Transportation Modeling with TransCAD

ECIV 340L, CEE, UofSC true
11-19-2021

RELATED RESOUCES:

INTRODUCTION

The purpose of this lab session is to perform travel demand modeling and logistics analysis using TransCAD. First, you need to download the demo files (TransCAD_demo.zip) from Blackboard. Then, open the TransCAD 8.0 (Academic 64-bit).

Part I: Travel Demand Modeling

Trip Generation (Production)

The goal of trip generation is to predict the number of trips that are generated by and attracted to each zone in a study area. This is only concerned with the number of trips that start and end in each zone, and not with making connections between origins and destinations of trips.

There are three primary tools that are used in modeling trip production:

Cross-Classification:

Cross-classification methods separate the population in an urban area into relatively homogeneous groups based on certain socio-economic characteristics. Then, average trip production rates per household or individual are empirically estimated for each classification. This creates a lookup table that may be used to forecast trip productions.

Regression Models:

Two types of regressions are commonly used. The first uses data aggregated at the zonal level, with average number of trips per household in the zone as the dependent variable and average zonal characteristics as the explanatory variables. The second uses disaggregated data at the household or individual level, with the number of trips made by a household or individual as the dependent variable and the household and personal characteristics as the explanatory variables.

Discrete Choice Models:

Discrete choice models use disaggregate household or individual level data to estimate the probability with which any household or individual will make trips. The outcome can then be aggregated to predict the number of trips produced.

Task 1: Generating Productions by Cross-Classification

Task 2: Estimating and Applying a Zonal Regression Model

Task 3: Estimating a Logit Model of Work Trip Production

Trip Generation (Attraction)

In many ways, estimating trip attractions is similar to estimating trip productions because the problem is the same: predicting the number of trips attracted by relating the number or frequency of trips to the characteristics of the individuals, the zone, and the transportation network. Thus, the three methods described above may also be used to estimate the number of trips attracted to a zone.

Trip Distribution

Trip distribution models are used to predict the destination choices of trip makers. Usually, in trip distribution a new flow matrix is forecasted based on estimates of future productions and attractions and estimates of the costs of travel between origins and destinations.

Trip distribution models can be applied at either an aggregate or a disaggregate level. Aggregate trip distribution models are typically used to predict flows between origin and destination zones. Three categories of aggregate trip distribution methods predominate in urban transportation planning:

Growth Factor Methods:

These involve scaling an existing matrix by applying multiplicative factors to matrix cells. These methods are usually encountered when there is no information available concerning the network interzonal distances, travel times, or generalized costs.

Gravity Models:

The typical inputs include one or more flow matrices, an impedance matrix reflecting the distance, time, or cost of travel between zones, and estimates of future levels productions and attractions. The gravity model explicitly relates flows between zones to interzonal impedance to travel.

Destination Choice Models:

These generally take the form of discrete choice models that evaluate utility functions for various potential destinations and ascribe probabilities to the same. The utilities can be a function of destination zone attributes as well as origin-destination skims.

Task 4. Applying a Growth Factor Model

Task 5: Applying a Gravity Model

Mode Choice

Mode choice models are used to analyze and predict the choices that individuals or groups of individuals make in selecting the transportation modes that are used for particular types of trips. Typically, the goal is to predict the aggregate share or absolute number of trips made by mode for each origin-destination pair. Mode choice models may also be applied on disaggregate lists of individuals and are often employed in Activity-Based Models. The modeling concepts underlying mode choice models may also be applied to predict other choices and shares such as household auto ownership splits, time of day choice, etc. The mode choice component of a travel demand model is estimated using individual-level data obtained from a survey, and forecasts are subsequently based upon aggregate, zonal segment level explanatory variables.

Discrete Choice Models

Discrete choice models are in many respects a substitute for regression models when the dependent variable is qualitative or categorical rather than continuous. Discrete choice models are formulated as stochastic models, in which the probability that a particular response is observed is a function of a set of explanatory variables. There are a variety of functional forms that can be proposed for the explanation of discrete choice. The Multinomial Logit (MNL) and Nested Logit (NL) models are used extensively.

Multinomial Logit (MNL) Model

MNL involves a decision-maker choosing exactly one alternative from a set of available discrete alternatives. For example, the figure below illustrated a mode choice situation in which individuals choose one of Drive Alone, Carpool or Bus to make certain type(s) of trips:

Task 6: Estimating an MNL Model

Task 7: Applying an MNL Model on a Matrix

Nested Logit (NL) Model

When dealing with choice models that include several alternatives, there often exist natural groupings of alternatives and/or a natural hierarchy to the decision being made. A nested logit model is often represented using a tree structure, which visually represents groupings or hierarchies of the choice alternatives. For example, in mode choice, the tree may be:

In this case, the decision process can be thought of as occurring at the two levels. First, the traveler decides whether to take transit or auto. If transit is chosen, the traveler further decides whether to take the bus or light rail. Similarly, if auto is chosen, the traveler decides whether to drive alone or carpool. However, note that it is in fact a join decision, in that whether a traveler chooses transit or auto depend on the characteristics of the lower level alternatives. The nested logit model provides a mathematical representation of this joint decision.

Task 8: Estimating a NL Model

Task 9: Applying a NL Model on a Matrix

Traffic Assignment

Traffic assignment models are used to estimate the traffic flows on a network. These models take as input a matrix of flows that indicate the volume of traffic between origin and destination (O-D) pairs. They also take input on the network topology, link characteristics, and link performance functions. The flows for each O-D pair are loaded onto the network based on the travel time or impedance of the alternative paths that could carry this traffic. TransCAD provides the widest array of traffic assignment procedures that can be used for modeling urban traffic. These procedures include numerous variants that can be used for modeling intercity passenger and freight traffic.

Traffic assignment is a key element in the urban travel demand forecasting process. The traffic assignment model predicts the network flows that are associated with future planning scenarios and generates estimates of the link travel times and related attributes that are the basis for benefits estimation and air quality impacts. The traffic assignment model is also used to generate the estimates of network performance that are used in the mode choice and trip distribution or destination choice stages of many models.

User Equilibrium (UE)

Assignment UE assignment is the recommended method for traffic assignment unless a more advanced model is employed. UE procedures use an iterative process to achieve a convergent solution, in which no travelers can improve their travel times by shifting routes. In each iteration, TransCAD computes network link flows that are based upon flow-dependent travel times. The formulation of the UE problem is a mathematical program and the Frank-Wolfe (FW) solution method is one of several employed in TransCAD. Another way to compute a UE assignment is to use an alternative algorithm such as path-based algorithm.

System Optimum (SO)

Assignment SO computes an assignment that minimizes total travel time on the network. Under SO assignment, no users can change routes without increasing the total travel time on the system, although it is possible that travelers could reduce their own travel times. SO assignment can be thought of as a model in which total system cost is minimized when travelers are told which routes to use. Obviously not a behaviorally realistic model, SO assignment can be useful in analyzing Intelligent Transportation System (ITS) scenarios.

Task 10: Performing a User Equilibrium Traffic Assignment

Multi-Modal Multi-Class Assignment (MMA)

MMA is a flexible master assignment routine designed for use in major metropolitan areas and is directly applicable in statewide or interregional models. The MMA model is a generalized cost assignment that lets you assign trips by individual modes or user classes to the network simultaneously. This method allows you to explicitly model the influence of toll facilities of all types, as well as HOV and HOT facilities. Each mode or class can have different network exclusions, congestion impacts (passenger car equivalent values), values of time, and toll costs.

Task 12: Performing Multi-Modal Assignment

Part II: Network and Route Analysis

Vehicle Routing Many businesses and government agencies transport goods from one or more central locations to a set of destinations. It is important to manage these operations efficiently, both to reduce operating costs and to ensure that pickups and deliveries adhere to reasonable service standards. This general problem is known as the vehicle routing problem. Solving the vehicle routing problem involves determining how many vehicles are required to service the destinations and developing a route and schedule for each one. Because there are many variations of the problem, it can be very difficult to solve. TransCAD provides a rich set of vehicle routing tools that solve various types of routing problems.

Task 13: Solving a Vehicle Routing Problem

Network Flow Models

TransCAD includes a set of procedures for solving network flow problems. These problems involve efficient delivery of goods or services and arise both in transportation and in many other contexts. For example: - You need to ship an inventory of goods from 15 warehouses to 100 retail centers, each with a given demand, and you want to determine which warehouses should service which retail centers to minimize the total transportation cost. - You need to route empty rail cars from their current locations to locations where they are required for new loads, taking into account rail traffic density by link.

Task 14: Solving a Transportation Problem and Displaying Output Flow

Facility Location Models

Facility Location Models are used to identify good locations for warehouses, hospitals, retail stores, manufacturing facilities, and other types of facilities. In general, the goal in locating such facilities is either to provide a high level of service, to minimize operating costs, or to maximize profits. TransCAD solves many different types of facility location problems, with applications in both the public and private sectors. Here are two examples: - You need to determine the best location for a new branch of a public library. Your goal is to provide the best overall level of access to city residents. - You need to determine the best location for a new police station. Your goal is to reduce the maximum distance a patrol car needs to travel from the station to a resident’s home.

Task 15: Solving a Facility Location Problem