Methodological Foundations of the Traffic Simulation Process | International Journal of Microsimulation

Cite this article as: J. Díaz-Ramírez, E. Nobil, J. Alberto Estrada-García, J. Adrián Martínez-Hernández, J. Ignacio Huertas; 2026; Methodological Foundations of the Traffic Simulation Process; International Journal of Microsimulation; 19(2); 37-69. doi: 10.34196/ijm.00342

Article
Figures and data
Jump to

Abstract

There is an urgent need to design and evaluate strategies that improve the mobility of people and goods in urban regions while reducing energy consumption and vehicular emissions. Traffic simulation is an increasingly powerful tool to address this challenge; however, it requires extensive input data and involves multiple steps and decisions, which can make the process complex. This paper provides researchers and practitioners with a comprehensive methodological framework for modeling and simulating urban traffic, along with key insights and practical recommendations. These contributions are grounded in both the literature and the authors’ experience across multiple traffic simulation applications, enabling the paper to offer guidelines that facilitate the implementation of urban traffic simulation while identifying areas that warrant further research. Unlike previous reviews, this study synthesizes practical lessons from real-world microsimulation projects and integrates methodological guidance applicable across different simulation scales.

1. Introduction

Traffic simulation is increasingly used by companies and institutions to evaluate scenarios and analyze urban mobility systems. A wide range of software tools support these applications, including SUMO, PTV VISUM, and PTV VISSIM. From a modeling perspective, substantial progress has expanded the available approaches for simulation and analysis. Practical guidelines developed in the early 2000s, such as the toolkits by the US Department of Transportation (Alexiadis et al., 2004; Dowling et al., 2004; Holm et al., 2007), remain widely used; however, they do not fully integrate current technical developments and require updates to reflect advances achieved over the past two decades. Methodologically, numerous approaches capture specific aspects of reality, yet recent developments have blurred the distinctions between approaches, tools, and applications. Although several toolkits exist, few integrate scale selection, data availability, and energy indicators within a unified framework.

Therefore, this study examines traffic simulation by outlining key steps and guidelines to support its practical application, while providing a basis for further development throughout the simulation process. The document follows a structured, step-by-step approach. It begins with an overview of traffic simulation and common modeling approaches, and then reviews available software tools, including a comparative discussion of their characteristics, applications, computational requirements, and key features relevant to implementation.

The study aims to serve as both a methodological framework and a practical guide for urban traffic simulation. It presents a software-independent approach that supports the analysis of mobility patterns and the evaluation of strategies across multiple spatial scales. In addition, by incorporating vehicle energy efficiency–based metrics within a broader mobility perspective, it shifts the focus from vehicle-centered analyses toward the characterization of people’s mobility patterns.

The manuscript is organized as follows. It first addresses simulation scale, defining scope and granularity, followed by a modeling approach for estimating travel demand using structured statistical procedures. It then analyzes simulation software tools, highlighting their strengths, limitations, and appropriate contexts. Recent advances are incorporated throughout, and the study concludes by identifying areas for future research.

2. Traffic Simulation Framework

Recognizing that data type, data quality, and resource availability critically influence the selection of an appropriate simulation model, this section first reviews the classification of traffic simulation models by scale, then introduces a general methodological framework applicable across scales, and finally outlines the fundamental principles for representing driving patterns.

2.1. Traffic Simulation Scales

Simulation models are classified into three categories, microscopic, mesoscopic, and macroscopic, according to their spatial and temporal resolution, as summarized in Table 1.

Table 1

Classification of traffic simulation models by scale

Scale	Level of detail	Region size	Mobility modeling	Assumptions	Modeling principles	Underlying algorithms	Software options
Microscopic	Each vehicle	Small city districts (e.g., freeway, arterial, corridor, etc.)	Individual dynamics of vehicles in intersections, specific roads, or small city zones.	Individual vehicles make decisions to get to their destination.	Individual mobility decisions in networks will represent the mobility system.	Longitudinal model, Lateral model, intersection model	Aimsun, Transmodeler, SUMO, PTV VISSIM, MATSIM, Cube Dynasim
Mesoscopic	Each vehicle. Movement is ruled by the average speed on the travel link.	City districts	Flows of vehicles between city zones by type of activity.	Vehicles moving in queues	Queues of individual vehicles that follow shockwave dynamics.	Longitudinal model, intersection model	Aimsun, Transmodeler, Mezzo Polaris, SUMO Meso
Macroscopic	Average vehicle dynamics, such as traffic density and average speed.	Entire cities	Flows of traffic between largely aggregated zones.	Aggregated vehicles flowing by main roads.	Aggregated vehicles flowing through roads based on fluid particle dynamics.	Flow model	Aimsun, Transmodeler, PTV VISSUM

Alexiadis et al., 2004; Algherbal and Ratrout, 2025; Azam et al., 2019; Lopez et al., 2018.

The selection of an appropriate modeling scale for a given application depends on multiple factors, including the required level of detail, availability and granularity of input data, the traffic performance indicators to be evaluated, computational resources and time constraints, capabilities of the simulation platform, and the scope of the decision-making objectives. For example, (Varga et al., 2023) highlights that the mesoscopic scale often balances resolution and computational cost, making it suitable for corridor-level applications instead of full network microsimulation.

Microscopic simulation models, or just microsimulation, represent traffic at the level of individual vehicles, capturing movement through dynamic variables such as position and velocity and explicitly modeling driver behavior and interactions (Alexiadis et al., 2004; Janson Olstam and Tapani, 2004; Nor Azlan and Md Rohani, 2018). Powered by car-following and lane-changing models, their main strength lies in the detailed representation of the traffic, enabling a multidimensional analysis. Examples are: localized infrastructure design, such as the evaluation of multiple design scenarios of a multi-lane roundabout in a small-scale context (Guerrieri and Khanmohamadi, 2024); the analysis of congestion dynamics (Uthpala et al., 2023), traffic management strategies, such as the effect of path widths by modeling cyclists’ complex behaviors like maneuver decisions and lateral movements, into the bicycle flow capacity (Brunner et al., 2024) or the analysis of bicycle flow characteristics in relation to infrastructure design (Pérez Castro et al., 2025); the impact of driving risk when a shockwave propagates in a queue at signalized intersections (Wang et al., 2025); the impact of an imperfect perception on the traffic flow dynamics, to be used in the study of mixed traffic of human and automated driving (Postigo et al., 2025); and the implementation of Intelligent Transportation Systems (ITS) (Alexiadis et al., 2004). This level of detail, however, entails high computational and data requirements, which limit the spatial extent of applications. A recent review of advancements in traffic simulation for road safety highlights the need for refined models that capture real-world traffic dynamics to improve predictive capabilities, further enhanced by integrating other models and machine learning algorithms (Mustapha et al., 2024).

At the other end of the spectrum, macroscopic models describe traffic using aggregate variables such as vehicular density, average flow, and the mean speed of a traffic stream (Alexiadis et al., 2004; Ding, 2011). These models operate at the level of are road segments rather than individual vehicles and do not explicitly represent behavioral components such as trip generation or mode choice, and validation involves replication of observed congestion patterns (Krista et al., 2004). Their reduced complexity allows the analysis of large-scale systems, including entire urban areas, and facilitates the use of aggregated data sources for calibration and validation. These models offer great capability to understand and make decisions on a large scale for strategic planning and the evaluation of network-wide conditions.

Between these two levels, mesoscopic models provide an intermediate representation by simulating individual vehicles (i.e., emissions, travel times, etc.) while governing their dynamics through aggregate relationships (i.e., speed-density or flow-density relationships that can be defined with few parameters) (Aimsun, 2025; Holm et al., 2007). This balance between detail and computational efficiency makes them suitable for applications at the corridor or subnetwork level. As examples, a scenario simulation framework based on vehicle trajectory big data was used to evaluate the impact of shifts in residential travel modes on carbon emissions, identifying peripheral areas as having the most significant carbon reduction potential (Li et al., 2025); and a mesoscopic traffic simulation model was developed to assess the vulnerability of road infrastructure in an urban area to closure scenarios caused by flooding (Marian et al., 2024).

2.2. Traffic Behavior Models

Key traffic behavior models, mainly considered in microscopic and mesoscopic simulations include gap-acceptance, speed adaptation, lane-changing, ramp merging, overtaking, and car-following (Janson Olstam and Tapani, 2004). Gap acceptance models determine the minimum acceptable spacing between vehicles, while speed adaptation refers to the adjustment of a vehicle’s speed at a given network location to conform to the road’s design speed. Lane-changing models describe drivers’ behavior governing lateral movements between lanes, and finally, car-following models describe drivers’ behavior governing longitudinal movements along the road by linking a vehicle’s motion to its interaction with preceding vehicles. Movement is determined by factors such as desired, average, and maximum speeds, responses to leading vehicles, and the conditions under which lane changes occur when necessary or feasible.

In mesoscopic simulations, car-following models typically neglect acceleration and deceleration, assuming that vehicles are either stopped or traveling at their desired speeds. This simplification can be significant in applications where traffic control devices strongly influence flow. In such cases, the absence of acceleration dynamics may lead to flow overestimation, which must be compensated through calibration of other parameters. As a result, mesoscopic simulation is primarily applied to large-scale road networks, offering reduced effort in network construction and calibration compared to microscopic models, particularly when aggregate flow effects can be adequately represented through parameter adjustment (Aimsun, 2025; Sun et al., 2020). Sun et al. (2020) summarized recent applications of mesoscopic simulation for assessing the impact of transportation strategies and system response to emergency scenarios in relatively large urban zones.

Table 2 presents a summary of the fundamentals of the most common car-following models used in traffic simulation tools. They are classified in models based on the safe distance premise to avoid collisions, models that consider the drivers’ perception and their reactions to stimuli from the vehicle ahead, and other more recent approaches. For reference, SUMO includes Krauss model as the default option together with alternatives such as the Intelligent Driver Model (IDM) and Wiedemann-like variants, while VISSIM relies on Wiedemann models.

Table 2

Car-Following traffic models

Class	Models	Description
Gap acceptance	Gipps model (Gipps, 1981)	Based on the premise that each driver sets limits on their desired braking and acceleration rates. The speed is based on a safe following distance to avoid a possible collision with the vehicle ahead. The behavior is controlled with response times, break rates, and maximum desired speed. This model is widely preferred for simulation purposes (Ranjitkar and Kawamura, 2005).
Gap acceptance	Krauss model (Krauß, 1998)	Alike but simpler than Gipps’ model. A stochastic model based on the same assumption that vehicles move without colliding. The model is set up with the same parameters as in the Gipps model. It is implemented in the SUMO simulator (Kanagaraj et al., 2013; Ranjitkar and Kawamura, 2005).
Psycho-physiological driver behavior	Gazis-Hermann-Rothery (GHR) model	Referred to as the general car-following behavior. The relationship between a leader and a follower vehicle is a stimulus-response type of function where the follower vehicle’s acceleration is proportional to its speed, the speed difference with respect to the leader, and the space headway. This type of model assumes that the follower reacts to arbitrarily small changes in the relative speed (Janson Olstam and Tapani, 2004).
	Leutzbach and Wiedemann model (1986)	It takes into consideration the influence of the driver’s perception on velocity control. A driver can be in different driving modes, such as free driving, approaching, following, and braking. The driver switches from one mode to another as soon as they reach a certain threshold, which can be expressed as a function of speed difference, space headway, etc. Versions of this model are used in VISSIM. (Kanagaraj et al., 2013; Ranjitkar and Kawamura, 2005).
	Fritzsche model (1994)	This model also considers human perception in defining the model regimes. For example, there is a threshold for the minimum speed difference that the driver perceives. In addition, it incorporates four thresholds for the follower’s space headway to its leader: Desired distance, the risky distance, the safe distance, and the braking distance (Janson Olstam and Tapani, 2004).
Cell based model	Nagel-Schreckenberg model (cellular – automata) (Nagel et al., 1998)	A computationally efficient stochastic model that simulates the traffic flow using only integer operations. It incorporates imperfections of driving using a noise term in the update rules. The model deals with single-lane traffic flow of cars moving in a one-dimensional cellular chain under periodic boundary conditions, which considers the desired maximum speed, vehicle acceleration, deceleration, random delays, and update of vehicle location (Liu, 2012; Ranjitkar and Kawamura, 2005).
Trajectory based model	Newell model (2002)	Based on the premise that the driver of the following vehicle drives as a shifted space trajectory of the leader vehicle. The space trajectory of the following vehicle is the same as that of the leader vehicle except for a translation in space and in time. It is driven by a time lag and a distance lag (Ranjitkar and Kawamura, 2005).
Trajectory based model	Intelligent Driver Model (IDM)	IDM is a car-following model that describes the dynamics of the positions and velocities of single vehicles using ordinary differential equations (Gora et al., 2020).

3. Methodological Framework for Traffic Simulation

Based on the literature review and the authors' experience in applying traffic simulation to evaluate mobility strategies, a comprehensive framework is proposed. Figure 1 outlines the process and illustrates its data-intensive nature, highlighting its strong dependence on the amount, quality, and type of data.

Figure 1

Download asset Open asset

Proposed methodology for implementing traffic simulation at any scale, aiming at evaluating strategies to improve mobility in a given region.

The framework consists of sequential steps. The first step, planning, defines the study’s objectives and scope, including the region of interest in terms of location, homogeneous sociodemographic zones, and available physical infrastructure such as the road network, public transport, vehicle fleet, and energy sources for passenger and freight transport. The second step, transportation data, involves the development, acquisition, or post-processing of the origin–destination matrix, supported by the identification of travel needs and major trip attraction centers. The third step, travel demand estimation, specifies the set of trips within the study region, including origin, destination, departure time, route, and mode of transport. The fourth step, simulation, represents the dynamic execution of these trips over time. The fifth step, calibration and validation, assesses the model’s ability to reproduce observed traffic conditions, with calibration involving parameter adjustment to improve accuracy (Dowling et al., 2004). Finally, the model is applied according to the objectives defined in the planning stage.

3.1. Planning

3.1.1. Scope Definition

The planning of a traffic modeling and simulation project is guided by the identification of key elements: the area of interest defining the network to be analyzed; the intended use of the model to evaluate traffic behavior, assess new strategies, or monitor performance; the required accuracy, defined as the degree to which model outputs match observed or accepted values; the key performance indicators used to support decision making; the required precision, understood as the unit of analysis for both inputs and outputs, such as individual vehicles, road segments, or zones; and the level of detail or resolution of the model; and the available resources, including expertise, software, time, computational capacity, budget, and data quality (Krista et al., 2004).

Regarding intended use, and following Krista et al. (2004), traffic simulation models support improved decision making in transportation management, the evaluation and prioritization of planning and operational alternatives, reductions in design time and costs, minimization of traffic disruptions, communication with stakeholders, operation of existing roadway capacity, performance monitoring, and recently, as an intermediate layer of cities’ digital twins (Huertas et al., 2026).

3.1.2. Zoning

Once the area of interest has been defined, the next step is to divide it into zones (e.g., households and businesses) with uniform characteristics. This process, known as zoning (Ortúzar and Willumsen, 2011) produces Transportation Analysis Zones (TAZs). Zoning typically involves analyzing activity centers, sociodemographic characteristics, and the transportation system to define appropriate zone boundaries.

In transport modeling, considerable effort is often devoted to subsequent stages, while zoning is treated largely as a matter of “common sense.” However, arbitrary TAZ boundaries can introduce statistical and geographical errors, becoming a major source of inaccuracy in the modeling process (Martínez et al., 2009). TAZ definitions shape the socioeconomic characteristics of subareas and, consequently, influence trip OD patterns. When TAZs lack homogeneity and compactness, trip generation and OD matrices may be poorly estimated. Moreover, predefined zoning systems may fail to capture the dynamic spatial and temporal evolution of land use.

Figure 2 shows an example of different zoning approaches for District Tec in Monterrey, Mexico. Figure 2a delimits District Tec and defines TAZs as “neighborhoods”, with political and administrative distinctions, while Figure 2b delimits District Tec in red line over the Basic geostatistical areas (AGEBs) defined by the National Institute of Statistics and Geography (INEGI) in Mexico. Zoning by neighborhoods can effectively represent mobility patterns, whereas zoning based on AGEBs, with extensive statistical information available at this aggregation level, facilitates integration with broader urban simulation layers, such as digital twins, as in Huertas et al. (2026).

Figure 2

Download asset Open asset

Zoning examples of the same area: District Tec in Monterrey, Mexico.

(a) TAZ as neighborhoods, (b) TAZ as AGEBs. Note: (a) Original map is a public consultation document from the Partial Urban Development Program of DistritoTec (ImplancMTY, 2021) (b) Original map is a public consultation document, retrieved from https://www.coneval.org.mx/.

Zoning approaches that determine the number and size of TAZs have generally evolved into four categories. The first is constraint-based zoning, in which each TAZ must satisfy modeling requirements such as compactness, population thresholds, or computational efficiency. Early spatial aggregation methods followed this approach by hierarchically grouping areas according to the similarity of selected attributes. The second is model-based zoning, which uses computational and transport-demand models to reduce forecasting inaccuracies and improve model performance. The third is criteria-based zoning, where zones are delineated according to planning, land-use, socioeconomic, and network-related characteristics. More recently, data-driven zoning has emerged, applying clustering techniques, machine learning, and large-scale mobility datasets to generate spatially homogeneous zones based on observed travel behavior and urban dynamics. Table 3 summarizes the main characteristics of some classical, widely recognized, and also recent zoning studies in each category.

Table 3

Zoning methods

Type	Author	Description
Constraint based	(Openshaw, 1977)	Maximization of the statistical precision of the estimation of the OD matrix cells.
	(Baass, 1981)	Trip generation/attraction homogeneity Adjustment of TAZ boundaries to political, administrative, or statistical ones. Minimization of intra-zonal trips
	(O’Neill, 1991)	Trip generation/attraction homogeneity Contiguity and convexity zones Compactness of TAZ shapes Exclusiveness (no doughnuts or islands) of zones Equity in terms of trip generation (small standard deviation across zones). Adjustment of TAZ boundaries to political, administrative, or statistical ones. Respect for physical separators Decision-makers' preferences are considered when defining the number of TAZs.
Model based	(Prosperi et al., 2021)	Traffic Zones Discretization and Origin-Destination Matrix Estimation by means of Fusion.
Model based	(Luo and Zhang, 2021)	TAZ are defined by combining static zoning and dynamic zoning, by adjusting the zones based on real-time traffic, enabling a maximum flow control strategy.
Data-Driven	(Martínez et al., 2009)	TAZs are defined based on actual travel demand patterns, using a density-smoothing algorithm.
Data-Driven	(Yang et al., 2022)	Clustering and GIS-based spatial analysis is used for grouping spatial units with similar mobility characteristics.
Criteria Based	(Crevo, 1991)	Minimize prediction errors associated with the spatial aggregation.
Criteria Based	(Ortúzar and Willumsen, 2011)	The aggregation error caused by the assumption that all activities are concentrated at the centroid of the TAZ should not be too large. Each zone must be compatible with other administrative divisions, particularly with census zones. TAZ should be as homogeneous as possible in terms of land use or composition. TAZ boundaries must be compatible with cordons, screen lines, and those of previous zoning systems. Avoid using main roads as zone boundaries. The zone configuration should facilitate the clear definition of centroid connectors. Zones do not have to be of equal size; if anything, they could be of similar dimensions in travel time units (i.e., smaller zones in congested areas).

3.2. Transportation Data Collection

Once TAZs have been established, the next stage in the traffic simulation modeling process is data collection and preparation. These data support subsequent steps, including transport demand modeling, simulation model development, and model calibration and validation. The process involves gathering available information and, when necessary, generating data through field campaigns, acquiring it from open or commercial databases (Chaparro Sierra and Huertas, 2024) or estimating missing variables such as attraction-zone capacities. The data are then processed for model implementation. One of the main limitations in traffic simulation arises at this stage when data quality is insufficient and resources for additional collection are limited. In such cases, the use of less data-intensive tools or broader modeling scopes was recommended (Alexiadis et al., 2004). Nowadays, machine learning techniques can estimate mobility patterns even from incomplete data. For example, INFOSTOP, a two-step stay-point detection algorithm for mobility trajectory analysis, was applied in the OD matrix construction presented by Chaparro Sierra and Huertas (2024).

Table 4 summarizes the required data, whose availability should be verified, as each component directly influences simulation accuracy and outputs. In general, traffic simulation data can be grouped into three categories: input data for the transport demand model, parameters for simulation model development, and driver-behavior characteristics used during simulation.

Table 4

Required input data

Type	Example	Description	Used in:
Type	Example	Description	TM	SM	DC
Network	Road geometry and capacity	Number of lanes, designated turning lanes, length of lanes, free flow speed.		X
	Traffic controls	Traffic signals’ location, signal timing, etc. Also known as geometric data, it consists of the number of lanes, designated turning lanes, length, and free flow speed. Also, it may include data on traffic control devices, such as the location of traffic signals and their signal timing.		X
	Level of service (LoS)	The measure of the quality of flow through a roadway segment or an intersection.		X
	Location of activity enters	Predefined TAZ (attraction and generation zones).	X
	Transit system	Public transport information (type, lines, stops, users, etc.)	X	X
Vehicles	Vehicle mix	Composition of the vehicle fleet in the study area.	X	X
	Maximum acceleration rate	The performance obtained from vehicle characteristics like weight and power varies according to the type of vehicle. In general, for lower speeds, the acceleration rates are greater.		X
	LoS of vehicles	Capacity (number of passengers), idle capacity,
	Vehicle lengths	It is the average length of the vehicles in the study area, which may vary according to the region and the needs of the population.		X	X
Traffic	Volume studies	The annual average daily traffic: Count location and duration using physical counting. It can be done with Inductive loop, magnetic, pneumatic road tubes, active or passive infrared, microwave radar, ultrasonic, passive acoustic or video image processing.		X
	Directional volume studies	Entry volumes, turning volumes, and turning movements at intersections.		X	X
	Demand	Traffic reports (modal preferences, travel habits, OD matrices, accident reports)	X	X	X
	Travel time and delays studies	Average time that it takes to go from point A to B through studies requiring a test vehicle (floating-car technique, average speed-technique, moving vehicle techniques) or not (license-plate method and the interview method).		X
	Capacity flow data	Maximum volume per segment, maximum entrance rate, maximum (saturation) flow.	X	X
People	Temporal location	GPS position at a given time	X
	Socio-demographics	Age, education, economic activity, income, family composition, etc.	X
Driving characteristics	Aggressivity level	Average speed, free speed, and other safety perception parameters.		X	X
	Modal split	Proportion of mode and type of cars used by drivers.	X	X	X
	Average occupancy	Use of HOVs, trip generating activities.		X	X
	Accident rates	Frequency, location, time of the day of incidents and accidents.		X	X

TM = Transport demand modeling; SMB = Simulation model building; DC = driving characteristics;

When developing a transport demand model, the required information depends on data availability and sources. In the best-case scenario, OD matrices are already available, substantially reducing data collection efforts. The main input components are: (a) modal preferences and travel habits, describing travel behavior between zones; and (b) the location, capacity, and activity centers that support the identification of trip generation and attraction areas within the defined TAZs. These inputs can also support the refinement of TAZ boundaries to better represent observed mobility patterns.

The second classification includes the information required for the simulation model to operate properly. It mainly comprises two components: (a) traffic network and transport systems, which support trip assignment and journey’s trajectories definition within the transport model; and (b) calibration and validation data, typically based on key performance measures such as traffic volume, level of service, travel and delay times, capacity, and queue lengths.

A third category of data, increasingly available through current technologies, consists of massive and passively collected mobility information from GPS-enabled mobile devices (Chaparro Sierra and Huertas, 2024; Eom et al., 2026). These data provide representative samples of individual mobility patterns with high spatial and temporal resolution, focusing on people rather than vehicles. Combined with emerging machine learning techniques for large-scale data processing, they offer a promising alternative to traditional approaches, addressing limitations such as high costs, limited representativeness, and low spatial and temporal resolution.

The fourth data type relates to driving characteristics (DCs). Depending on the simulation software, additional parameters may be required to better represent real-world driving conditions, as variations in driving behavior can significantly affect simulation outputs. For instance, traffic characteristics may differ across TAZs. Indicators such as modal split and accident reports can provide insights into driver aggressiveness, the use of dedicated or high-occupancy vehicle lanes, and trip-generating activities, supporting more accurate parameter tuning.

3.3. Travel Demand Estimation

Travel demand models are widely used to predict travel demand and analyze mode, activity, and destination choices based on current conditions. However, they are not intended to evaluate the dynamic interactions between vehicles and transportation systems. Within travel demand estimation, the literature mainly distinguishes between two research directions: trip-based and agent-based (or activity-based) modeling (Harb et al., 2022; Nguyen et al., 2025).

3.3.1. Trip-based Travel Estimation

Trip-based modeling focuses on trip generation as the primary unit of analysis, representing travel demand through aggregated regional mobility patterns. As the classical approach, it models individual daily travel as independent origin–destination trips derived from statistical and census-based data, typically provided by from government institutions, public agencies, or organizations managing publicly available information. These datasets support the analysis of population distribution and demographic characteristics, making the approach suitable for evaluating long-term policies and broad mobility trends.

3.3.2. Agent (or Activity)-based Travel Estimation

Activity-based (or agent-based) modeling extends the trip-based approach by representing individuals as active agents whose travel behavior is shaped by attributes such as family status, occupation, and activity patterns. In this framework, trip generation becomes a more complex process aimed at capturing daily trip chains and the activities driving individual movements. By tracking daily trajectories within a broader system, it provides greater granularity and more dynamic insight into short-term behavioral changes than aggregated population-based models. Table 5 summarizes both approaches.

Table 5

Transport demand models

	Trip-based approach (TBA)	Activity-based approach (ABA)
Scale	City or large urban region	Urban district or smaller
Type of simulation	Macroscopic	Mesoscopic or microscopic
Assumption about trips	Production and attraction of trips by zone (Ortúzar and Willumsen, 2011)	Trips generated and constrained by the activities of individuals
Spatial assumptions	Each zone only contains the centroid.	Zones contain individual households and activity centers (Codeca, 2022)
Approach	Individual, not related trips (Ortúzar and Willumsen, 2011)	Interrelated tours and trip chains
Required data	Zone aggregated data	Individual data by household and individual attributes
Comments	Data must be complete for each zone; fewer approximations are acceptable	Approximations at the individual level can lead to a good representation by aggregation.

At a more detailed level, activity-based modeling represents trips as OD journeys, often structured as simple chains from home to a primary destination (e.g., work or commerce) and back home. However, real behavior frequently departs from this structure. After reaching a primary destination, individuals may include secondary stops (e.g., errands or drop-offs) or combine multiple intermediate purposes in stochastic ways. This expanded representation enables more realistic modeling of trip chains with variable intermediate stops and path dependencies (Krista et al., 2004).

Examples of activity-based models include the evaluation of multimodal transportation shares for reducing traffic congestion (Uthpala et al., 2023), the development of an open-source agent-based traffic simulation tool using cellular automata (Górka and Małecki, 2024), the analysis of sudden travel behavior changes in mixed traffic flow with automated vehicles (Wang et al., 2025), and large-scale pedestrian movement modeling (Kaziyeva et al., 2023).

This richer trip definition has several implications: it enables simulation models to incorporate coupling between stop scheduling and mode choice, time flexibility, and route reshaping in response to intermediate opportunities. In doing so, it also better reflects observed travel diaries, allowing microsimulation engines to allocate more realistic time budgets across activities. Integrated frameworks such as the BayABM model combine Bayesian networks with activity-based architectures to link choice and chaining behavior at the individual level (Luo et al., 2024). Likewise, project-based activity–agent microsimulation models have recently been proposed to overcome the limitations of simple trip chaining, enabling flexible representation of multiple stop patterns across days (Miller, 2024).

Recent research in large-scale activity-based demand generation (Nguyen et al., 2025) refines the generation of chained activity tours beyond single-trip cycles. Similarly, Moeckel et al. (2024)’s ABIT framework (2024) extends weekly activity patterns to capture habitual sequences beyond simple home–work–home loops. Moreover, Harb et al. (2022) integrate multi-purpose chaining in short-term travel demand modeling to better represent mixed trip behaviors, such as combining business and shopping stops within a single tour. Figure 3 exemplifies representations of single chains (TBA approach) versus more complex activity chains (ABA approach).

Figure 3

Download asset Open asset

Examples of OD matrices representations (a) single trips, one mode, (b) activity chains.

The primary input for these models is population survey data, based on the assumption that most regional mobility originates from the individuals who inhabit the area. Using statistically representative samples, travel habit surveys collect information that can be mathematically translated into regional trip patterns (Tamin and Willumsen, 1989). Consistent with the two modeling approaches, surveys may also be trip- or activity-based. Trip-based surveys focus on trips generated by households and aggregate them into travel volumes, as in Díaz-Ramírez et al., 2023, whereas activity-based surveys emphasize the activities that motivate movements across locations and times (McNally, 2008) as in Eom et al. (2026).

Survey data are then processed to estimate origin–destination (OD) matrices, one of the most critical inputs for simulation models. These matrices may vary in aggregation level and can be static or time-dependent (e.g., by day or time period). As shown in Table 6, OD matrix estimation remains essential, even when not explicitly required by simulation software.

Table 6

Different formulations of OD matrices required in simulation models

Input for simulation software	Description	Transformation to OD matrix
# Trips from centroid to centroid once a day	Software packages may require input trips that go from one location to another.	This is a direct formulation of the OD matrix, which could result from the generation stage of the transportation model, and can be requested as an input for each location.
# Trips from centroid to centroid over periods of time (hourly, 15 minutes)	Simulation software may require assigning trips at different time periods as the simulation resolution increases.	This is a formulation that is extended into a 3^rd dimension of time intervals (Fellendorf and Vortisch, 2010).
# Trips from centroid to centroid of different modes over periods of time, etc.	As more detail is used in the simulation, additional disaggregated information could be included as input for simulation software packages.	This formulation extends the previous OD matrices and may include several factors that differentiate the traffic as intended.
Probability that a user has a specific location, activity chain, and mode of transport	Meso- and microscopic simulation software will require a broad set of highly disaggregated data.	This formulation of the OD is disaggregated and may be obtained from the previously explained multidimensional OD matrix. It may require calculating the conditional probabilities of the individual’s travel characteristics (Codeca, 2022; Eom et al., 2026).

3.3.3. The Four-stage Travel Demand Model

Travel demand can be modeled through a four-stage framework Ortúzar and Willumsen (2011) for trip-based modeling, whose stages are described below.

Generation of trips stage: This stage of the transport model estimates the total volume of trips generated and attracted by each TAZ in the region. Depending on modeling objectives, trips are typically classified by purpose and origin (home-based or non-home-based) (Krista et al., 2004). Two complementary models are developed: one for trip production and another for trip attraction. In the production model, trips are assumed to originate from residential zones, operating at the household level. Trips to work and school destinations are represented at the zone level within the attraction model. A closed-system assumption is commonly adopted when external mobility is considered negligible. Additionally, many mobility studies incorporate an urban planning component, requiring long-term analysis to anticipate changes in regional travel demand. Since the activity-based approach relies on the disaggregated analysis of generation, TAZs are divided into households and workplaces. Its implementation requires explicit spatial considerations for assigning trip origins and destinations.

Distribution: At this stage, the mobility relationships between zones are defined through spatial linkage. All trips generated in the previous stage are assigned to an origin and a destination zone, which may be the same zone, an adjacent zone, or a more distant zone (Krista et al., 2004). In general, distribution models assume that travel time is perceived as undesirable; consequently, trips tend to concentrate in nearby areas and progressively decrease with increasing distance. In the ABA, distributing travel demand requires accounting for disaggregate user characteristics, such as access to work or school opportunities that satisfy individual needs, including those within the same household. This stage also incorporates the spatial structure of the network and its alignment with user requirements.

Modal Choice: This stage models the transport mode choice process used to execute the distributed trips from the previous stage. It is based on the concept of attractiveness or utility of alternatives within a user’s choice set. Accordingly, modes with attributes better suited to specific user conditions have a higher probability of being selected to complete trips. These attributes may be evaluated at the individual or group level, based on shared location or characteristics, to estimate each mode’s utility. Examples include travel time, travel cost, waiting time, and accessibility (e.g., proximity to stops or vehicle ownership). In the activity-based approach (ABA), modal choice is implemented through models that estimate the probability of selecting each mode using individual characteristics, such as access to transport options, household and workplace locations, and network spatial structure. This results in a disaggregate model that predicts mode choice at the individual level. Repeating this process across all simulated individuals enables the derivation of aggregate mode shares.

Trip assignment: The final step in travel demand modeling determines the trajectory users select to reach their destinations using the chosen mode. This requires explicit consideration of the network available for each mode. For urban road systems, congestion effects, roadway type, and connectivity to TAZs must be accounted for. Trip assignment within the ABA approach involves evaluating all feasible paths across the network, incorporating traffic conditions and travel times along the selected itinerary. An additional step is required to complete demand modeling: iterative feedback to the distribution and mode choice stages to explore alternative solution states and ensure consistency with the initial OD matrix. This process is referred to as equilibrium between the transportation system and its users.

Each stage may be modeled using different methods in a hybrid framework, adapting travel demand modeling to available data and expertise. Codeca et al., 2015 developed a hybrid activity-based simulation model of Luxembourg using SUMO, combining count-based trip generation with agent-based distribution and mode choice, and applying maximum likelihood estimation for route assignment that minimizes individual travel times.

Table 7 explains the methods reviewed in the literature related to travel demand modeling. Five categories are included: simple methods that update old OD matrices with new available information, theoretical methods that are based on laws of physics and phenomena from different disciplines, and are applied by analogy; and counting-based methods that are based on observations of flows and are based on assignment equations (Bierlaire, 1997); estimation methods that calculate unknown parameters of equations and functions that model demand (Ben-Akiva and Bierlaire, 1999); and machine learning methods that are algorithms used for multiple types of tasks that can model travel demand (Salas et al., 2022), which can be identified as the current trend.

Table 7

Methods for travel demand modelling

Category	Method	Approach		Model Stage				Level			Details	Sources
Category	Method	TBA	ABA	G	D	M	A	Mi	Me	Ma	Details	Sources
Simple methods	Growth factor	X		X						X	Updating an old matrix using the actual number of trips.	(Furnes, 1965)
Simple methods	Tri- proportional	X		X						X	Growth factor considering cost distribution.	(Bierlaire, 1997)
Theoretical methods	Gravitational	X			X			X		X	Analogy with Newton’s gravitational law.	(Casey, 1955)
	Entropy maximization	X				X	X	X	X		Derived from the second law of thermodynamics.	(Ortúzar and Willumsen, 2011)
	Agent-based		X	X	X	X	X	X	X	X	Derived from cellular automata theory, using individual beha-vioral rules to define aggregate transportation.	(Zhang and Levinson, 2004)
Counting-based methods	Proportional assignment	X		X	X			X	X	X	Consider that the flow proportion using a specific link is inde-pendent of the complete traffic flow.	(Dial, 1971)
Counting-based methods	Restricted capacity assignment	X			X			X	X		Based on Wardrop equilibrium (1952), taking congestion into account. The flow proportion depends on the complete traffic flow.	(Beckman et al., 1956)
Estimation methods	Linear least squares	X		X	X			X	X	X	Fit the parameters of a model that minimizes the difference between estimated and real flows	(Bierlaire, 1997)
Estimation methods	Maximum Likelihood estimation		X		X	X		X	X	X	With a probability model and statistical hypotheses to make statistical inference.	(Ben-Akiva and Bierlaire, 1999)
Machine Learning methods	Classification		X			X		X	X	X	Algorithms based on decision trees that optimize the modeling of mode choice based on decision rules.	(Hillel et al., 2021)
	Clustering		X	X	X			X	X	X	Unsupervised algorithms that group zones in the study region. Based on individual characteristics.	(Hafezi et al., 2019)
	Regression	X	X	X	X		X	X	X	X	Supervised algorithms that forecast the demand for travel, the distribution of activities of trips, and can forecast.	(Rocha et al., 2021)

TBA: Travel-based approach, ABA: Activity-based approach, G: Generation stage, D: distribution stage, M: Mode choice stage, A: Trip assignment stage, I: Microscopic simulation level, E: Mesoscopic simulation level, and A: Macroscopic simulation level.

Table 7 illustrates how a single method can be applied across multiple modeling stages. This table also suggests the high potential for hybridization of the travel demand process by combining methods at each stage or using the same technique at different stages.

3.4. Building the Simulation Model

After defining the objectives and modeling approach, appropriate traffic simulation software must be selected. The platform should match the modeling approach, analysis scale, and project objectives, while model development relies on input data generated in earlier stages.

Open-source platforms include SUMO (Lopez et al., 2018), MATSim (Balmer et al., 2008), TRANSIMS, MOTUS, TRITONE, and OpenTrafficSim (Raju and Farah, 2021), Commercial alternatives include the PTV VISSIM and PTV VISUM platforms (Fellendorf and Vortisch, 2010), Aimsun (Casas et al., 2010), and CORSIM (Halati et al., 1997; Raju and Farah, 2021; Saidallah et al., 2016). Several studies have reviewed the evolution and capabilities of these simulation tools, while comprehensive catalogs and comparison resources are also available online (University of Leeds and ITS, 2000).

Raju and Farah (2021) reviewed the development of sixteen traffic microsimulation platforms throughout this century, identifying SUMO, VISSIM, and AIMSUN as the most widely adopted, followed by PARAMICS and CORSIM. SUMO and VISSIM have been extensively applied to evaluate the externalities associated with urban mobility policies and are increasingly used in emerging research on autonomous vehicles and connected mobility. More recent comparative studies have evaluated traffic simulation platforms according to modeling scale, licensing schemes, platform compatibility, visualization capabilities, application objectives, and support for intelligent transportation systems and connected or autonomous vehicles (Algherbal and Ratrout, 2025).

3.4.1. Traffic Simulation Software

Although several reviews of simulation software are available, this study provides an independent comparison of existing tools to assess their suitability for traffic simulation applications. Table 8 summarizes the main characteristics of these platforms.

Table 8

Most common simulation software tools

Software	Cost & License	User-Friendliness	Modeling Approach	Visualization	Strengths (Power)	Limitations
Aimsun	Commercial, expensive; academic licenses available	User-friendly GUI, steep learning curve	Microscopic, mesoscopic, hybrid	Strong 2D/3D visualization	Flexible (multi-res), good for large urban projects	Costly, proprietary
TransModeler	Commercial (Caliper Corp.)	GUI is okay but less intuitive than Aimsun/VISSIM	Micro + meso + dynamic traffic assignment	3D visualization (moderate quality)	Strong for planning + operations integration	Less global user base, fewer academic users
SUMO	Free, open-source	Moderate (steep learning initially), script-based	Micro + meso	2D GUI, limited native 3D (extendable via Unity/ Blender)	Scalable (city-wide), great for AI, CAV, RL	Graphics not polished; scripting needed
PTV VISSIM	Commercial, very expensive	Very user-friendly GUI	Microscopic only	Excellent 2D/3D visuals	Most realistic driver behavior (Wiedemann), industry standard	High cost, slower on very large networks
PTV VISUM	Commercial, costly	GUI-focused, planner-friendly	Macroscopic (strategic planning)	Visualization is more schematic than realistic	Excellent for demand modeling, network assignment	Not for micro-level traffic operations
MATSim	Free, open-source	Code-based, not very beginner-friendly	Agent-based, mesoscopic	Limited built-in; external tools for visuals	Huge-scale simulations (whole regions)	Requires coding skills, less visual
Cube	Commercial (Bentley)	Planner-oriented GUI	Macroscopic (demand forecasting)	Limited visualization	Widely used for travel demand modeling	Not for microscopic operations
Dynasim	Commercial (PTV legacy)	Less common now	Microscopic	Decent visualization	Earlier detailed micro tool	Mostly replaced by VISSIM
Mezzo	Free (academic, KTH Sweden)	Basic, not very user-friendly	Mesoscopic (stochastic simulation)	Very limited	Efficient meso-sim, academic research	Not widely used, poor visuals
Polaris	Free (open source, Argonne National Lab)	Requires coding	Agent-based, mesoscopic	Limited	Strong research tool for large networks	Steep learning, low adoption

Based on Table 8, widely used traffic simulation software exhibits distinct advantages depending on the application context. For projects with sufficient budgets and product-oriented requirements, such as government or privately funded transportation infrastructure initiatives, commercial platforms including PTV VISUM and PTV VISSIM provide robust and comprehensive solutions. In contrast, SUMO is particularly well suited for research due to its open-source nature, active community, and collaborative ecosystem. From the authors’ experience, SUMO provides high-quality outputs, supports multiple data formats for interoperability with external platforms, and offers Python-compatible modules that facilitate advanced research without licensing constraints, but requires extensive programming effort for data integration and customization. It can also be integrated with external applications such as Unity to improve visual representation.

Figure 4 summarizes the key features identified in the analysis. Each axis represents a key evaluation dimension: Integration, the ability to connect with external data or tools; Visualization, the quality of graphical outputs; Scalability, the capacity to handle different network sizes; Cost, referring to licensing and expenses; Co-simulation, the ability to work with other simulators; Updating, the frequency of updates and level of support; Multimodal, the ability to model multiple transport modes; and Open-source, the availability of source code.

Figure 4

Download asset Open asset

Comparison of available traffic simulation software capabilities

Figure 4 compares some of the authors’ highest-ranked and most widely adopted traffic simulation platforms, highlighting open-source tools such as MATSim and SUMO for their flexibility and continuous development, as well as commercial solutions such as VISSIM and VISUM, both developed by PTV Group, for their advanced visualization capabilities and extensive professional application (Algherbal and Ratrout, 2025).

The Appendix A summarizes recent applications of SUMO, MATSim, and VISUM reported in the literature.

3.5. Calibration and Validation

In traffic simulation, calibration consists of adjusting model parameters to improve the model’s ability to reproduce local traffic conditions (Holm et al., 2007), whereas validation evaluates the calibrated model’s capacity to accurately represent travel behavior using metrics independent from the calibration process, thereby supporting generalization and predictive capability.

Table 9 summarizes the main parameters associated with each traffic model, although these are not exclusive to a particular model or software. Additional parameters include vehicle attributes and legal or desired driving limits. Olstam and Tapani (2004) provide an extensive description of the parameters commonly used in the first two traffic model categories presented in Table 9.

Table 9

Metrics used for calibrating traffic models.

Traffic model	Software	Parameter	Description
Gap acceptance	Based on Gipps (AIMSUM)	Look ahead distance or Safety gap or safety distance or Headway or min distance between vehicles	Necessary distance to avoid a collision if the leader decelerates heavily
Psycho-physiological driver behavior	Based on GHR models (MITSIM) SUMO VISSIM Based on Fritzsche (PARAMICS) Based on Wiedeman model	(Driver’s) Reaction time or Time headway	Can be the same for all drivers (macro) or specific per type of vehicle, per level of congestion (micro), for example. It influences density and flow.
		(Driver’s) magnitude of the reaction (acceleration, deceleration, or retardation rates)	Can influence travel time delay and average speed
		Car-following parameters (α, β, γ)	Proportionality rations between acceleration and speed, speed difference between follower and leader, and space headway.
		Max waiting time, Standstill distance, Speed model, critical gap, follow up gap.	The gaps can influence the maximum capacity of a subordinate flow within a node. The max waiting time is the longest time a vehicle of subordinate flow can wait to enter a node.
		Action points	Thresholds where the driver changes his/her behavior.
		Threshold for perception of negative and positive speed differences	Below these thresholds, the follower doesn’t perceive the speed differences.
		Desired time gap, risky time gap Safety time gap	Required to compute the thresholds: Desired distance, risky distance, safe distance, and braking distance.
		Several fix and random numbers for different thresholds.	Required to compute the thresholds: Desired distance between stationary vehicles, desired min following distance at low-speed differences, max following distance, approaching point, and decreasing, and increasing speed differences.
Cell based model	MATSIM	Length of the road, length of the cell, probability of decrease speed, max speed, probability of switch lane, initial density, number of cars, acceleration rate.	Parameters required to simulate one-lane highway traffic model with the Cellular automaton model.
Trajectory based model		Time lag, Distance lag Desired speed, Desired time gap, Minimum gap, Max acceleration, Comfortable deceleration	Required by the IDM model.

(Olstam and Tapani, 2004; Sun et al., 2020)

3.5.1. Goodness-of-fit Measures

Although earlier studies often limited validation to visual inspection of metric behavior without quantitative assessment of differences (e.g., Ding, 2011; Olstam and Tapani, 2004), current computational capabilities enable the systematic evaluation of multiple error functions.

During validation, model parameters may undergo optimization procedures (Treiber and Kesting, 2013), based on design of experiments, trial-and-error calibration, or sensitivity analysis (Azam et al., 2019; Kanagaraj et al., 2013). These approaches seek goodness-of-fit measures that minimize error functions, satisfy threshold criteria, or support statistical hypothesis testing. Error is defined as the difference between simulated outputs and observed data. Commonly used measures include the GEH statistic for traffic volumes; absolute and relative errors (Giraldo et al., 2025; Hale et al., 2015; Krajzewicz et al., 2002; Ranjitkar et al., 2005), cosine similarity (Giraldo et al., 2025), mean absolute percentage error (MAPE), root mean square error (RMSE), (Patwary et al., 2021; Treiber and Kesting, 2013), and coefficient of determination (R²). Table 10 describes the most common and recommended error functions in traffic simulation, where $y^{o}$ is the observed value and $y^{s}$ the simulated value.

Table 10

Recommended error functions in traffic simulation

Function	Description	Formula	Recommendation / Assumption/Strength
R² Coefficient of determination	Indicates how well the simulated variable explains the real one.	$R^{2} = 1 - \frac{\sum {(y^{o} - y^{s})}^{2}}{\sum {(y^{o} - \bar{y})}^{2}}$	Assumes linear correlation. Easy to understand, and common [0,1]
GEH statistic (Geoffrey E. Havers)	Normalizes the relative differences with respect to the magnitude of the compared flows.	$G E H = \sqrt{\frac{2 \sum {(y^{o} - y^{s})}^{2}}{y^{o} + y^{s}}}$	Mitigates the misleading effects that relative differences may produce. <5: good fit, > 10: poor fit
MAPE Mean absolute percentage error	Indicates the average percentage difference between observed and predicted values.	$M A P E = \frac{100}{n} \sum^{n} \| \frac{y^{o} - y^{s}}{y^{o}} \|$	Easy to interpret as a percentage [0 – 100] <10%: good fit
RMSE Root Mean Square Error	Statistical measure that estimates the standard deviation of the errors distribution.	$R M S E = \sqrt{\frac{\sum {(y^{o} - y^{s})}^{2}}{n}}$	In the same units as the variable. Not a standard range. Penalizes large errors.
Relative differences with Cosine similarity	Similarity between two vectors or tensors by means of the dot product between them.	$R D = 1 - \frac{\| y^{o} \cdot y^{s} \|}{‖ y^{o} ‖ ‖ y^{s} ‖}$	For multi-dimensional variables. Easy to read. Range [0,1] >0.9: good fit

The literature highlights several key considerations for calibration processes: (a) prioritizing global rather than purely local calibration based on gap measures; (b) selecting intuitive and plausible parameter values, preferably initialized from published data; (c) using data from representative locations rather than inflection or boundary points; and (d) beginning with zero-conflict conditions or scenarios without flow on the major road (Sun et al., 2020; Treiber and Kesting, 2013).

3.5.2. Model Performance Metrics

The suitability of a simulation model depends on its intended purpose, which determines the performance metrics to be monitored, calibrated, and estimated throughout the simulation process. Without aiming to be exhaustive, Table 11 summarizes the most common output metrics, together with their level of detail and application scale. Among these, travel time, travel speed, density, and saturation stand out due to their broad applicability across different purposes and scales, as well as their availability in most simulation software. The units are given in SI units. Some of the metrics described in Table 11 may also serve as input data for the calibration process.

Table 11

Model performance metrics

Metric	Units	Description	Detail			Scale		Purpose
Metric	Units	Description	Mi	Me	Ma	N	I	I	O
Travel time	min/veh sec/veh	Average travel time of all vehicles of the simulation or in a specific facility or trajectory.	x	x	x	x		x	x
Travel Speed	km/h	Rate of motion (expressed in distance per unit of time)	x	x	x	x		x	x
Travel Distance	km/veh	Average extent of the space between the trip origin and the destination, measured along a vehicular route.	x	x	x	x		x	x
Delay	Sec or min	Additional travel time experienced by travelers at speeds less than the free flow (posted) speed.	x	x		x			x
Speed ratio	-	Degree of traffic flow through intersections (speed adjacent to the access point over speed along the corresponding road section).	x	x	x	x			x
Stopping frequency	% Stops per time-space bin	Average relative frequency of stops per time, space, or segment bin. Shown in histograms.	x	x		x			x
Volume of Traffic or Traffic Flow	% Vehicles or persons per time-space bin	Number of persons or vehicles passing a point (or intersection) on a roadway per time interval. Shown in histograms.			x	x		x	x
Congestion of streets	sec or min	Time lost per trips, level of congestion per road.			x	x		x	x
Saturation (flow) rate	% Vehicles per time bin	Maximum rate of flow of traffic in a road segment per time interval.	x	x	x	x		x	x
Density	veh/km or veh/km-lane or % Capacity of lane/road	Number of vehicles on the roadway segment averaged over space.	x	x	x	x		x	x
Queue Length	km or # Vehicles	Length of queued vehicles waiting to be served by the system.	x	x		x			x
Mode Split	% Persons	Percentage of travelers using each travel mode (SOV, HOV, transit, bicycle, pedestrian, etc.).				x		x	x
Time-Space Diagrams	km/h per (time-space bin)	3D graphs or surface response plots showing the speed at various times and spaces (positions).				x			x
Time-Flow Diagrams	% of maximum flow	Graphs show the flow behavior at various times with different densities.				x			x
Cost of travel (due to delays)	Time	The sum of the volume delay, turn penalty and junction delay functions.	x		x	x			x
Cost of travel	$ or time	Dynamic cost functions depend on the experienced travel time.	x	x		x	x		x

Mi: Microscopic or specific, Me: Mesoscopic, Ma: Macroscopic or network-wide, I: Input, O: Output, N: Network-wide, I: vehicle or user specific.

As an example of the application of these metrics in a validation exercise, Figure 5 presents the normalized traffic volume profile obtained from vehicle counts and from the corresponding traffic simulation model. Figure 5a illustrates the temporal profiles, while Figure 5b presents the correlation diagram between both variables, together with the metrics summarized in Table 10.

Figure 5

Download asset Open asset

Validation example based on traffic volume profiles using five goodness-of-fit measures.

3.6. Analysis of Scenarios

Simulation scenarios represent variations of the current state of the modeled system, typically defined from a validated baseline scenario, and are used to assess implementation feasibility, roadway resilience, or the effects of population growth. Simulation outputs may be analyzed at the network level, aggregated by roadway or zone, or at the level of individual vehicles. These outputs quantify changes in system performance resulting from the evaluated scenarios and should remain comparable to the validation data collected during the initial stages of the methodology. Examples of simulation outputs are summarized in Table 12.

Table 12

Examples of expected results from different scenarios

Simulated scenario element modified	Examples	Expected results	Examples of Sources
Network	Add/remove stoplights, lanes, entrances, intersection designs, etc. Change of lane uses (bikes per vehicles) Different modal orientations of integrated urban transportation systems	Changes in congestion, in travel times, in transferring time, integration effectiveness.	(Fierek and Zak, 2012)
Vehicle	Change buses for electric buses Penetration of connected and automated vehicles	Changes in emissions Changes in congestion and network capacity Changes in accident rate and road safety	(Raju and Farah, 2021)
Traffic flow management	Lane sorting strategies Converting HOVs into CACC (Cooperative Adaptive Cruise Control) lanes Cooperative ITS (Intelligent Transportation Systems)	Changes in traffic streams, operational traffic performance	(Codeca and Härri, 2018; Raju and Farah, 2021)
Policy implementation	Floating cars versus local taxis fleets	Differences in passenger wait time, pickup trip time	(Maciejewski et al., 2016)

4. Implementation Recommendations

Experience with modeling and simulation using SUMO has highlighted several factors that can substantially affect performance and execution. Efforts to improve realism and reproduce the intrinsic characteristics of mobility patterns generally require modifications in three main areas: network configuration, additional coding for trip generation, and increased framework complexity.

4.1. Network improvements

The network model should accurately represent the mobility patterns of the study area and capture key characteristics such as speed, congestion levels, and primary routing structures. Although simulations can still be executed without an ideal model, the resulting outputs may become difficult to validate against real-world data. During initial runs, execution failures may occur, and the presence of errors or warnings can substantially increase computation time, often resulting in incomplete or unsuccessful simulations.

Pedestrian Path Flow: When importing maps from GPS-based sources such as OpenStreetMap, pedestrian routes are often assigned to edges as unidirectional elements. On one-way roads, this causes both sidewalks to inherit the same pedestrian direction, potentially generating errors, warnings, and pedestrian congestion during simulation. To avoid these issues, all edges should include bidirectional pedestrian paths and maintain proper connectivity with adjacent edges to ensure continuous routes. Figure 6a illustrates this adjustment. Without these modifications, simulations may experience significant pedestrian jams and execution errors.

Figure 6

Download asset Open asset

Examples of improvement implementations in the network.

(a) Pedestrian path flows, (b) Controlled intersections.

Parking Stations Definition: Parking stations should be defined as having realistic capacities and accurate locations across both commercial and residential areas. Although simplifying them as indefinite points may be acceptable for targeted simulations, it can introduce bias and inconsistencies in broader analyses. While a detailed definition increases modeling complexity and computation time, it provides higher-resolution and more reliable results.

Traffic Light Programs: Traffic signal timing and prioritization strongly influence simulation outcomes. Modifications to traffic light programs can increase traffic density, generate congestion, and extend computation time. Accurate traffic signal data obtained from official sources or field measurements are therefore recommended to ensure reliable simulation performance. Figure 6b presents an example of a detailed signal-controlled roundabout intersection.

Intersection and Lane Maneuvers: Lanes at intersections should be reviewed and adjusted to reflect local traffic conditions and right-of-way priorities. Imported network maps frequently lack detailed representations, often simplifying intersections into basic configurations. In addition, parking station locations may impose restrictions on adjacent lanes, which should be considered to ensure realistic simulation behavior.

4.2. Activity-based trips implementation

Activity-based trip generation is widely used for modeling traffic demand by generating trips for individual mobility actors according to activity distributions and modal shares. Although conceptually straightforward, the complexity of activity chains depends on the desired level of resolution.

Complex activity chains remain less frequently addressed in the literature. In SUMO, the SAGA model presents limitations in scripting and trip-file generation, complicating the implementation of chains that combine multiple activities, stops, or transport modes. These chains may require returning to previous locations or ending at specific destinations. Nevertheless, they can represent realistic daily routines; for example, an adult may drop off children at school, commute to work, leave during the day to pick up the children, and then return to work. The required level of detail depends on the intended modeling precision. Figure 7a presents examples of these activity chains.

Figure 7

Download asset Open asset

Schemes of activity-based trips.

(a) activity chains considering work, study and entertainment. (b) multimodal activity chains. (c) combined multimodal and multi-activity chains.

Incorporating multiple transport modes within an activity chain further increases variability across trip segments, as illustrated in Figure 7b. Although individual short trips may rely on different modes of transport, the overall modal distribution should remain consistent throughout the simulation.

Implementation challenges remain significant when defining parameters for activity chains that combine multiple activities and transport modes, as illustrated in Figure 7c. Even with advanced AI-assisted programming and coding tools, configuring these interactions continues to be a time-consuming and complex task.

4.3 Simulation Size

The spatial extent of a simulation influences the choice of modeling approach and software. While larger, detailed areas provide a better understanding of complex traffic phenomena, computational constraints and available resources limit the feasible size and scope.

Micro-Scale Simulation: Micro-scale simulations generally focus on limited areas, such as individual streets or neighborhoods, involving relatively few intersections, trips, and agents. As network complexity increases, accurate information on traffic signal programs, lane configurations, and intersection maneuvers becomes increasingly important. In SUMO, the NetEdit module supports the selection, modification, and definition of these elements.

Large-Area Simulation Challenges: Generating detailed trips over a large area can be computationally intensive or infeasible. The SAGA framework supports pre-generation of trip plans, allowing the simulation to focus on trip execution.

Parallel Processing: Recent SUMO developments incorporate multiprocessing capabilities for operations such as parking station rerouting, substantially reducing computation time in networks with extensive parking infrastructure.

Boundary Effects: Restricting the analysis to a sub-region may introduce inconsistencies associated with incomplete trips crossing the simulation boundary. Trips can be classified as: (1) entirely within the study area, (2) originating within the area and leaving it, (3) entering the area and remaining within it, or (4) traversing the area. Proper classification is essential to preserve realism in micro-scale simulations.

Overall, these techniques and recommendations support a more streamlined simulation workflow and facilitate the extraction of meaningful insights from digital mobility models. Although specific software platforms are commonly associated with particular simulation scales, many tools can be adapted across scales. Advances in computational capabilities and modeling methodologies increasingly enable the integration of multiple approaches to overcome individual limitations.

5. Conclusions

Traffic models are essential tools for representing and understanding vehicular movement in urban networks. They support the evaluation of mobility strategies, infrastructure optimization, and the assessment of environmental impacts such as energy consumption and emissions. This review consolidated the methodological and computational foundations of traffic microsimulation, synthesizing the key stages required for realistic mobility representation, from data collection to calibration and validation.

The proposed six-step framework provides a structured methodology adaptable to microscopic, mesoscopic, and macroscopic scales according to data availability, computational resources, and decision-making objectives. It integrates multimodal and activity-based approaches to improve the representation of real-world mobility dynamics. The comparative assessment of simulation platforms highlights complementary strengths among existing tools. Open-source platforms such as SUMO and MATSim provide flexibility and scalability, whereas commercial tools including PTV VISSIM, PTV VISUM, and Aimsun offer advanced visualization and calibration capabilities. Recent integrations with platforms such as Unity, CARLA, and Python APIs continue expanding the analytical and visualization capabilities of urban mobility models.

Overall, microsimulation has evolved beyond the replication of traffic flows into a robust framework for sustainable transportation planning and policy evaluation. Advances in digital twins, real-time data assimilation, and AI-driven optimization are further strengthening its role in the development of resilient and intelligent transportation systems.

Future research should address several emerging challenges, including advanced data-driven methods for configuring complex activity chains, the integration of individualized data from personal sensing devices to improve dynamic travel demand estimation, and the use of parallel and next-generation computing approaches to support larger and more detailed simulations. Incorporating daily variations in traveler behavior and demand will also be essential for improving the predictive and prescriptive capabilities of future microsimulation frameworks.

Appendix A

Applications per Software

SUMO applications

Recent applications of SUMO in traffic research highlight its versatility. One study demonstrated how Unmanned Aerial Vehicles (UAV) data, when combined with SUMO, can capture the dynamics of traffic shockwaves at intersections. It revealed that rear-end collision risk propagates with these waves and concentrates at the rear of queues (Wang et al., 2025). Another investigation enhanced the realism of cyclist behavior by incorporating empirical insights from the SimRa dataset to design distinct cyclist profiles and improve intersection logic, thus overcoming the limitations of the default bicycle model (Karakaya et al., 2023). Large-scale applications are also evident. A co-simulation environment linking CARLA, an open-source simulator for autonomous driving research, with SUMO has been developed to evaluate the performance of automated vehicles under calibrated traffic demand, enabling robust system-level testing of safety and efficiency (Yilmaz-Niewerth et al., 2024). At the urban scale, commuting patterns in Aichi Prefecture were simulated using calibrated trip and vehicle data, successfully reproducing accident-related patterns and informing prevention strategies (Suzuki et al., 2024).

The combination of SUMO with external visualization engines has opened new ground in transport modeling. A framework integrating Unity 3D and Blender with OpenStreetMap data enables real-time visualization and multi-scale emission measurement (Nagy et al., 2025). On the methodological side, comparative analyses of demand generation tools embedded in SUMO have shown that tool selection directly influences network connectivity, emissions, and re-routing behavior, factors critical for reliable vehicular network simulations (Barbecho Bautista et al., 2022). Policy-oriented research has also expanded, demonstrating that congestion in Zurich can only be avoided if more than 50 % of car users shift to bicycles (Fulton et al., 2025). Similarly, analyses of urban form have revealed that rectangular block designs consistently outperform triangular and radial configurations in both efficiency and sustainability (Essamlali et al., 2025).

The integration of SUMO with external visualization engines and methodological tools has broadened its application in traffic research. Frameworks such as SUMITY enable real-time visualization and multi-scale emission assessment, while the choice of demand generation tools has been shown to affect network connectivity, emissions, and rerouting behavior. SUMO is increasingly applied in policy and planning studies, including evaluations of bicycle infrastructure and urban block design, demonstrating its value for assessing efficiency and sustainability across diverse urban contexts.

Energy and environmental dimensions have also been modeled in detail. The incorporation of fine-grained energy variables into SUMO-based bus simulations has demonstrated how stop spacing, vehicle type, and passenger load influence energy consumption, highlighting the framework’s potential for digital-twin development García-Cerrud et al. (2024). Integration of live video data with SUMO experiments has enabled quantification of heterogeneous transport behavior with over 95 % classification accuracy, demonstrating the benefits of linking field data and simulation for enhanced safety management (Ravindran et al., 2025). SUMO has also been connected with analytical hierarchy process methods to evaluate public acceptance of transport technologies, showing that NGVs lead in acceptance due to cost and infrastructure reliability. At the same time, EVs perform best environmentally but remain limited by affordability and network constraints (España et al., 2025).

Collectively, these studies illustrate the versatility of SUMO when combined with empirical data, co-simulation platforms, or decision-analysis tools. The open-source framework not only enhances methodological rigor but also supports practical insights for safer, greener, and more socially acceptable mobility systems.

VISUM applications

Beyond open-source environments, proprietary tools like VISUM have also been extensively used to explore sustainable transport strategies, multimodal planning, and policy interventions. VISUM-based studies frequently integrate GIS data, participatory methods, and external models to support comprehensive decision-making. The mapping of transport externalities, such as emissions, noise, and congestion, has enabled the identification of critical blackspots for mitigation strategies (Sampaio et al., 2022). Bicycle traffic modeling in Warsaw has been incorporated into a broader regional transport framework to support cycling policy evaluation (Jacyna et al., 2017). Pedestrian-oriented analyses combining demand modeling, microscopic simulation, and participatory approaches have demonstrated the usefulness of multi-criteria decision processes for evaluating pedestrianization alternatives (Gülhan et al., 2025). Evaluations of Madrid’s “Centro” low-emission zone have shown notable improvements in pedestrian air quality following car-access restrictions (Sánchez et al., 2021).

Further research has explored the link between emissions and urban form. Bottom-up emission modeling integrating the Irish National Transport Model with VISUM found that electric vehicle adoption yields greater CO₂ savings exits (Charly and Caulfield, 2025). Simulation of alternative morphologies in Aveiro, Portugal, indicated that compact urban layouts reduce emissions, whereas dispersed forms exacerbate traffic pollution (Augusto et al., 2024). Analyses of depopulation in Łódź, Poland, revealed that urban shrinkage promotes decentralization, longer commutes, and a weaker service core, with significant implications for planning (Sahebgharani et al., 2024).

Data-driven and digital integration approaches are increasingly prominent. The use of smartphone records for demand estimation has demonstrated the potential of mobile data to complement traditional surveys in scenario testing (Montero et al., 2022). Linking OD-matrix calibration and Python routines with VISUM has advanced digital-twin development and mobility replication for long-term decision-making (Pala et al., 2025). Comparative evaluations of static and dynamic tolling have shown that dynamic link-based schemes enhance welfare under autonomous and shared-autonomous vehicles, while static tolls perform better in mixed fleets (Shatanawi et al., 2022). Finally, multimodal scenario assessments along the Lao Cai–Hanoi–Hai Phong–Quang Ninh corridor have indicated that integrated strategies improve connectivity while reducing emissions (Nguyen et al., 2025).

Together, VISUM-based studies indicate that simulation is increasingly applied not only for technical modeling but also for evaluating policy instruments, sustainable designs, and social trade-offs. The findings show that VISUM, often integrated with complementary data sources and analytical methods, functions as a versatile decision-support tool across contexts ranging from emissions reduction and cycling promotion to pricing, pedestrianization, and corridor-level investment planning.

VISSIM applications

Compared with open-source platforms, VISSIM is typically employed when greater microsimulation detail or explicit driver–vehicle interaction modeling is required. Applications span logistics, urban networks, safety, and behavioral analysis. City logistics modeling has demonstrated that intersection redesign can reduce congestion and promote sustainable freight operations (Kučera and Chocholáč, 2021). During the COVID-19 pandemic, simulations of Calgary’s Marlborough transit station tested distancing rules and spatial redesigns, revealing that simple, low-cost interventions could balance infection risk and operational efficiency (Miao and Saidi, 2025). In the heterogeneous traffic conditions of Mangalore City, calibrated microsimulation identified short- and long-term measures that improved flow and reduced travel times (Bandi and George, 2020). They also evaluated the impact of vehicle and driver characteristics on the traffic volume prediction in the same city (Bandi and George, 2020). At the freeway level, the evaluation of ramp metering strategies in Thessaloniki confirmed that both fixed and actuated control can improve performance and reduce environmental impacts (Mitkas and Politis, 2020).

Recent developments include the “Virtual Emergency Lane” (VEL) strategy, which reallocates lanes dynamically during incidents to enable rescue vehicle access, yielding reductions in delay and improved emergency response efficiency (Wang et al., 2025). Broader methodological reviews have summarized advances in driving-behavior simulation using artificial intelligence, covering calibration techniques, car-following models, heterogeneous traffic, and AI-based optimization of driver responses (Al-Msari et al., 2024). Pedestrian–vehicle conflict modeling has also benefited from VISSIM’s integration with the SSAM safety assessment tool, achieving over 90% prediction accuracy and demonstrating that mitigation measures can reduce conflicts by up to 51% (Hussain et al., 2024). Collectively, these studies demonstrate that VISSIM has evolved into a flexible platform for analyzing city logistics, public transit resilience, roadway performance, incident response, and safety evaluation—particularly when combined with AI or surrogate safety models.

Overall, the growing body of research using VISSIM underscores its strength in detailed, behavior-based simulation, where human decision-making, congestion dynamics, and safety risks are central. Unlike open-source frameworks, VISSIM allows fine calibration of driver, vehicle, and infrastructure interactions, making it a preferred tool when precision in microscopic modeling is required, especially in studies linking operational performance to safety and emergency management.

MATSim applications

In contrast, MATSim focuses on multi-agent, activity-based simulation, offering a powerful environment for evaluating policies, demand-responsive transport, and the integration of emerging mobility services. Recent work has demonstrated MATSim’s evolution through co-simulation, extensions, and large-scale urban applications. Integrated frameworks combining MATSim with external fleet simulators have enabled realistic modeling of ride-pooling and heterogeneous services without additional extensions (Yang et al., 2024). A validated methodology for cycling in England incorporated a “quietness” attribute to capture environmental effects on route choice, providing results transferable to other regions (Alvarez Castro et al., 2024). Enhancements to the MATSim Open Berlin scenario improved activity-based demand generation, automated calibration, and included commercial traffic, increasing model realism relative to earlier versions (Rakow et al., 2025). Automated generation of lanes and traffic signals from OpenStreetMap has further streamlined modeling and enabled the testing of adaptive signal strategies (Ziemke and Braun, 2021).

Additional studies have employed MATSim to assess on-demand mobility solutions. Prebooking mechanisms have been introduced into demand-responsive transport simulations (Hörl et al., 2023), and reinforcement learning approaches such as Q-learning have been applied to fleet rebalancing (Chouaki et al., 2022). Further developments include modeling peer-to-peer car sharing through synchronization concepts for limited vehicle and parking resources, tested in a Berlin case study (Hörl et al., 2024). Integration of MATSim with the FEATHERS activity-based model has expanded its policy scope to encompass activity participation and location choice (Ziemke et al., 2021).

Taken together, these studies demonstrate that MATSim functions as a highly adaptable simulation ecosystem extending far beyond traditional traffic flow analysis. From cycling and pedestrian policy assessment to digital-twin development and fleet-based mobility services, MATSim enables detailed representation of behavioral dynamics, incorporation of advanced algorithms, and evaluation of long-term sustainability strategies. Its agent-based foundation makes it a key tool for linking individual decision processes with large-scale mobility planning.

References

1
Products and Solutions
1. Aimsun
(2025)
2
Traffic Analysis Toolbox Volume I, Traffic Analysis Tools Primer
(2004)
3
A comparative analysis of currently used microscopic, macroscopic, and mesoscopic traffic simulation software
1. EA Algherbal
2. NT Ratrout
(2025)
Transportation Research Procedia 84:495–503.
https://doi.org/10.1016/j.trpro.2025.03.101
- Google Scholar
4
Review of driving-behaviour simulation: VISSIM and artificial intelligence approach
(2024)
Heliyon 10:e25936.
https://doi.org/10.1016/j.heliyon.2024.e25936
- Google Scholar
5
A MATSim model methodology to generate cycling-focused transport scenarios in England
(2024)
Journal of Urban Mobility 5:100078.
https://doi.org/10.1016/j.urbmob.2024.100078
- Google Scholar
6
Assessing the impact of different urban morphology scenarios on air pollutant emissions distribution
1. B Augusto
2. D Lopes
3. S Rafael
4. MC Coelho
5. J Ferreira
(2024)
The Science of the Total Environment 950:S0048-9697(24)05491-3.
https://doi.org/10.1016/j.scitotenv.2024.175341
- Google Scholar
7
Calibration of microsimulation model for tight urban diamond interchange under heterogeneous traffic
1. M Azam
2. OC Puan
3. SA Hassan
4. N Mashros
(2019)
IOP Conference Series 527:012077.
https://doi.org/10.1088/1757-899X/527/1/012077
- Google Scholar
8
Transportation Research Record
1. KG Baass
(1981)
Design of zonal systems for aggregate transportation planning models, Transportation Research Record.
- Google Scholar
9
Agent-based simulation of travel demand
1. M Balmer
2. K Meister
3. M Rieser
4. K Nagel
5. KW Axhausen
(2008)
2nd TRB Conference on Innovations in Travel Modeling (ITM 2008).
- Google Scholar
10
Microsimulation modelling in VISSIM on short-term and long-term improvements for mangalore city road network
1. MM Bandi
2. V George
(2020)
Transportation Research Procedia 48:2725–2743.
https://doi.org/10.1016/j.trpro.2020.08.243
- Google Scholar
11
Calibration of vehicle and driver characteristics in VISSIM and ANN-based sensitivity analysis
1. MM Bandi
2. V George
(2020)
International Journal of Microsimulation 13:79–101.
https://doi.org/10.34196/IJM.00219
- Google Scholar
12
How does the traffic behavior change by using SUMO traffic generation tools
(2022)
Computer Communications 181:1–13.
https://doi.org/10.1016/j.comcom.2021.09.023
- Google Scholar
13
Studies in the Economics of Transportation
(1956)
OR 7:146.
https://doi.org/10.2307/3007560
- Google Scholar
14
5–33, Discrete Choice Methods and their Applications to Short Term Travel Decisions, p, 10.1007/978-1-4615-5203-1_2
1. M Ben-Akiva
2. M Bierlaire
(1999)
5–33, Discrete Choice Methods and their Applications to Short Term Travel Decisions, p, 10.1007/978-1-4615-5203-1_2.
- Google Scholar
15
Mathematical models for transportation demand analysis
1. M Bierlaire
(1997)
Transportation Research Part A 31:86.
https://doi.org/10.1016/S0965-8564(97)88373-5
- Google Scholar
16
Microscopic simulation of bicycle traffic flow incorporating cyclists’ heterogeneous dynamics and non-lane-based movement strategies
(2024)
Simulation Modelling Practice and Theory 135:102986.
https://doi.org/10.1016/j.simpat.2024.102986
- Google Scholar
17
173–232, Traffic Simulation with Aimsun, p, 10.1007/978-1-4419-6142-6_5
1. J Casas
2. JL Ferrer
3. D Garcia
4. J Perarnau
5. A Torday
(2010)
173–232, Traffic Simulation with Aimsun, p, 10.1007/978-1-4419-6142-6_5.
- Google Scholar
18
The law of retail gravitation applied to traffic engineering
1. HJ Casey
(1955)
Traffic Quarterly, 9, http://onlinepubs.trb.org/Onlinepubs/trr/1991/1305/1305-010.pdf.
- Google Scholar
19
Determination of urban traveling patterns with high spatial and temporal resolution by using GPS cell phone data
1. LF Chaparro Sierra
2. JI Huertas
(2024)
IEEE Access 12:169643–169658.
https://doi.org/10.1109/ACCESS.2024.3492003
- Google Scholar
20
Transport emission modelling based on a bottom-up approach to facilitate sustainable transport planning
1. A Charly
2. B Caulfield
(2025)
Sustainable Futures 9:100435.
https://doi.org/10.1016/j.sftr.2025.100435
- Google Scholar
21
Implementing reinforcement learning for on-demand vehicle rebalancing in MATSim
(2022)
Procedia Computer Science 201:134–141.
https://doi.org/10.1016/j.procs.2022.03.020
- Google Scholar
22
Luxembourg SUMO Traffic (LuST) Scenario: 24 hours of mobility for vehicular networking research
1. L Codeca
2. R Frank
3. T Engel
(2015)
2015 IEEE Vehicular Networking Conference (VNC), 10.1109/VNC.2015.7385539.
- Google Scholar
23
Monaco SUMO Traffic (MoST) Scenario: A 3D Mobility Scenario for Cooperative ITS
1. L Codeca
2. J Härri
(2018)
43–29, Monaco SUMO Traffic (MoST) Scenario: A 3D Mobility Scenario for Cooperative ITS, p, 10.29007/1zt5.
- Google Scholar
24
SAGA: An activity-based multi-modal mobility scenariogenerator for SUMO
1. L Codeca
(2022)
SUMO Conference Proceedings 1:39–58.
https://doi.org/10.52825/scp.v1i.99
- Google Scholar
25
Impacts of Zonal Reconfigurations on Travel Demand Forecasts
1. CC Crevo
(1991)
Transportation Research Record.
- Google Scholar
26
A probabilistic multipath traffic assignment model which obviates path enumeration
1. RB Dial
(1971)
Transportation Research 5:83–111.
https://doi.org/10.1016/0041-1647(71)90012-8
- Google Scholar
27
Predicting transport mode choice preferences in a university district with decision tree-based models
(2023)
City and Environment Interactions 20:100118.
https://doi.org/10.1016/j.cacint.2023.100118
- Google Scholar
28
Modeling and Simulation of Highway Traffic Using a Cellular Automaton Approach
1. D Ding
(2011)
Uppsala Universitet.
- Google Scholar
29
Traffic Analysis Toolbox Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software
(2004)
Traffic Analysis Toolbox Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software.
- Google Scholar
30
Scenario-based mobility simulation for shrinking cities using ABATA: A case study of Iksan, South Korea
1. JK Eom
2. K-S Lee
3. S Ko
4. JH Min
(2026)
Journal of Transport Geography 131:104564.
https://doi.org/10.1016/j.jtrangeo.2026.104564
- Google Scholar
31
Social acceptance assessment of alternative urban sustainable transport technologies in intermediate cities using AHP and SUMO
(2025)
World Development Sustainability 6:100229.
https://doi.org/10.1016/j.wds.2025.100229
- Google Scholar
32
Impact of urban block shape on traffic and air quality: A SUMO-based comparative study of rectangular, radial, and triangular forms
(2025)
Transportation Research Interdisciplinary Perspectives 31:101413.
https://doi.org/10.1016/j.trip.2025.101413
- Google Scholar
33
63–93, Microscopic Traffic Flow Simulator VISSIM, p, 10.1007/978-1-4419-6142-6_2
1. M Fellendorf
2. P Vortisch
(2010)
63–93, Microscopic Traffic Flow Simulator VISSIM, p, 10.1007/978-1-4419-6142-6_2.
- Google Scholar
34
Planning of an integrated urban transportation system based on macro – simulation and MCDM/A Methods
1. S Fierek
2. J Zak
(2012)
Procedia - Social and Behavioral Sciences 54:567–579.
https://doi.org/10.1016/j.sbspro.2012.09.774
- Google Scholar
35
Impact of radical bike lane allocation on bi-modal urban road network traffic performance: A simulation case study
(2025)
Case Studies on Transport Policy 22:101583.
https://doi.org/10.1016/j.cstp.2025.101583
- Google Scholar
36
Time function iteration
1. KP Furnes
(1965)
Traffic Engineering and Control, 7.
- Google Scholar
37
Introducing fine grained energy consumption variables into a public passenger transport simulation in SUMO
(2024)
Procedia Computer Science 232:1890–1899.
https://doi.org/10.1016/j.procs.2024.02.011
- Google Scholar
38
A behavioural car-following model for computer simulation
1. PG Gipps
(1981)
Transportation Research Part B 15:105–111.
https://doi.org/10.1016/0191-2615(81)90037-0
- Google Scholar
39
Fundamentals of Driving Patterns and Driving Cycles
(2025)
91–102, Driving Cycle Validation, Fundamentals of Driving Patterns and Driving Cycles, Springer Nature, p, 10.1007/978-3-031-76863-7_6.
- Google Scholar
40
Microscopic traffic simulation models for connected and automated vehicles (CAVs) – state-of-the-art
(2020)
Procedia Computer Science 170:474–481.
https://doi.org/10.1016/j.procs.2020.03.091
- Google Scholar
41
ABMTrafSimCA: An agent-based modelling open-source software tool for traffic modelling and simulation based on cellular automata
1. P Górka
2. K Małecki
(2024)
SoftwareX 27:101813.
https://doi.org/10.1016/j.softx.2024.101813
- Google Scholar
42
A multi-objective method for large multi-lane roundabout design through microscopic traffic simulation and SSAM analysis
1. M Guerrieri
2. M Khanmohamadi
(2024)
Sustainable Futures 8:100313.
https://doi.org/10.1016/j.sftr.2024.100313
- Google Scholar
43
Selectıng pedestrianization types for sustainable urban mobility: A case study of Trabzon, Turkey applicatıon
(2025)
Case Studies on Transport Policy 21:101554.
https://doi.org/10.1016/j.cstp.2025.101554
- Google Scholar
44
A time-use activity-pattern recognition model for activity-based travel demand modeling
(2019)
Transportation 46:1369–1394.
https://doi.org/10.1007/s11116-017-9840-9
- Google Scholar
45
Traffic Congestion and Traffic Safety in the 21st Century: Challenges, Innovations, and Opportunities
1. A Halati
2. H Lieu
3. S Walker
(1997)
CORSIM - corridor Traffic Simulation Model, Traffic Congestion and Traffic Safety in the 21st Century: Challenges, Innovations, and Opportunities.
- Google Scholar
46
Optimization-based assisted calibration of traffic simulation models
(2015)
Transportation Research Part C 55:100–115.
https://doi.org/10.1016/j.trc.2015.01.018
- Google Scholar
47
Estimating short-term travel demand models that incorporate personally owned autonomous vehicles
1. M Harb
2. JL Walker
3. J Malik
4. G Circella
(2022)
Travel Behaviour and Society 26:279–289.
https://doi.org/10.1016/j.tbs.2021.10.008
- Google Scholar
48
A systematic review of machine learning classification methodologies for modelling passenger mode choice
1. T Hillel
2. M Bierlaire
3. M Elshafie
4. Y Jin
(2021)
Journal of Choice Modelling 38:100221.
https://doi.org/10.1016/j.jocm.2020.100221
- Google Scholar
49
Traffic Analysis Toolbox Volume IV: Guidelines for Applying CORSIM Microsimulation Modeling Software
(2007)
50
Resource-constrained replanning in MATSim applied to the simulation of peer-to-peer car sharing services
1. S Hörl
2. J Sobieraj
3. S Axer
4. H Rewald
(2023)
Procedia Computer Science 220:698–703.
https://doi.org/10.1016/j.procs.2023.03.091
- Google Scholar
51
Evaluating prebooked on-demand mobility services using MATSim
1. S Hörl
2. T Chouaki
3. O Ludwig
4. H Rewald
5. S Axer
(2024)
Procedia Computer Science 238:763–770.
https://doi.org/10.1016/j.procs.2024.06.089
- Google Scholar
52
Holistic urban mobility assessment in urban centers: Developing an integrated digital twin tool for Tec District
(2026)
Journal of Urban Management, 10.1016/j.jum.2026.04.010.
- Google Scholar
53
An integrated VISSIM-SSAM approach to predicting and mitigating pedestrian crashes and severity along urban crossings
(2024)
Case Studies on Transport Policy 15:101153.
https://doi.org/10.1016/j.cstp.2024.101153
- Google Scholar
54
Programa Parcial de Desarrollo Urbano del Distrito Tec
1. ImplancMTY
(2021)
55
Modelling of bicycle traffic in the cities using VISUM
(2017)
Procedia Engineering 187:435–441.
https://doi.org/10.1016/j.proeng.2017.04.397
- Google Scholar
56
Comparison of Car-Following Models
1. J Janson Olstam
2. A Tapani
(2004)
Comparison of Car-Following Models.
- Google Scholar
57
Evaluation of different vehicle following models under mixed traffic conditions
(2013)
Procedia - Social and Behavioral Sciences 104:390–401.
https://doi.org/10.1016/j.sbspro.2013.11.132
- Google Scholar
58
Achieving realistic cyclist behavior in SUMO using the SimRa dataset
(2023)
Computer Communications 205:97–107.
https://doi.org/10.1016/j.comcom.2023.04.015
- Google Scholar
59
Large-scale agent-based simulation model of pedestrian traffic flows
(2023)
Computers, Environment and Urban Systems 105:102021.
https://doi.org/10.1016/j.compenvurbsys.2023.102021
- Google Scholar
60
An example of microscopic car models validation using the open source traffic simulation SUMO
(2002)
Proceedings of Simulation in Industry 14th European Simulation Symposium.
- Google Scholar
61
Microscopic modeling of traffic flow: Investigation of collision free vehicle dynamics
1. S Krauß
(1998)
D L R - Forschungsberichte, 8.
- Google Scholar
62
Traffic Analysis Toolbox Volume II: Decision support methodology for selecting traffic analysis tools
(2004)
- Google Scholar
63
Design of the city logistics simulation model using PTV VISSIM software
1. T Kučera
2. J Chocholáč
(2021)
Transportation Research Procedia 53:258–265.
https://doi.org/10.1016/j.trpro.2021.02.033
- Google Scholar
64
Assessing the potential for traffic carbon emission reductions through residential travel mode shifts: insights from massive vehicle trajectory data and scenario simulations
1. S Li
2. S Li
3. S Xu
4. H Xu
5. X Chen
6. Q Mu
7. Z Tan
(2025)
International Journal of Applied Earth Observation and Geoinformation 142:104684.
https://doi.org/10.1016/j.jag.2025.104684
- Google Scholar
65
A cellular automaton traffic flow model with advanced decelerations
1. Y Liu
(2012)
Mathematical Problems in Engineering, 2012, 10.1155/2012/717085.
- Google Scholar
66
Microscopic Traffic Simulation using SUMO
1. PA Lopez
2. E Wiessner
3. M Behrisch
4. L Bieker-Walz
5. J Erdmann
6. Y-P Flotterod
7. R Hilbrich
8. L Lucken
9. J Rummel
10. P Wagner
(2018)
2018 21st International Conference on Intelligent Transportation Systems (ITSC. pp. 2575–2582.
https://doi.org/10.1109/ITSC.2018.8569938
- Google Scholar
67
Subdivision of urban traffic area based on the combination of static zoning and dynamic zoning
1. J Luo
2. Q Zhang
(2021)
Discrete Dynamics in Nature and Society 2021:1–17.
https://doi.org/10.1155/2021/9954267
- Google Scholar
68
An integration modeling framework for individual-scale daily mobility estimation
1. N Luo
2. A Nara
3. HL Khoo
4. M Chen
(2024)
Travel Behaviour and Society 34:100650.
https://doi.org/10.1016/j.tbs.2023.100650
- Google Scholar
69
Large-scale microscopic simulation of taxi services. Berlin and Barcelona case studies
(2016)
Journal of Ambient Intelligence and Humanized Computing 7:385–393.
https://doi.org/10.1007/s12652-016-0366-3
- Google Scholar
70
Quantifying the vulnerability of road networks to flood-induced closures using traffic simulation
(2024)
Transportation Engineering 17:100262.
https://doi.org/10.1016/j.treng.2024.100262
- Google Scholar
71
A traffic analysis zone definition: a new methodology and algorithm
(2009)
Transportation 36:581–599.
https://doi.org/10.1007/s11116-009-9214-z
- Google Scholar
72
The Four Step Model
1. MG McNally
(2008)
73
Development of public transit measures to mitigate the impact of COVID-19 on pedestrians and station performance using PTV vissim simulation
1. BH Miao
2. S Saidi
(2025)
Transportation Research Procedia 82:2294–2312.
https://doi.org/10.1016/j.trpro.2024.12.187
- Google Scholar
74
Project-based agent/activity microsimulation of travel demand: theory, experience & future prospects
1. EJ Miller
(2024)
Procedia Computer Science 238:111–118.
https://doi.org/10.1016/j.procs.2024.06.005
- Google Scholar
75
Evaluation of alternative Ramp Metering scenarios on freeway on-ramp with the use of microscopic simulation software Vissim
1. DZ Mitkas
2. I Politis
(2020)
Transportation Research Procedia 45:483–490.
https://doi.org/10.1016/j.trpro.2020.03.042
- Google Scholar
76
Demand model estimation from smartphone data. An application to assert new urbanistic development scenarios
(2022)
Transportation Research Procedia 62:270–277.
https://doi.org/10.1016/j.trpro.2022.02.034
- Google Scholar
77
Advancements in traffic simulation for enhanced road safety: A review
(2024)
Simulation Modelling Practice and Theory 137:103017.
https://doi.org/10.1016/j.simpat.2024.103017
- Google Scholar
78
Two-lane traffic rules for cellular automata: A systematic approach
1. K Nagel
2. DE Wolf
3. P Wagner
4. P Simon
(1998)
Physical Review E 58:1425–1437.
https://doi.org/10.1103/PhysRevE.58.1425
- Google Scholar
79
Eclipse SUMO and Unity 3D integration for emission studies based on driving behaviour in virtual reality environment
(2025)
Procedia Computer Science 257:396–403.
https://doi.org/10.1016/j.procs.2025.03.052
- Google Scholar
80
Using visum traffic simulation model to evaluate the traffic network on the economic corridor: Lao Cai – Hanoi – Hai Phong – Quang Ninh
1. TA Nguyen
2. HV Nguyen
3. TA Pham
4. DT Do
5. HC Pham
6. SS Tu
(2025)
Transportation Research Procedia 85:83–93.
https://doi.org/10.1016/j.trpro.2025.03.137
- Google Scholar
81
Large-scale activity-based demand generation modeling: A literature review and exploration of potential approaches
(2025)
Transportation Engineering 20:100329.
https://doi.org/10.1016/j.treng.2025.100329
- Google Scholar
82
Overview of application of traffic simulation model
1. NN Nor Azlan
2. M Md Rohani
(2018)
MATEC Web of Conferences 150:03006.
https://doi.org/10.1051/matecconf/201815003006
- Google Scholar
83
Comparison of Car-Following Models
1. JJ Olstam
2. A Tapani
(2004)
Swedish National Road and Transportation Research Institute.
- Google Scholar
84
Developing optimal transportation analysis zones using GIS
1. WA O’Neill
(1991)
ITE Journal 61:12.
- Google Scholar
85
Optimal zoning systems for spatial interaction models
1. S Openshaw
(1977)
Environment and Planning A 9:169–184.
https://doi.org/10.1068/a090169
- Google Scholar
86
Modelling Transport
1. J de D Ortúzar
2. LG Willumsen
(2011)
Wiley.
https://doi.org/10.1002/9781119993308
- Google Scholar
87
Digital Twin for mobility: simulation integrated approach for demand matrices calibration with empirical data
1. AL Pala
2. G Felici
3. V Torrisi
4. D Donato Russo
5. A Salvatore
6. P Leonardi
7. G Stecca
8. M Ignaccolo
9. G Inturri
(2025)
Transportation Research Procedia 86:516–523.
https://doi.org/10.1016/j.trpro.2025.04.065
- Google Scholar
88
Metamodel-based calibration of large-scale multimodal microscopic traffic simulation
(2021)
Transportation Research Part C 124:102859.
https://doi.org/10.1016/j.trc.2020.102859
- Google Scholar
89
Empirical study of bicycle traffic characteristics relevant for microscopic simulation
(2025)
Journal of Cycling and Micromobility Research 3:100054.
https://doi.org/10.1016/j.jcmr.2024.100054
- Google Scholar
90
Evaluation of perception performance modeling in microscopic traffic simulation
(2025)
Transportation Research Procedia 86:459–466.
https://doi.org/10.1016/j.trpro.2025.04.058
- Google Scholar
91
Traffic Zones Discretization and Origin-Destination Matrix Estimation by means of Transport Demand and Satellite Data Fusion
1. T Prosperi
2. G Moser
3. N Sacco
4. F Rebora
(2021)
2021 IEEE International Intelligent Transportation Systems Conference (ITSC. pp. 3217–3222.
https://doi.org/10.1109/ITSC48978.2021.9565123
- Google Scholar
92
Evolution of traffic microsimulation and its use for modeling connected and automated vehicles
1. N Raju
2. H Farah
(2021)
Journal of Advanced Transportation 2021:1–29.
https://doi.org/10.1155/2021/2444363
- Google Scholar
93
Advancing the MATSim open berlin scenario: improvements in transport scenario generation and calibration methods
(2025)
Transportation Research Procedia 86:732–739.
https://doi.org/10.1016/j.trpro.2025.04.091
- Google Scholar
94
Car-Following models: an experiment based benchmarking
1. P Ranjitkar
2. A Kawamura
(2005)
Journal of the Eastern Asia Society for Transportation Studies 6.
https://doi.org/10.11175/easts.6.1582
- Google Scholar
95
Car-following models: an experiment based benchmarkinG
(2005)
Journal of the Eastern Asia Society for Transportation Studies, 6.
- Google Scholar
96
Measurement and analysis of heterogeneous road transport parameters using smart traffic analyzer and SUMO simulator:an experimental approach
(2025)
Measurement 242:116233.
https://doi.org/10.1016/j.measurement.2024.116233
- Google Scholar
97
Applying optimization algorithms for spatial estimation of travel demand variables
(2021)
Transportation Research Interdisciplinary Perspectives 10:100369.
https://doi.org/10.1016/j.trip.2021.100369
- Google Scholar
98
Analyzing the effect of depopulation on the spatial structure of the city of Łódź, Poland: development and application of an integrated land use and transportation model
(2024)
Habitat International 143:102992.
https://doi.org/10.1016/j.habitatint.2023.102992
- Google Scholar
99
A comparative study of urban road traffic simulators
(2016)
MATEC Web of Conferences 81:05002.
https://doi.org/10.1051/matecconf/20168105002
- Google Scholar
100
A systematic comparative evaluation of machine learning classifiers and discrete choice models for travel mode choice in the presence of response heterogeneity
(2022)
Expert Systems with Applications 193:116253.
https://doi.org/10.1016/j.eswa.2021.116253
- Google Scholar
101
Mapping of individual transportation traffic-related externalities in an intercity corridor
(2022)
Transportation Research Procedia 62:672–679.
https://doi.org/10.1016/j.trpro.2022.02.083
- Google Scholar
102
Evaluation of emissions in traffic reduction and pedestrianization scenarios in Madrid
(2021)
Transportation Research Part D 100:103064.
https://doi.org/10.1016/j.trd.2021.103064
- Google Scholar
103
Implications of static and dynamic road pricing strategies in the era of autonomous and shared autonomous vehicles using simulation-based dynamic traffic assignment: the case of budapest
(2022)
Research in Transportation Economics 95:101231.
https://doi.org/10.1016/j.retrec.2022.101231
- Google Scholar
104
Practical guidance for using mesoscopic simulation tools
1. B Sun
2. J Appiah
3. BB Park
(2020)
Transportation Research Procedia 48:764–776.
https://doi.org/10.1016/j.trpro.2020.08.078
- Google Scholar
105
Development and evaluation of an urban-scale traffic simulation for reducing the number of traffic accidents
1. K Suzuki
2. T Uchitane
3. N Mukai
4. K Iwata
5. N Ito
6. Y Jiang
(2024)
Procedia Computer Science 246:490–499.
https://doi.org/10.1016/j.procs.2024.09.429
- Google Scholar
106
Transport demand model estimation from traffic counts
1. OZ Tamin
2. LG Willumsen
(1989)
Transportation, 16, 10.1007/BF00223044.
- Google Scholar
107
Microscopic calibration and validation of car-following models – a systematic approach
1. M Treiber
2. A Kesting
(2013)
Procedia - Social and Behavioral Sciences 80:922–939.
https://doi.org/10.1016/j.sbspro.2013.05.050
- Google Scholar
108
SMARTEST Project. Review of Micro-Simulation Models
1. University of Leeds
2. ITS
(2000)
109
Analyzing transportation mode interactions using agent-based models
(2023)
SN Applied Sciences 5:357.
https://doi.org/10.1007/s42452-023-05609-z
- Google Scholar
110
Optimizing vehicle dynamics co-simulation performance by introducing mesoscopic traffic simulation
(2023)
Simulation Modelling Practice and Theory 125:102739.
https://doi.org/10.1016/j.simpat.2023.102739
- Google Scholar
111
A simulation study on traffic conflict risk under shockwave propagation at signalized intersections
1. J Wang
2. H Yang
3. Y Zhu
4. K Xie
5. Z Yan
6. Q Pu
(2025)
Accident Analysis & Prevention 218:108081.
https://doi.org/10.1016/j.aap.2025.108081
- Google Scholar
112
Research on virtual emergency lane optimization and control strategy based on VISSIM expressway
1. Y Wang
2. M Zhang
3. X Yang
4. T Wei
5. X Zhang
6. L Wang
(2025)
Ain Shams Engineering Journal 16:103750.
https://doi.org/10.1016/j.asej.2025.103750
- Google Scholar
113
How to improve urban transportation planning in big data era? a practice in the study of traffic analysis zone delineation
1. B Yang
2. Y Tian
3. J Wang
4. X Hu
5. S An
(2022)
Transport Policy 127:1–14.
https://doi.org/10.1016/j.tranpol.2022.08.002
- Google Scholar
114
A co-simulation system that integrates MATSim with a set of external fleet simulations
1. H Yang
2. E Wong
3. H Davis
4. JYJ Chow
(2024)
Simulation Modelling Practice and Theory 134:102957.
https://doi.org/10.1016/j.simpat.2024.102957
- Google Scholar
115
Developing a comprehensive large-scale co-simulation for replication of automated driving in urban traffic scenarios
(2024)
Transportation Research Procedia 78:522–529.
https://doi.org/10.1016/j.trpro.2024.02.065
- Google Scholar
116
Agent-based approach to travel demand modeling: exploratory analysis
1. L Zhang
2. D Levinson
(2004)
Transportation Research Record 1898:28–36.
https://doi.org/10.3141/1898-04
- Google Scholar
117
Automated generation of traffic signals and lanes for MATSim based on openstreetmap
1. T Ziemke
2. S Braun
(2021)
Procedia Computer Science 184:745–752.
https://doi.org/10.1016/j.procs.2021.03.093
- Google Scholar
118
Expanding the analysis scope of a MATSim transport simulation by integrating the FEATHERS activity-based demand model
(2021)
Procedia Computer Science 184:753–760.
https://doi.org/10.1016/j.procs.2021.04.022
- Google Scholar

Article and author information

Author details

Jenny Díaz-Ramírez

Tecnologico de Monterrey, Sustainable Energy Lab. Expedition FEMSA Building, Av. Garza Sada 2501 Sur. CP 64700, Monterrey, Mexico

"This ORCID iD identifies the author of this article:" 0000-0001-5820-7512
Erfan Nobil

Universidad de las Américas Puebla, Puebla de Zaragoza, Puebla, Mexico

"This ORCID iD identifies the author of this article:" 0000-0001-5904-5199
Juan Alberto Estrada-García

University of Michigan at Ann Arbor, Michigan, United States

"This ORCID iD identifies the author of this article:" 0000-0003-2308-1897
Jesús Adrián Martínez-Hernández

Western University, London, Ontario, Canada

"This ORCID iD identifies the author of this article:" 0000-0001-9062-4949
Jose Ignacio Huertas

Tecnologico de Monterrey, Sustainable Energy Lab. Expedition FEMSA Building, Av. Garza Sada 2501 Sur. CP 64700, Monterrey, Mexico

For correspondence
jhuertas@tec.mx

"This ORCID iD identifies the author of this article:" 0000-0003-4508-6453

Funding

No external funding for this article is reported. Internal Grant: This work was partially funded by the Challenge-Based Research Funding Program (Tecnologico de Monterrey). Grant 5607TM-10-393 MAITEC.

Acknowledgements

Special thanks to RELIEVE, the Latin American Network of Researchers in Energy and Vehicles, for supporting the collaboration among the research team.

Publication history

Version of Record published: August 25, 2026 (version 1)

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Article PDF

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

BibTeX
RIS

Open citations (links to open the citations from this article in various online reference manager services)

Mendeley

Classification of traffic simulation models by scale

Car-Following traffic models

Proposed methodology for implementing traffic simulation at any scale, aiming at evaluating strategies to improve mobility in a given region.

Zoning examples of the same area: District Tec in Monterrey, Mexico.

Zoning methods

Required input data

Transport demand models

Examples of OD matrices representations (a) single trips, one mode, (b) activity chains.

Different formulations of OD matrices required in simulation models

Methods for travel demand modelling

Most common simulation software tools

Comparison of available traffic simulation software capabilities

Metrics used for calibrating traffic models.

Recommended error functions in traffic simulation

Model performance metrics

Validation example based on traffic volume profiles using five goodness-of-fit measures.

Examples of expected results from different scenarios

Examples of improvement implementations in the network.

Schemes of activity-based trips.

Author details

Jenny Díaz-Ramírez

Erfan Nobil

Juan Alberto Estrada-García

Jesús Adrián Martínez-Hernández

Jose Ignacio Huertas

For correspondence

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Categories and tags