Office of Planning, Environment, & Realty (HEP)
The TRansportation ANalysis and SIMulation System (TRANSIMS) is a set of travel modeling procedures designed to meet the State DOTs' and MPOs' need for more accurate and more sensitive travel forecasts for transportation planning and emissions analysis.
TRANSIMS outputs detailed data on travel, congestion, and emissions; information that is increasingly important to investment decisions and policy setting. Because TRANSIMS simulates and tracks travel by individuals, the benefits to and impacts on different geographies and travel markets can be evaluated as well. Furthermore, TRANSIMS has the capability to evaluate highly congested scenarios and operational changes on highways and transit systems.
TRANSIMS is based on four primary modules: population synthesizer, activity generator, route planner, and traffic microsimulator. Using these components, TRANSIMS estimates activities for individuals and households, plans trips satisfying those activities, assigns trips to routes, and creates a microsimulation of all vehicles, transportation systems, and resulting traffic in a given study area.
TRANSIMS differs from previous travel demand forecasting methods in its underlying concepts and structure. These differences include a consistent and continuous representation of time; a detailed representation of persons and households; time-dependent routing; and a person-based microsimulator. These advances are producing significant changes in the travel forecasting process. To date the TRANSIMS models have been tested with data from Dallas, Texas, and Portland, Oregon.
TRANSIMS has been developed by researchers at the Los Alamos National Laboratory and is available commercially through the IBM TRANSIMS Solution Center. The design for TRANSIMS is based on requirements in the Intermodal Surface Transportation Efficiency Act, the Transportation Equity Act for the 21st Century and Clean Air Act Amendments. The development of TRANSIMS has been funded by the Federal Highway Administration, the Federal Transit Administration and the Office of the Assistant Secretary for Transportation Policy of the United States Department of Transportation, and by the Environmental Protection Agency.
More information on TRANSIMS can be found on the TRANSIMS Open Source site.
TRANSIMS simulates the second-by-second movement of persons and vehicles throughout a region. A simple set of rules is used to govern vehicle movement. A 7.5-meter grid describes vehicle location. This grid results in rough instantaneous speed estimates for individual vehicles however average link speeds are estimated reasonably well for time intervals as short as 15 minutes. The time varying data from the Microsimulator provides a more accurate representation of queuing and congestion over the course of an entire day compared to methods used for typical static equilibrium assignment models.
TRANSIMS does not simulate complex driver behaviors, detailed pedestrian movements, or precise intersection operations. Exacting operational analyses of highway facilities, including simulation of advanced traffic control technologies, are more appropriate for sub-area application of highly specialized traffic simulation models.
Recent applications of TRANSIMS are purposefully drawn from 4-step travel demand models. Person-travel needs have been estimated with trip-based and tour-based models. Travel destinations and modes are modeled with nested choice methods currently used in practice. Iterative feedback is used to refine travel choices. However, the significant differences derive from modeling individual travel choices and simulating these choices using internally consistent definitions of time and personal travel schedules.
The move from aggregate, parameter-driven methods to agent-based, rule-driven methods pre-dates TRANSIMS and has been a long while in the making. Regional microsimulation adds a key strategic capability to the study of travel and transportation. Improved service information for individual travelers allows for better modeling of travel scheduling and peak spreading phenomena. Better speed and acceleration estimates support more detailed congestion, fuel, emission, and safety analyses. The technology embodied in TRANSIMS gives planners and decision-makers an ability to consider transportation issues and prospective solutions in greater depth than ever before.
The Population Synthesizer module in TRANSIMS uses Census data to build synthetic households for the study area, and also uses land use data to locate the households relative to the transportation network. The output of the Population Synthesizer module is the synthetic households with a set of information associated with each household and each individual living in that household. It also provides the household location in the TRANSIMS network including the information on vehicles belonging to each household. The data input and output for the Population Synthesizer can be summarized as shown in the figure below.
Input/Output flow for Population Synthesizer data
Synthetic households are usually generated for a block group within a census tract, and each synthetic household is classified either as family, non-family or individuals living in group quarters such as dormitories. Each individual has an associated set of demographics which may consist of age, income, status, etc. These demographics are matched closely to the demographics of the real household. Using the census data of STF-3A and PUMS, the population synthesizer can estimate a proportion of real households for each block group in each demographic category using iterative proportional fitting (IPF). Synthetic households are drawn from the PUMS data for each block group according to these resulting proportions. TRANSIMS not only creates synthetic households but also matches the demographics between synthetic households and real households. This includes the household structure, the individual incomes and ages, those individuals who work and those who attend school, as shown below.
Final IPF Table for example Block Group
|Householder’s Age (HHAGE)|
Corresponding proportions of family households for example Block Group
|Householder’s Age (HHAGE)|
The PUMS data for the example Block Group 1
|Householder’s Age (HHAGE)|
Use weighted sampling to get 63 households from the group of 56 in PUMS bin.
Example Household sampled from the PUMS cross class bin
|Status||Head of Household||Spouse||Child|
|Work Status||Worker||Worker Hence:||Student|
|Household Weight, wp||201|
After the synthetic households are created, they are distributed spatially to approximate regional population distribution. The Population Synthesizer associates these households to activity locations on the walk link of the TRANSIMS network according to land-use characteristics associated with the activity locations on that link, as shown in the figure below.
The Population Synthesizer creating synthetic households and placing them on the network
Then vehicle ownership is generated for each synthetic household as given by the PUMS data. All the outputs obtained from the Population Synthesizer are used as inputs for the Activity Generator module.
TRANSIMS generates a list of activities for each individual in a synthetic household by using the Activity Generator module. These activities are based on demographic surveys and activity surveys collected from real households in the study area. The demographic survey contains information about characteristics of each individual household in the survey sample of real households used to match the demographics characteristic of the synthetic households obtained from the census PUMS data. The activity survey obtained from the sample of real households includes travel and event-participation information for each individual household member over a period of one or more days. The household activity survey, the synthetic household obtained from the Population Synthesizer module and the network data constitute the input list to the Activity Generator.
Input/Output data flow for Activity Generator
The assignment of activities from survey households to synthesized households is done based on household demographic characteristics. The demographics of synthetic households must match the demographics of the survey households. A classification and regression tree algorithm (CART) is used to group the survey households having similar activity time patterns according to these demographic characteristics.
Example CART tree using Household Demographics
The end nodes in this tree represent survey households that have similar activity patterns based on the classified demographic characteristics. The matching is done by selecting a survey household in the end node and giving its 24-hour activities to a synthetic household that has the same demographic path to that end node.
The activity matching is done for each individual member of the synthetic household based on age, gender, and relation. Each activity assigned to each individual has as associated activity type (i.e. work, shopping, school, etc), duration, mode preference, beginning time, and ending time. The figure below shows an example of the activity list for an individual synthetic household.
Example of activity list of a synthetic household in TRANSIMS
All activities in TRANSIMS separated by time and location require travel between them. Therefore, a travel mode to each activity is assigned. However, the Route Planner searches for the best of all possible modes to execute the travel between two activities. To locate the non-home activities, TRANSIMS use a model that considers the zonal attractiveness value, the travel times between activities and the intensity of activities within the zone.
The Route Planner module in TRANSIMS produces route plans for every individual according to the activity list generated from Activity Generator. Moreover, the Route Planner also selects the shortest-time path in the network for each individual trip. In addition to the activity list from the Activity Generator, the inputs to the Route Planner module include TRANSIMS network (Transit data and Network data), the vehicle file and the link travel times as feedback from the microsimulator as shown below.
Input/Output flow data of Route Planner module
TRANSIMS assumes that people choose their routes by selecting the shortest time path, given an upper time limit constraint from the activity survey data. TRANSIMS uses a unique algorithm, label-constrained, time-dependent shortest path-finding, in order to select routes for each trip plan of an individual member of a synthetic household.
For example, a trip plan may consist of a trip from home to work, from work to shopping, and then from shopping to home. Each trip plan is a set of trips that represent each individual’s movement between his/her desired activities through the network. Each trip may consist of several legs that are composed of nodes and links which are traversed with a single travel mode. Each leg starts and ends at an activity location, parking location, or transit stop.
A trip from office to home consisting of several legs
In order to find the route for each traveler, the Route Planner transforms the TRANSIMS network to an Internal Planner network for routing purpose. The TRANSIMS network provides information about streets, intersections, signals, parking, activity locations, and transit modes within a road transportation network. Route Planner use this information to construct the Internal Planner Network which consists of nodes, link, travel time on each link, and the possible travel mode on each link. The Route Planner views the network as a set of layers which each belong to each mode as is shown below.
Internal Route Planner network representation
The example below depicts the TRANSIMS network representation of two streets with a bus stop and an activity location on each street. The transit layers can be split into many different transportation layers such as bus route 1 layer, bus route 2 layer, rail 1, etc. The first figure shows a bi-directional street link in the TRANSIMS network between street nodes 1 and 2 and nodes 3 and 4. The second figure- shows the constructed Internal Network for this example.
There are two bus routes connecting the bus stops. Note that each bus stop splits into three nodes. For example, a bus stop BS1 splits into a node for the bus shelter for passengers (S1), a node for the bus-stop place for the bus route 1 (BS1R1), and a node for the bus-parking place for the bus route 2 (BS1R2).
TRANSIMS network representation of two streets with a bus stop and an activity location on each street. There are two bus routes connecting the bus stops.
The Corresponding Internal Network representation
The Corresponding Layers of the Internal Network
There are five different layers in this Internal Network. All activity locations are always placed on the walk layer, while the intersection nodes and parking location are placed on the street layer. The bus layers contain the bus stations, and the two bus route layers. Conceptually, nodes 1, 2, 3, and 4 appear in two different layers, the walk layer and the street layer, even though these appearances correspond to the same nodes in the TRANSIMS network. Generally, each activity location is attached to a corresponding parking location and/or transit stop. Each parking or transit stop must be explicitly connected to appropriate activity locations in the walk-network using process links.
The Route Planner uses trip information to build the network as follows. The first leg of the trip is the walking leg in which the Route Planner searches for possible paths within the walking layer of the network to obtain a walking route from the walking location to the parking location of the individual’s vehicle. When such a path is found, a series of least-cost driving links in the street layer are found to obtain a route to a parking location near his house. A walk route is then developed to move the traveler from the parking lot to his home.
Trips that cannot meet each individual’s goal are fed back to the activity list to choose a new activity time, location or mode for travel to the activity. When all trip plans are done, TRANSIMS passes this information to the Traffic Microsimulator in order to execute these plans.
The Traffic Microsimulator module in TRANSIMS executes travel plans and computes the overall intra- and inter-modal transportation system dynamics. The Traffic Microsimulator is updated every second to ensure that dynamic vehicle behaviors are captured with enough fidelity to generate realistic overall traffic behavior. Interactions of travelers produce emergent traffic behaviors, such as congestion, which consequently are used to compute vehicle emissions. The input and output data flow of the microsimulation is shown in the figure below.
Input/Output flow data of Traffic Microsimulator module
The input to the Traffic Microsimulator module consists of TRANSIMS network, vehicle file, and traveler plans obtained from the Route Planner. Each individual moves from one activity to another according to the plan obtained from the Route Planner, using combinations of modes such as walking, driving or riding in a vehicle. All vehicle movements are simulated in detail to include driving on roads, stopping for signals, accelerating, decelerating, changing lanes, stopping to pick up passengers, etc.
Vehicles follow a set of rules that guarantee that no vehicle collisions will occur. This movement is accomplished by using a cellular automata principle. Each section of roadway is divided into cells as shown in the figure below.
Cellular Automaton Microsimulation
Each cell either contains a vehicle or is empty. This simulation is carried out in discrete timesteps. For each second, the vehicle decides whether to accelerate, brake, or change lanes in response to the nearby vehicles in the grid. The simulation guarantees that each vehicle makes decisions based on the state of every other vehicle in its surrounding at the same time.
There are three major types of output from the Traffic Microsimulator:
Traveler Event Output Data — which report almost everything that happens to a traveler.
Summary Data — which consists of spatial and temporal data. Spatial summaries include data aggregated over user-defined sections of roadway along the street networks, for example, densities and total flow in a 150-meter section. Temporal summaries include data about travel times along streets at various times of day.
Snapshot Files — Traffic animation can be produced from the snapshot files, which contain time, position, and velocity information for each vehicle in the simulation.
|Different From 4-Step Models||Similar to 4-Step Models|
|Framework||The microsimulation and travel behavior sub models in TRANSIMS share data easily. Model iteration works with individual travelers rather than aggregate demand data. Time is a continuous variable throughout the framework.||TRANSIMS provides a collection of software modules and utilities to build models of travel demand and transport networks.|
|Travel demand generation||TRANSIMS models and tracks travel for each individual and vehicle. The TRANSIMS activity model uses a statistical sampling method to draw an activity pattern for a modeled household from a demographically similar survey household.||TRANSIMS may estimate demand from a zonal rate-based model.|
|Destination choice||Destination choice for each trip is made for each traveler rather than aggregate trips by purpose by zone.||Zonal interchanges can be estimated with a user defined gravity model. Choice of intermediate destinations on a tour may also be estimated.|
|Mode choice||TRANSIMS applies a user defined mode preference for each traveler. All travel starts and ends with a walk segment and transfers are explicitly routed as walks. For a given trip, the TRANSIMS Router may find that walking is a better choice than a vehicle and use walk for that particular trip.||Modes are user defined. The mode preference model structure is user defined. Applications to date use zone-zone auto and transit impedance with terminal characteristics as preference variables. The Router makes the final walk vs. vehicle choice. Walking, bicycling, SOV, auto passenger, transit bus, school bus, fixed guideway transit, and park-and-ride have been modeled to date.|
|Route choice and simulation||TRANSIMS uses a routing network representing all available transport services with facility service quality aggregated to fifteen-minute intervals. Path finding is link based. The Router relies on the Microsimulator to provide expected link travel times. The Microsimulator is an integral part of the TRANSIMS tool set.||TRANSIMS finds the shortest time path between two points on a network. Cost variables can be incorporated using a value of time for each traveler. All-or-nothing, iterative capacity-restrained, and (Nash) equilibrium assignment methods can be replicated.|
|Different From 4-Step Models||Similar to 4-Step Models|
|Time||Traveler itineraries and subsequent vehicle uses are scheduled according to a travel survey and refined based on simulation results. Demand is estimated to the minute and simulated to the second.||Trip time of departure can be estimated using diurnal factors. Results can be aggregated to the hour, period or entire day.|
Stop signs, traffic signalization, and intersection lane geometries need to be coded.
Land Use, Population and Employment data
Monitoring and Surveys
A link-node format is used to describe topology. Existing network data or a commercial GIS dataset can provide the basis for an initial highway network. Existing headway and route data may be used to generate initial transit files, schedules and run details. The network database needs to be appropriately corrected and calibrated.
Land Use, Population and Employment data
Monitoring and Surveys
|Computing Environment||A Linux server or Linux cluster is used for model application. The framework is scalable to allow more computers to be used on larger, more complex problems.||Model results are analyzed using Windows or Unix workstations.|