“Simulation is the modeling of individual vehicle movements on a second or sub second basis for assessing the traffic performance of highways and street systems, transit and pedestrians” (Source: FHWA Traffic Analysis Toolbox). Travel demand models, which are sometimes called static models or deterministic models, assume there is no variability in driver-vehicle characteristics and they assess the operations of segments in the roadway or transit system rather than modeling the individual vehicles using those roadway system segments.For example, travel demand models calculate and report volume, speed and volume/capacity ratios for each segment of a freeway or arterial. Simulation models, on the other hand, simulate the characteristics and interactions of individual vehicles as they move through the system. They essentially produce trajectories of vehicles as they move through the network, interacting with control devices and with each other. The processing logic for simulation includes algorithms and rules describing how vehicles move and interact with each other, including car following, acceleration, deceleration, lane changing, and passing maneuvers. This provides more detailed and accurate results for the system being studied as compared to static models. Because of this, the output of simulation models, when properly calibrated, is more accurate, realistic and detailed than deterministic travel models.
The HCM can evaluate the impacts of certain congested network scenarios, such as on a freeway network or isolated intersections as long as the effect, limitations, and context of the results is understood. Yet, other congested scenarios are beyond the limitations of the HCM, such as signalized arterial networks or freeway-arterial interfaces and freeway queues which extend to adjacent segments and cause congestion. However, HCM analyses can still be useful in the scenarios beyond its limits. For instance, the HCM enables the analyst to quickly assess the existing conditions of a given location, confirm “hot spot” locations, and compared to the complexity of simulation is an easy tool to aid in the realization of the cause and effects of modified geometry and operational schemes at specific locations. Additionally, the results of the HCM may aid in the establishment of a simulation analysis purpose, need, scope and physical network limits.
Since Highway Capacity Manual results are typically not accurate for highly congested conditions and where the volume/capacity ratio exceeds 1.0, in those cases simulation should be considered. Also, where downstream congestion affects another roadway segment, HCM results can be misleading (such as where a bottleneck on a freeway extends back to other segments further upstream). For these reasons, and because simulation can be more closely calibrated to actual conditions, microsimulation modeling is often the best tool for highly congested conditions as opposed to HCM or other methods.
Realizing that the HCM and other methods have certain limitations, the Federal Highway Administration (FHWA) and other partnering agencies and practitioners have recommended complimenting the use of the HCM with other traffic analysis tools. This is driven by a desire to better understand and support a robust operations analysis and decision making for future improvements and investments in the transportation system. One recurring example in which additional tools are being considered is for scenarios where one or more features of an interchange or corridor operate in unstable or over-saturated conditions for a given design period, such as the highly congested peak periods. Due to the interdependence of the individual features of a corridor or interchange, simulation increases the understanding of how an improvement alternative may function in the future.Simulation helps more accurately test various types of improvements such as physical changes to the system, Intelligent Transportation Systems improvements, managed lanes and toll lanes and person throughput.
Simulation tools utilize algorithms that consider and reflect the interaction of individual vehicles throughout the given roadway network. As a Stochastic Tool, simulation tools assign probabilities to many of the decisions drivers make on a sub-second level (for example; whether or not to make a lane change) for the purpose of better reflecting the variations in response seen in actual driver’s actions in the field.Random numbers are generated within the simulation tool to account for the fact that drivers do not always make the same decisions under the same conditions. As a result, a fixed set of assumptions and known conditions could generate different output results in separate runs if these random decisions were slightly different. To account for this, an analyst should perform multiple runs using the same assumptions and conditions but with different random seeds as inputs to those multiple runs, along with a statistical analysis to increase their confidence of the overall analysis results. Single runs that are not representative of the random nature of these tools will reduce the credibility of the analysis and reduce the acceptance of the results.Enough random seed runs should be completed to ensure that the ‘noise’ associated with the of the stochastic decisions made within a simulation model is minimized to a confidence level that the analyst is comfortable with.
The respective documentation may be developed commensurate to the needs and the scope of the project. But typically, documentation can include a model methodology memo, a model calibration report and model run results report including tables, graphics and animations with performance measure results.
While varying somewhat by software platform, microsimulation models can produce a wide range of detailed performance measures including, but not limited to, speeds, delay, travel time, trip length, vehicle miles of travel, vehicle hours of travel, stopped time, number of stops, congested queue lengths and level of service. Travel demand models can also produce some of these same performance measures as output, but the results are less accurate and detailed due to the deterministic nature of travel demand models and the fact that they analyze the system at an aggregate level segment by segment and across the analysis period rather than for each vehicle, as with the simulation models. The performance measures reported by simulation models are thus the result of the sum total of all of the individual vehicles that are traveling through the system, and can be reported at various different roadway elements (links, nodes, signals, etc.) and at varying times throughout the simulation analysis period. Simulation models can also produce realistic animations of the system being analyzed and produce animated video clips of the system as output to demonstrate potential future conditions to decision makers and stakeholders interested in improvement projects.
Typical data requirements for developing and running a simulation model include detailed network geometric conditions (lanes by type, ramps, ramp meter locations, gore point locations, intersection configurations, auxiliary lane locations and length, etc.), control data (signal timing and signage), traffic flow count data (daily and hourly), counts by classification (trucks, buses, auto), locations of queues, slowing and bottlenecks, corridor travel times and travel demand information (trip origins and destinations).
Simulation can be used for many types of transportation systems including freeways, arterial roadways, intersections, freeway-to-freeway interchanges, freeway/arterial interchanges, managed lanes and tolled lanes and transit facilities.
Simulation can be used to test many types of improvements including adding new lanes (mixed flow, general purpose, tolled or managed lanes), changing signal timing patterns, adding auxiliary lanes, removing lanes (road diets), adding bus rapid transit to a corridor, changing interchange configurations, work zone changes, various Intelligent Transportation System deployments, Active Transportation Management Strategies (ATMS) and Integrated Corridor Management (ICM), and even future technologies such as connected and autonomous vehicles.
There are a number of factors that impact the time and effort in applying simulation tools. In summary, it is due to the time and complexity in collecting adequate field data for calibrating the model, setting up and coding a base model, adequately calibrating the base model to the observed conditions, and performing multiple runs for various improvement alternatives.
This varies depending on the tool/software platform and it is suggested that the vendor be consulted to understand the details and sensitivity of a respective product. It however is important to understand that the placement and coding of links, nodes and connectors in certain combinations, order and spacing may impact the reported measure of effectiveness for a given location in the network. While two models representing the same network may have similar global operating outputs and animation, a specific link, node or connector may have drastically different numerical output due to how the network was coded.As such, it is important to follow a set of good practices and consistency in how a network is coded and represented across a network and between simulated alternative scenarios.
Typically the simulation model should fully cover the congested study area to be studied.For example, on a freeway, all segments that experience congestion in a contiguous area should be included to cover the extent of the current congestion or bottleneck. This is not always feasible as urban congestion can cover large areas, and in these cases, the area can be chosen to best replicate the likely future improvement area.If congestion extends beyond the study area, the model can be adjusted at the end points to replicate the congestion beyond the study area.Mesoscopic models can generally cover a larger network area due to the lower level of detail and more simplified vehicle interaction models, which reduce the overall associated computer resources needed to model the same study area.
A Dynamic Traffic Assignment (DTA) model estimates the evolution and propagation of congestion through detailed models that better simulate the point to point demands for travel, the sometimes dynamic nature of the network supply and their complex interactions. DTA models seek to provide more detailed means to represent the interaction between travel choices, traffic flows, and time and cost measures in a temporally coherent manner. DTA models aim to describe time-varying network and demand interaction using a behaviorally sound approach.
The equilibrium-seeking DTA models are based on iterative algorithmic procedures that describes individual route choice adjustment. Within each iteration, simulated drivers learned from the previous iterations of what typical operating conditions are and may adjust to a less costly (either in time and/or tolls) route. DTA can also refer to non-equilibrium or ‘one-shot’ simulations, where many vehicles follow certain habitually used paths regardless of the experienced congestion, but a smaller portion of the population would know of and react to changing network performances (e.g. via radio, 511, DMS, or smart phone traffic apps). One-shot DTA simulations are appropriate for simulating non-recurring congestion events, such as short term work zones or vehicle crashes.
Mesoscopic DTA models are suitable for large-scale network applications. Unlike a static model, a DTA model can describe time-dependent dynamics of traffic and replicate the interactions between travelers and the transportation network. Additionally, dynamic attributes of a network (such as opening/closing lanes, varying toll rates, turns prohibitions, etc.) can be properly represented and modeled, rather than a static or average condition.
A simulation-based DTA model is an analysis tool to address complex and dynamic transportation operations and planning issues of urban road networks. For example, a simulation-based DTA model can be used to assess the impacts of ITS and non-ITS technologies on the transportation network, to support decision-making for work zone planning and traffic management, to evaluate different congestion pricing schemes, to plan for special events and emergency situations including evacuation scenarios, to assess crash or non-recurring events, and to perform traffic assignment analyses in conjunction with classical four-step demand models or emerging activity-based and tour-based models.
Each of commonly used simulation software packages have different attributes and are difficult to compare in a broad sense. Each has its own advantages and potential disadvantages depending on the specific nature and details of the study that needs to be completed, and there is no single ‘best’ simulation software.
As part of any traffic analysis project, a review of the possible analysis tools should be conducted at the onset of the project.At this time the pros and cons of each software specific to the needs of the study can be discussed.Further details on each of the simulation software packages can be obtained from the software vendors and their websites.
Mesoscopic simulation is best described as being a level of traffic modeling that is less detailed than a microscopic simulation but more detailed than a macroscopic travel demand model. There is a wide variation in the types of models that are referred to as mesoscopic. In recent years, more traditional simulation software packages are adding the ability to simulation at a mesoscopic level of simulation, further complicating the definition between micro and meso simulation tools.While a DTA model can be microscopic or mesoscopic, some microscopic simulation models are less suitable for large-scale applications and may not fully consider an individual traveler’s route choice decisions across a larger network like is often done in a mesoscopic model.As a result, the terms DTA and mesoscopic are often used interchangeably, albeit incorrectly.
Mesoscopic models fall into two generally types; those that model individual lanes and those that do not model lanes, but rather the whole link.Lane-based simulation is generally more resource intensive than link-based models, but provide better estimates of bottleneck conditions related to over-saturated turn lanes and freeway ramps.Individual vehicles and route choices are usually still simulated, but are sometimes grouped into platoons of vehicles to further simplify the travel time calculations. Traffic controls are also modeling in a wide variety of details depending on the software package, from none being modeled at all to full actuation-based signal timings.
In general, meso simulations algorithms do not use the micro model’s detailed car-following and lane changing models calculated on sub-second level calculations to determine how a vehicle moves across links.Instead, meso model algorithms rely on more aggregated speed/flow/density relationships predicted by traffic flow theory to estimate travel times for simulated vehicles under the demand loads present when the vehicle uses that link.As such, there are a lot fewer calculations needed, and meso simulations can run much faster than simulation model as a trade-off for the lower fidelity algorithms.This allows meso models to cover larger networks, often approaching sub-regional or even city-scale simulation models.