Activity-Based Modeling System for Travel Demand Forecasting
Chapter 1: Introduction
1.1 Background
Over the past couple of decades, the emphasis of transportation planning has shifted from the construction of new infrastructure to the effective management of travel demand. This shift has been brought about by rising social, environmental, and economic concerns coupled with a realization that building one's way out of congestion is only a temporary solution to serving the increasingly complex patterns of travel demand that evolve over time. Federal legislative acts such as the Clean Air Act Amendments, 1990 and the Intermodal Surface Transportation Efficiency Act, 1991, serve as key examples of this shift in transportation planning emphasis.
In this regard, the decade of the 1980s saw an increased interest in the development and implementation of Travel Demand Management (TDM) strategies. These strategies were aimed at effectively managing and distributing travel demand, both in the spatial and temporal dimensions. For example, flexible work hours helped shift commute related peak-period trips to off-peak periods. However, these strategies alone were not able to alleviate air quality, traffic congestion, noise, and safety problems associated with an ever-rising travel demand. As a result, new strategies termed Transportation Control Measures (TCMs) have been embraced by the transportation planning community. These measures are sophisticated and complex in nature, the exact impacts of which are unknown. However, they are not only intended to effectively manage existing travel demand, but also to reduce travel demand through the suppression and selective elimination of trips. Specifically, these measures tend to target peak-period commute trips and single-occupant vehicle (SOV) automobile trips, the two types of trips that contribute most to traffic congestion, fuel consumption, and emissions.
As increasing numbers of urban areas began considering TCMs, it became apparent that traditional travel demand forecasting and planning methods, that are primarily derived from trip-based four-step procedures, are not able to address the complex questions raised by TCM implementation. Relationships among human travel behavior patterns and the attitudes, values, and constraints that determine these patterns are extremely complex in nature, and traditional forecasting methods do not explicitly model these relationships in a theoretically sound framework.
An alternative approach which has the potential of offering effective and practical tools for TDM and TDM analysis is the activity-based approach. It was conceived in the travel behavior research arena in 1970s. Activity-based approaches explicitly recognize that travel demand is derived from the need to pursue activities that are dispersed in time and space. Moreover, these approaches recognize the inter-dependence among decisions for a series of trips made by an individual. They also recognize the interactions among various members of the household, that arise when household members allocate resources (such as household vehicles) to themselves, assign and share tasks, and jointly engage in activities. As such, it has been argued that activity-based approaches provide a theoretically and conceptually stronger framework within which travel demand modeling may be performed.
Because activity-based approaches attempt to treat travel behavior in more rigorous and realistic manners, they tend to focus on details and demand more data. Furthermore, activity-based approaches have been more of a conceptual framework than specific methods that are accompanied with quantitative tools. In fact, applications of activity-based approaches to travel demand forecasting or quantitative policy analysis are practically non-existent. Activity-based approaches are by no means a "proven" concept.
This study is probably the first attempt to develop and implement a full-fledged activity-based policy analysis tool for a metropolitan region and thereby examine whether activity-based approaches can be put to practical use. In particular, the study attempts to determine whether an operational activity-based tool can be developed while utilizing available data, supplemented by a medium-scale survey that can be conducted with modest mounts of monetary and time resources.
Although results of this study indicate that activity-based approaches in fact lead to viable policy tools, the experimental nature of this study must be born in mind by the reader of this report. It is also noted that it is not the intent of the report to assert in any way that activity-based approaches are the only approaches to travel demand forecasting and policy analysis. To the contrary, it is believed that non single approach or model system is suited for all study objectives; activity-based approaches are believed to be effective in the types of analysis contained in this report, while other approaches, including the trip-based, four step model systems, will continue to be useful tools in other types of analysis.
1.2 Study Objectives
The Metropolitan Washington Council of Governments (MWCOG) as part of the Travel Model Improvement Program (TMIP), jointly sponsored by the U.S. Department of Transportation (DOT) and the U.S. Environmental Protection Agency (EPA) engaged RDC, Inc. to conduct an applied research study to determine the feasibility of using activity-based methodologies to evaluate selected TDM policies. To perform this study using large-scale regional data, RDC, Inc., implemented a prototype of its Activity-Mobility Simulator (AMOS) which is a dynamic micro-simulator that replicates household responses to TDM measures.
To implement and test AMOS in the Washington, D.C. metropolitan area, RDC's approach consisted of the following activities:
The TDM measures to be tested within the activity-based framework of AMOS were selected in collaboration with MWCOG and Federal sponsors. Of the more than 50 identified individual and combined TDM measures, six were selected for evaluation ranging from targeted premium charges for using personal vehicles (e.g., congestion pricing) to incentives for using alternatives to personal vehicles (e.g., improved pedestrian facilities). Appendix A describes the initial set of TDM measures identified, and the process used in selecting the TDM measures addressed in the study.
In collaboration with MWCOG, RDC administered an elaborate survey of over 650 commuters in the metropolitan area designed to collect stated-preference responses to the selected TDM measures, revealed by daily time-use (activity) patterns both inside and outside the home, daily travel patterns, detailed commute trip attributes, and demographic and socio-economic data. This AMOS survey was the basis for estimating AMOS model parameters essential in evaluating TDM responses in the Washington, D.C. metropolitan area.
The AMOS prototype system was configured to maximize the use of existing pertinent data available within the MWCOG jurisdiction. MWCOG's data bases including the MWCOG 1994 Household Travel Survey data (trip diary data) and relevant network data provided baseline travel patterns for the Washington, D.C. metropolitan area.
The AMOS prototype system was tested and used to assess the selected TCMs in the Washington, D.C. metropolitan area. MWCOG provided the necessary sample data on nearly 100 households located in the study area to evaluate the commuter responsiveness to the selected TCMs.
1.3 AMOS Features
Over the past two years, the RDC, Inc., research team has developed and implemented the AMOS prototype intended to serve as a short-term transportation planning and policy analysis tool. AMOS is an activity-based micro-simulator of daily human activity and travel patterns, which focuses on the adaptation and learning process that people exhibit when faced with a change in the transportation environment. AMOS simulates a new activity-travel pattern that a person is likely to adopt in response to a TDM measure. This is accomplished through the implementation of several AMOS modules, namely:
Baseline Activity-Travel Analyzer. The baseline activity-travel analyzer reads individual trip records, compares them with the network data for logical consistency and missing information, and then generates a coherent baseline activity-travel pattern for each individual. All consistent baseline activity-travel patterns are used by the remaining AMOS system components.
TDM Response Option Generator. This module creates the "basic" response of an individual to a TDM strategy. It is a neural network model that is trained by using revealed-preference and stated-preference data obtained from AMOS survey. The baseline travel pattern from the Baseline Activity-Travel Analyzer, demographic and socio-economic attributes, and TDM characteristics under investigation serve as inputs to this module. The outputs of this module are the behavioral responses. The TDM measures are characterized by their cost changes, travel time changes, mode attribute changes, and imposition or relaxation of constraints.
Activity-Travel Pattern Modifier. This module constitutes the activity-trip re-sequencing and re-scheduling algorithm. It provides one or more alternative activity-travel patterns based on the response provided by the TDM Response Option Generator. The inputs of this module include the baseline activity-travel patterns, network data, land-use data, socio-economic and demographic characteristics, and the response options from the TDM Response Option Generator. The output of this module is a modified activity-travel pattern. The feasibility of a modified activity-travel pattern is checked for consistency and logic against a set of rule-based constraints.
Evaluation Module and Acceptance Routines. This component evaluates the utility associated with a modified activity-travel pattern generated by the Activity-Travel Pattern Modifier. Operationally, its built-in acceptance routines assess whether a modified activity-travel pattern will be accepted or rejected on the basis of a human adaptation and learning model incorporating a set of search termination rules.
Statistics Accumulator. This module reads all feasible accepted activity-travel patterns provided by the Evaluation Module and generates descriptive and frequency statistics on a daily basis. These descriptive and frequency statistics include vehicle miles traveled, number of trips by mode and by time of day, number of stops by purpose, trip chains, activity duration by purpose, travel times by purpose, vehicle occupancy, cold and hot starts, etc. In conjunction with baseline travel patterns, it can provide measures of change in travel characteristics.
As such, AMOS consists of a series of inter-related components that collectively serve as a comprehensive transportation planning and policy analysis tool. AMOS abandons some of the questionable assumptions in the trip-based four-step procedures, and embraces several new concepts that are theoretically sound and lead to more robust TCM impact predictions.
1.4 Study Conclusions
This project represents the first implementation of a full-fledged activity-based model system for transportation planning and policy analysis. Despite the theoretical arguments that warrant their practical applications, activity-based approaches remained within the domain of academia for nearly two decades. The development of AMOS and its implementation in the Washington, D.C., metropolitan area, therefore, represents a significant step forward in transportation planning and policy analysis. The development is especially significant considering the importance of travel demand management in the current planning contexts set forth by the Clean Air Act Amendments and Intermodal Surface Transportation Efficiency Act.
In the project, a micro-simulation model system which is capable of producing travel demand forecasts based on principles of activity-based analysis has been constructed and implemented in the Washington, D.C., metropolitan area, and applied to a selection of TDM measures using a sample of trip diaries from the 1994 MWCOG survey. The achievements of this effort can be summarized as follows.
The project has demonstrated that the activity-based model system can be implemented in a metropolitan area using data available from a typical metropolitan planning organization (MPO), such as trip diary data, network travel time data, and land-use inventory data (the only additional data needed for AMOS implementation are stated-preference survey results from the area which are used to customize a component of AMOS to the area residents' responsiveness to TDM measures).
It has been shown that travel demand forecasts can be developed while treating the daily travel pattern in its entirety, without breaking it into individual trips and thereby compromising the interdependencies and continuities that exist across the series of trips made by a traveler.
This also implies that practical capabilities have been developed to assess TDM impacts more cohesively while accounting for secondary and tertiary changes in a traveler's daily travel pattern that are brought about as results of a primary change in response to a TDM measure (for example, if a SOV (single-occupant vehicle) commuter, who stops on the way to and from work to drop off and pick up a child at day-care, switches to carpooling in response to congestion pricing (primary change), then new, two round-trip SOV trips may be made between the home and day-care to drop off and pick up the child).
The AMOS survey designed in this project has shown that the stated-preference questions developed in this project have produced credible results (except for the case of a particular synergy combination of two TDM measures), and that the survey can be applied to obtain information vital for the assessment of potential effectiveness of alternative TDM measures.
The AMOS survey data produced rich statistical results that have revealed the characteristics of responses commuters would show when faced with TDM measures; for example, female commuters who make stops on the way to or from work tend not to change their travel in response to a TDM measure.
The numerical examples using the sample of MWCOG trip diary data have shown the AMOS prototype is capable of producing aggregate statistics of travel demand at levels that are comparable to the conventional trip-based model systems (except that the current version of AMOS operates with static zone-to-zone travel time matrices rather than internally conducting network assignment).
It is worthy to note that the development of the AMOS prototype incorporates a number of theoretical concepts, such as "adaptation behavior" and "time-space constraints," into a practical model system which fully utilizes the data that are maintained by a typical MPO.
1.5 Outline of Report
This report consists of eight more sections. Sections 2 and 3 discuss the trip-based four step process, and the features that can be either augmented or replaced by an activity-based travel demand methodology such as AMOS. Sections 4 and 5 discuss the basic concepts and analytical techniques which are the foundation of AMOS, and its applicability in evaluating TDM policies. Section 6 defines the TDM policies selected for evaluation in the Washington, D.C. metropolitan area, implementation of AMOS with the MWCOG network data, and the application of AMOS to MWCOG household records. Sections 7 and 8 discuss the results of the TDM policy analysis, and implications for future activity-based travel demand modeling.
