical constructs embedded in a software tool. They typically entail a collection of models representing interacting agents or processes. Different physics have been used to represent flow processes in these platforms. The main differentiation has been in terms of the detail of representation, with micro, meso, and macro being the main labels for this differentiation. Dynamic traffic assignment (DTA) tools have been developed with all three types of physics; the discussion here primarily applies to particle-based simulation, where individual vehicles/entities are tracked and used in conjunction with either meso- or micro-level physics.

Examples of simulation-based DTA platforms used in research, practice, or both, include DYNASMART–P and DynaMIT–P, which were originally developed for the Federal Highway Administration (FHWA) to support intelligent transportation system (ITS) deployment studies. Both platforms combine particle-based mesoscopic simulators with pathfinding algorithms for traveler route choice decisions; however, certain important details differ, with implications for the ability to represent various aspects of CV deployment. DYNASMART–P and DynaMIT–P have also formed the basis of online Traffic Estimation and Prediction Tools (TrEPS) intended for real-time traffic prediction to support traffic management functions; most notable road weather applications of TrEPS tools include the FHWA-supported deployment in Salt Lake City(1) and the Integrated Modeling for Road Condition Prediction (IMRCP) project in Kansas City.(2) There are several offshoots of the original DYNASMART–P framework, including VISTA, DynusT, DTALite, and DIRECT—all share the same modeling philosophy, though with possibly important differences for weather and CV impact modeling. Of these, only DYNASMART has been configured, calibrated, and deployed to specifically capture the impact of road weather on traffic operations. (See references 2, 3, 4, 5, 6, 7, and 8.)

In addition to university-generated tools, commercial platforms for meso-level DTA have emerged, generally as a complement to static macroscopic assignment tools, or as add-ons to microscopic simulators. Examples include TransModeler® (TransCAD®), Dynameq® (Emme), Cube Voyager, and Visum (related static platforms by the same vendor in parentheses). There is considerable variation across the commercial packages, which can be somewhat opaque in the absence of documented refereed publications describing these tools. This is a limitation for CV-related development, which requires detailed knowledge of and access to specific algorithmic components.

A third category of simulation-based network modeling tools, originally intended as agent-based activity-based demand models, includes the FHWA-funded TRANSIMS and its evolution into MATSim in Switzerland. The latter adopts a non-standard cellular-automata traffic flow representation that is not necessarily consistent with traffic flow theories, but allows fast computation for large networks, albeit when not seeking to reach equilibrium states. It also allows flexibility to route agents and execute elaborate rule-based activity schedules.

In addition to DTA platforms, which have largely supplemented or co-opted static macro network tools for new model investment by agencies, the other category of simulation platforms consists of microscopic simulation tools primarily intended for traffic operational applications. Originating in the 1970s with FHWA-supported NETSIM, which subsequently evolved into the current CORSIM, the domain experienced substantial commercial growth with the advent of ITS and adaptive signal control strategies that required fine-grained representation for design and evaluation. Three primary commercial platforms appear to dominate the marketused internationally are Vissim, Aimsun, and PARAMICS. Like CORSIM—and NETSIM before it—these are time-based, discrete event, discrete particle simulators with heavy reliance on Monte Carlo methods to generate random variable realizations of a driver’s every maneuver. With similar underlying logic (albeit different specific behavioral rules for drivers), the products have sought to differentiate through the quality, look, features, and functionalities of their graphical user interfaces.

Other traffic operational microscopic platforms have also been developed and have gained some traction, usually in specialized markets. These include TransModeler microscopic simulator, which is patterned in part on MITSim, and is built on a TransCAD network; INTEGRATION, which evolved from a mesoscopic version to a microscopic platform; Cube Avenue, and the open-source SUMO (Simulation of Urban MObility) developed at the German DLR Institute of Transportation Systems.

Off-the-shelf commercial packages for either strategic or operational applications are generally not capable of representing the particular aspects of CVs that impact both operational performance and users’ behavior. In some instances, modification of certain aspects through application programming interfaces (API) is possible, but control of how the API is used in the overall simulation is generally not available. For this reason, researchers have developed special-purpose tools focused on the particular questions of interest to their scope of the study. These are typically not comprehensive or integrated platforms, but simplified representations in all but those aspects essential to the question of interest.

A key question for developers and agencies interested in developing an AMS capability for CVs is whether to add CV capabilities to existing platforms, thereby taking advantage of graphical user interfaces and other useful components, or whether to integrate a special-purpose tool into a larger, custom-targeted platform built around those capabilities. The simple answer is: it depends on several factors, including the structure and logic of the platform itself, and the degree to which the software could accommodate the desired features. For an agency, tool selection also depends on budget, tool availability, existing models, and staff expertise.

Functional Requirements

To guide the assessment of existing AMS tools for the purpose of this study, the following list of functional requirements was identified:

Ability to represent weather events
- Assumes the ability to represent different weather scenarios and impacts thereof in simulation (rain, snow, heavy snow) via built-in capability
Facility types (freeways, arterials, etc.)
Road weather management strategies
Traffic and operational conditions
- Traffic events such as incidents and work zones
Weather and traffic parameters
CV data/deployment parameters
Transportation system performance measures
Assumptions (i.e., impacts of CV information/data)
Calibration and validation requirements
Computing/programming requirements to use the tool
Corridors/networks where they have been used and transportation agencies involved
Applicability/transferability to different corridors/networks
Availability of the tools for general/public use
Telecommunication aspects of CVs, and performance under bad weather
Use of CV data for estimation and prediction
Ability to use CV data for management strategies

Assessment of Tools against Above Requirements

CV technology is expected to affect the operational performance of transportation systems in different aspects,⁽⁹⁾ including safety,^(10,11) mobility,^(12,13) and sustainability.⁽¹⁴⁾ The technology is expected to improve traffic safety through reduction of accidents caused by human error, increase throughput⁽¹⁵⁾ through driving at higher densities with the help of highly responsive CVs, and improve traffic control⁽¹⁶⁾ at intersections (see references 17, 18, 19, 20, 21, 22, 23, and 24) through wireless communications.

To evaluate those impacts effectively, the distinct behavior of CVs⁽²⁵⁾ needs to be captured in the AMS tools. Given the required detail at the individual vehicle level, the logical type of methodology consists of traffic microsimulation. Microsimulation provides the highest degree of detail in capturing the characteristics of CVs, including car-following behavior, lane-changing, sensor range, wireless communications,⁽²⁶⁾ reaction time, etc. Microsimulation is the only type of simulation capable of simulating mixed traffic conditions at different CV market penetrations as each vehicle is simulated individually. Therefore, most of the prior/current studies on the operational performance impacts of CVs relied on microsimulation tools. The main limitation for this type of simulation is the computing power it requires to process and analyze the high amount of detail associated with the simulated vehicles. This can limit the amount of time and the network (area of study) size for which simulations are run.

For strategic-level CV analyses of large regional networks, detailed microsimulation of all traffic maneuvers is unnecessary and impractical. Developing macroscopic relations for either facilities or networks requires observation of actual systems at different penetration levels of the technologies, which is not possible under the current situation because these technologies remain in very early stages of test deployment. It is possible to rely on microsimulation experiments conducted for facilities and subnetworks to produce macroscopic fundamental diagrams that could then be used in mesoscopic simulation-based network modeling tools. Mesoscopic models provide a fidelity that is in between microscopic and macroscopic models. A recent example of incorporating the market penetration of CVs in a mesoscopic tool was illustrated by Mittal et al.;⁽²⁷⁾ the input speed-density parameters were generated using a special-purpose microsimulation tool. Trajectories obtained from the network simulations then formed the basis for calibrating link-level as well as networkwide fundamental diagrams (NFD).

In addition to modeling mixed flow impacts of CV systems, modeling emerging traffic control and management strategies enabled by the new technology is also challenging.⁽²⁸⁾ An important aspect of emerging control algorithms is wireless telecommunications. However, most AMS tools lack an abstract representation of telecommunications, its performance, and its impacts on driving behavior in their models. Microsimulation is also used in this case for modeling those strategies, as it provides enough details to capture the interactions between the vehicle’s control devices (or lack of them) and the infrastructure.

Commercially available simulation tools, such as Vissim and Aimsun, recently introduced the capability to model CV systems, and also have means to enable users to code their own models; this is the primary mechanism used by researchers to evaluate CV alternatives with these tools. However, the parameters for these models would still need to be assumed, as the data required to calibrate them remain difficult to obtain.

In addition to coding special CV characteristics into existing tools, some researchers have developed simulation platforms specifically designed to model mixed traffic with CV systems. The tools included are those that meet certain thresholds of usability and responsiveness to the main considerations in this project (road weather, CVs). The review will describe how these tools address the functional capabilities.

The assessment of simulation tools involves two major aspects: (1) modeling connected environments and their impact on overall system’s operational performance and (2) adverse weather-related behavior and management strategies. To the best of the authors’ knowledge, table 1 assesses all mesoscopic simulation tools, and table 2 assesses all microscopic tools..

The functional requirements in chapter 2 are grouped into three categories reflecting the study objectives:

Weather event modeling and related management strategies
Capabilities pertaining to modeling CVs and connected environments
General simulation capabilities and features

Each category corresponds to a block of rows in table 1 and table 2, with specific functionalities corresponding to individual rows within each block.

Table 1. Mesoscopic simulation tools assessment based on identified requirements.
Capability/Requirement	General Platforms	DYNASMART (P, X)	DynaMIT (P, X)	DTALite	MEZZO	DynusT	Dynameq®
Category: Weather
Represent weather events (rain, snow, heavy snow)—system integrates external operational conditions and simulates effects on transportation system performance under different scenarios.	NO	YES	YES	NO	NO	NO	NO
Represent road weather management strategies—system has built-in capability to model and execute management strategies: Variable message signs (VMS) Variable speed limits (VSL) Snowplow routing (SPR) De-icing application (DIA)	NO	YES (VMS, VSL, SPR) PARTIAL (DIA)	NO	PARTIAL (VMS, VSL); not road-weather specific NO (SPR, DIA)	NO	PARTIAL (VMS, VSL); not road-weather specific NO (SPR, DIA)	NO
Represent weather with respect to traffic parameters.	YES	YES	YES	YES	YES	YES	YES
Category: CV Environment
Model different effects of telecommunication technologies on connected drivers.	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions
Model sensor performance and reliability aspects that directly influence vehicle performance.	NO	NO	NO	NO	NO	NO	NO
Integrate information flow through V2I/V2V/V2X communications within AMS system.	NO	NO	NO	NO	NO	NO	NO
Use new data sources from CV systems, and traditional data sources, (e.g., loop detectors) for traffic estimation and prediction.	NO	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing
Use new data sources from CV systems, traditional data sources (loop detectors), or management strategies. Calibrate/recalibrate model parameters of connected-manual driving as actual trajectories are available.	NO	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing
Category: General
Model different facility types (freeways, arterials, etc.).	YES	YES	YES	YES	YES	YES	YES
Model various traffic and operational conditions—system enables users to define different operational scenarios to be simulated.	YES	YES	YES	YES	YES	YES	YES
Retrieve transportation system performance measures.	Scenario comparison (highway networks, public transit services & socioeconomic), transport performance statistics, matrix histograms, environmental analyses (noise, emissions), accident data analysis*	Network-, link-, and vehicle-level performance indicators	Travel guidance: departure time, travel mode, and route; network/link performance indicators: including travel time, flows, speeds, and densities	Network/link-level average values of traffic measurements: flows, densities, speeds, travel times, vehicle counts, and queues	Link-level average speed, inflow/outflow, densities and queue length; vehicle-level travel time, distance traveled, and route switched; OD pair summary statistics (vehicle generated, arrived, total travel time, and distance travelled)	Link/lane-level and path-level average values of traffic measurements: flows, densities, speeds, travel times, vehicle counts, and queues	Link/lane-level and path-level average values of traffic measurements: flows, densities, speeds, travel times, vehicle counts, and queues
Post-processing of outputs.	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable
Calibration and validation requirements.	YES	YES	YES	YES	YES	YES	YES
Applicability/transferability to different corridors/networks.	YES	YES	YES	YES	YES	YES	YES
Corridors/networks where they have been used, and transportation agencies involved.	Various	Various	Various	Various	Various	Various	Various
Availability of the tools for general/public use (open source).	NO	YES	NO	YES	YES	YES	NO
Ability to modify or build on source code (API available).	Can be modified via scripting language (in most cases)	Codes can be modified	Codes can be modified	Open source; codes can be modified	Open source; codes can be modified	Open source; codes can be modified	Using a Python®-based API to implement advanced strategies
Cost to modify API.	High	High	High	High	High	High	High
Computational power (runs on regular PC).	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size

Table 2. Microscopic simulation tools assessment based on identified requirements.
Capability/Requirement	Vissim®	Aimsun®	PARAMICS	TransModeler®	NUTC Platform	FHWA-STOL/UCLA Platform	PATH Platform	ECO-CACC Integration
Category: Weather
Represent weather events (rain, snow, heavy snow)—system integrates external operational conditions and simulates effects on transportation system performance under different scenarios.	NO	NO	NO	NO	NO	NO	NO	NO
Represent road weather management strategies—system has built-in capability to model and execute various management strategies: Variable message signs (VMS) Variable speed limits (VSL) Snowplow routing (SPR) De-icing application (DIA)	PARTIAL (VSL); not road-weather specific NO (VMS, SPR, DIA)	PARTIAL (VSL); not road-weather specific NO (VMS, SPR, DIA)	PARTIAL (VMS, VSL); not road-weather specific NO (SPR, DIA)	NO	PARTIAL (VSL); not road-weather specific NO (VMS, SPR, DIA)	NO	NO	NO
Represent weather with respect to traffic parameters.	YES	YES	YES	YES	NO	YES	NO	NO
Category: CV Environment
Model different effects of telecommunication technologies on connected drivers.	NO	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions	YES; V2I and V2V	YES; V2I and V2V	No representation of wireless telecommunications in current versions	No representation of wireless telecommunications in current versions
Model sensor performance and reliability aspects that directly influence vehicle performance.	NO	NO	NO	NO	YES	NO	NO	NO
Integrate information flow through V2I/V2V/V2X communications within the AMS system.	NO	NO	NO	NO	YES	NO	NO	NO
Use new data sources from CV systems, and traditional data sources (e.g., loop detectors), for estimation and prediction.	NO	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing
Use new data sources from CV systems, and traditional data sources (e.g., loop detectors) or management strategies. Calibrate/recalibrate model parameters of connected-manual driving as actual trajectories become available.	NO	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing	Yes; might need pre-processing
Category: General
Model different facility types (freeways, arterials, etc.).	YES	YES	YES	YES	YES	YES	YES	YES
Model various traffic and operational conditions—users define different operational scenarios to be simulated.	YES	YES	YES	YES	YES	YES	YES	YES
Retrieve transportation system performance measures.	Scenario comparison statistics of transport performance, matrix histograms, environmental analyses (noise, emissions), crash data*	Network-, link-, and vehicle-level performance indicators	Travel guidance: departure time, travel mode, and route; network/link performance indicators including travel time, flow, speeds, and densities	Network/link-level average values of traffic measurements: flows, densities, speeds, travel times, vehicle counts, and queues	Link-level average speed, inflow/outflow, densities, and queue length; vehicle-level travel time, distance traveled, and route switched; OD pair summary statistics	Link/lane-level and path-level average values of traffic measure-ments: flows, densities, speeds, travel times, vehicle counts, and queues	Link/lane-level and path-level average values of traffic measure-ments: flows, densities, speeds, travel times, vehicle counts, and queues	Link/lane-level and path-level average values of traffic measure-ments: flows, densities, speeds, travel times, vehicle counts, and queues
Post-processing of outputs required.	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable	YES, if outputs not configurable
Calibration and validation requirements.	YES	YES	YES	YES	YES	YES	YES	YES
Applicability and transferability to different corridors/networks.	YES	YES	YES	YES	YES	YES	YES	YES
Corridors/networks where they have been used, and transportation agencies involved.	Various	Various	Various	Various	Various	Various	Various	Various
Availability of the tools for general/public use (open source).	NO	NO	NO	YES	YES	YES	YES	NO
Ability to modify or build on source code (API available).	Can use API via Python®, Microsoft® Visual Basic, C, C++, and others to implement advanced strategies/Vissim–COM program-ming	Can use Aimsun micro API to implement advanced strategies/applications	Can use script language or API to implement advanced strategies	Open source; codes can be modified	Open source; codes can be modified	Can use API via Python, Microsoft Visual Basic, C, C++, and others to implement advanced strategies/Vissim–COM programming	—	—
Cost to modify API.	Moderate	Moderate	Moderate	High	High	High	High	High
Computational power (runs on regular PC).	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size	x-server, depending on model size

The following conclusions emerged from the assessment of available capabilities against the functional requirements identified for AMS to support use of CVs for road weather management:

The majority of mesoscopic tools are not weather-ready, except for DYNASMART– (P, X), which has been applied in several road-weather related studies, and, to some degree, DYNAMIT, which has been used with limited calibration in one previous weather-related study. However, virtually all existing mesoscopic tools offer the possibility of specifying traffic model parameter values that have been calibrated to weather conditions as input values.
Virtually all off-the-shelf microscopic simulation tools are not explicitly weather-ready. However, they all allow users to specify parameter values for the driver behavior models that reflect adverse weather. For example, the Federal Highway Administration-Saxton Transportation Operations Lab/University of California, Los Angeles (FHWA-STOL/UCLA) tool has adopted calibrated driver behavioral models using naturalistic driving data to reflect various weather features.
Representation of road weather management strategies is available only partially to the mesoscopic simulation user, except for DYNASMART, which has been applied in previous studies to analyze the impact of road-weather related variable message signs (VMS), variable speed limits (VSL), snowplow routing (SPR) and de-icing application (DIA). Most models offer some form of VMS with limited integration of specific behavior related to road weather messaging.
Microscopic simulation tools offer VMS and VSL modeling capability at the individual vehicle level, with road-weather features modeled through driver behavioral parameters. SPR and DIA capabilities have not been evaluated through microscopic tools in the literature, though similar approaches as used for VMS/VSL can be applied for small networkwide SPR and DIA.
CV features are beginning to emerge for both microscopic and mesoscopic simulation approaches. However, calibration remains limited. Most development has taken place through modification to microscopic simulation tools, particularly the car-following rules and lane-changing mechanisms that govern human drivers.
For mesoscopic tools, one application has relied on microscopic simulation to generate performance profiles under different CV market penetration rates, which serve as fundamental diagrams relating speed to density in the meso model.
There remain gaps in simulation of CV capabilities with respect to data collection and weather-related management; at the microscopic levels, we can expect availability of such features to take place over time, though the absence of data for calibration of the micro level remains a limitation.

In summary, application of either mesoscopic or microscopic AMS tools requires additional adaptation of the tools in question. At the meso level, DYNASMART has extensive cumulative experience in road weather applications. At the micro level, most off-the-shelf tools are not configured for the combined impact of CVs and road weather, but they offer flexibility to implement such capabilities through APIs and other mechanisms. The recent FHWA-STOL/UCLA platform has developed customized functions for CV and weather features and has been used for CV/automated vehicle (AV) applications.

Review of Selected Analysis, Modeling, and Simulation Tools

Describing and predicting roadway conditions and events that may impact travel across a road network require a broad range of data and analytical tools. These data and tools span multiple scientific and engineering disciplines. The assessment in the previous section addressed the most commonly available mesoscopic and microscopic modeling tools. This section only includes recent special-purpose tools developed specifically to capture CV system features and/or weather road traffic management.

Traffic condition and network performance models describe and predict the operational state of the traffic system, and its future evolution over time and space. The traffic state characteristics typically of interest consist of the traffic volume (flow), traffic density, speed along the different parts of the network, and the variation of these characteristics over time.

Modeling and prediction of traffic condition may be performed at the single link, freeway system, major arterials, corridor, and network (or subnetwork) levels over the short term, such as a 15-minute prediction, or they may extend to medium and longer terms on the order of hours or possibly days. Traffic condition and network performance models usually combine current traffic state observations (from sensors) with historical traffic data to generate more accurate predictions than would be possible with either current observations or historical information alone. Traffic system performance is not merely a repeat of past occurrences, but is influenced by the unique situations and prevailing events that day or at any given time. However, history can be a guide and starting point from which to pivot and adjust prediction to reflect the measured state of the system and unfolding events.

Given the role and recurrence of historical patterns, data-driven approaches based on statistical models and artificial intelligence methods play a role in data mining and pattern recognition and matching. However, without sufficient data coverage and—more importantly—when the identified patterns are disrupted by exogenous factors and events (e.g., severe weather, occurrence of one of more crashes in a given location), it is necessary to go back to the underlying systematic effects and relations rooted in traffic physics and behavioral modeling to predict the evolution of traffic flows in the system under consideration. Hence the rationale for prediction approaches based on traffic modeling and simulation, used as a basis for control actions and predictive interventions to mitigate the negative impact of these events. The integration of statistical and data-driven models with simulation-based approaches has recently emerged as a direction for development.

Special Purpose Tools and Platforms

Integrated Wireless Telecommunication Traffic Simulation – NUTC

The integrated wireless telecommunication traffic simulation platform developed by Talebpour et al.⁽²⁹⁾ is a special-purpose tool for simulating mixed traffic conditions on freeways in a connected environment. The platform integrates three different driving behaviors: regular (non-connected and non-automated) vehicles, CVs, and AVs, in addition to modeling vehicle-to-infrastructure (V2I)/vehicle-to-vehicle (V2V) wireless telecommunications. For regular vehicles, the platform uses a stochastic car-following model introduced by Hamdar et al.⁽³⁰⁾ and extended by Talebpour et al.⁽³¹⁾ The model is based on the Prospect Theory⁽³²⁾ and captures drivers’ crash-avoiding behavior while maintaining a desired speed. For modeling CVs, the platform relies on a deterministic car-following model introduced by Kesting et al.⁽¹⁵⁾ Finally, for AVs, the platform uses a car-following model based on the previous simulation studies by Van Arem et al.⁽³³⁾ and Reece and Shafer.⁽³⁴⁾ For lane-changing, the platform uses a game-theoretical approach that endogenously captures the effects of additional information through wireless communications on lane-changing decisions.

With respect to modeling wireless telecommunications, the Node Mobility Model (ns–3) was integrated with the microscopic vehicular traffic simulation framework. Thus, the positions of the vehicles are governed by the micro rules in the simulator, including whatever messages may be received, as transmitted by ns–3 to the vehicles in their evolving positions. Thus, the integrated simulation tool addresses some of the limitations of general platforms by explicitly modeling wireless telecommunications and its impact on the driving behavior of CVs. The tool is restricted to modeling freeway sections, which limits its capability of evaluating CV impacts on arterials or urban networks.

FHWA-STOL/UCLA CAV Simulation Tool – STOL/UCLA

FHWA-STOL has developed a series of Vissim driver model APIs (and component object model [COM] codes) for modeling different CAV applications, such as cooperative adaptive cruise control (CACC), signalized intersection approach and departure, cooperative merge, and speed harmonization. This platform was originally developed as an FHWA simulation CV/AV simulation guidebook, and has been gradually enhanced to address the following needs:

Modeling AV behavior: calibrated using available field experiment data; cooperative automation considers collaborative behavior between vehicles, e.g., platooning, cooperative merge, and eco-approach for departure at signalized intersections.
Modeling wireless communication for CV applications to consider potential communication performance (package delay, drops): communication model calibration using the data collected at FHWA-STOL and Michigan Safety Pilot, and considers communication performance under various conditions.
Reflecting weather conditions through calibrated driver behavioral parameters: recent research at the University of Wyoming and Turner-Fairbank Highway Research Center (TFHRC) used Strategic Highway Research Program 2 (SHRP2) Naturalistic Driving Study data⁽³⁵⁾ to capture driving behavior in a variety of weather conditions. The calibrated Wiedemann 1999 car-following model has been incorporated into the tool.

Connected Vehicle Traffic Simulation Tool – PATH

The CACC traffic simulation tool developed by PATH⁽³⁶⁾ is a microsimulation model implemented using the Aimsun API and Micro–SDK. It represents multilane freeway traffic operations with varying levels of market penetration of vehicles equipped with adaptive cruise control (ACC), CACC, and close-formation platoon capabilities. The major components of the microscopic traffic model include the vehicle dispatching model, human driver model, and ACC/CACC model. The vehicle dispatching model determines how a modeled vehicle enters the simulation network and the distribution of different types of vehicles across the multilane highway. It is intended to generate very high volumes of vehicles at the source section, under steady state conditions, as it is essential for simulating CACC strings that have much shorter time gaps between consecutive vehicles.

The human driver model is built upon the basic framework of the Next Generation Simulation (NGSIM) oversaturated flow model proposed by Yeo et al.,⁽³⁷⁾ which partitions the driver’s car-following and lane-changing behavior into multiple fundamental driving modes. The ACC/CACC driving behavior models in the tool are based on empirical models developed by Milanes and Shladover,⁽³⁸⁾ in which CACC-equipped vehicles can form strings with short gaps. Drivers of CACC-equipped vehicles can also exit their closely coupled string and switch off CACC to make lane changes or exit the freeway. Although the CACC system implementation relies on information received from the leading vehicle in the CACC string as well as from the immediately preceding vehicle, the empirical models used in the simulation provide a simplified description of the closed-loop vehicle-following dynamics.

ECO-CACC Traffic Simulation Tool (Extension of INTEGRATION) – Virginia Tech

The ECO-CACC traffic simulation tool is an extension⁽³⁹⁾ of the microsimulation tool INTEGRATION that was developed by Rakha et al.^(40,41) The objective of the implemented ECO-CACC system is to reduce fuel consumption of CACC vehicles driving through multiple intersections. It uses traffic signal phasing and timing (SPaT) data transmitted through V2I communications and vehicle queue predictions to estimate fuel-optimal speed limits. The advisory speed limits are then sent to CACC vehicles via V2I communications to prevent them from completely stopping at intersections.

The general INTEGRATION platform is a microscopic traffic assignment and simulation tool⁽⁴⁰⁾ that allows detailed analysis of vehicle movements and lane-changing maneuvers on a network. Its main capabilities include detailed estimation of vehicle fuel consumption and emissions and the expected number of crashes using a time series model. The traffic assignment module within the platform uses different heuristics to assign vehicles to links on the network. The cruise control behavior emulated by ECO-CACC is different from typical CACC systems. Rather than vehicles communicating to form platoons, ECO-CACC vehicles communicate only with traffic signals when driving near intersections through V2I technology. This limits the tool’s usefulness to evaluate general CACC systems. The tool also lacks the capability to model the information flow through wireless telecommunications and its effect on driving behavior.

DYNASMART Integrated Platform

DYNASMART is a (meso) simulation-based intelligent transportation network planning tool. The model can be configured to run offline or online. The offline model (DYNASMART–P) includes dynamic network analysis and evaluation, and the online model (DYNASMART–X) provides short-term and long-term prediction. DYNASMART–P models the evolution of traffic flows in a traffic network resulting from the travel decisions of individual drivers. It is designed for use in urban areas of various sizes and is scalable, in terms of the geometric size of the network, with minimal degradation in performance. DYNASMART–P can also model the fine details of transportation networks such as zones, intersections, links, origins, and destinations. Inheriting the core simulation components from DYNASMART–P, the primary feature of the online operational tool (DYNASMART–X) is the capability of interacting with multiple sources of information and providing estimates of network traffic conditions and predictions of network flow patterns.

In a previous active traffic demand management (ATMD)/dynamic mobility application (DMA) project,⁽³⁾ DYNASMART was used to model, test, and evaluate weather-responsive operations and several potential strategies in the Chicago testbed. During inclement weather events, traffic flow pattern and behaviors had changed, and thus initial strategies and applications for normal weather should have also been modified to mitigate weather impact on traffic. The system supports operators’ decision-making in deploying alternative strategies given specific operational conditions. Various potential strategies were designed and tested, with particular focus on snowplow routing and speed harmonization.

Analysis, Modeling, and Simulation Tools Summary

Several AMS tools have been proposed to study the traffic impacts of CV systems. Those tools include general-purpose, commercially available AMS platforms in addition to customized research models developed to answer specific questions related to CV systems. There has been particular growth in the latter category.

Weather impacts on traffic depend on the severity of rain, snow, or other conditions. Researchers have already incorporated or have begun to include weather impacts in analyses using traffic simulation modeling tools. Adjustable weather factors allow these models to simulate realistic traffic situations in inclement weather. Furthermore, weather databases provide adequate weather data to deploy the weather module for traffic operations and management. Micro-level simulation platforms Aimsun and Vissim are also modeling and capturing the effect of adverse weather conditions by adjusting car-following and lane-changing behavior of vehicles. Such efforts require extensive site- and configuration-specific parameter calibration.

Microsimulation tools are the dominant methodology for evaluating CV operational impacts; researchers have either sought to extend general-purpose platforms through programming interfaces and development kits or have built their own prototype simulation tools. Most of the existing microsimulation tools with CV capabilities can model mixed traffic flows and CV related policies/controls; however, several other key features remain lacking relative to modeling of weather impacts on CVs.

CV features are beginning to emerge for both microscopic and mesoscopic simulation approaches. However, calibration remains limited. Most development has taken place through modification to microscopic simulation tools, particularly the car-following rules and lane-changing mechanisms that govern human drivers. There remain important gaps facing simulation of CV capabilities, with respect to data collection and weather-related management; at the microscopic levels, we can expect availability of such features to take place over time, though the absence of data for calibration of the micro level remains a major limitation.

In summary, application of either mesoscopic or microscopic AMS tools requires additional adaptation of those tools. Several promising examples have been identified to provide a starting point for an AMS capability to support CV-enabled road weather application. These require varying degrees of modification, adaptation, and calibration for effective deployment to support agency decision-making in adverse weather conditions.

previous | next