Practical Innovations and Experience in ABM Equilibration with Network Assignments
Corresponding Author: Chrissy Bernardo, WSP
Presented By: Chrissy Bernardo, WSP
Equilibration of transportation demand and network supply is an essential component of any travel demand model. The equilibration process both ensures that travel demand model become policy sensitive and also that a unique objective solution is modeled for each scenario. With a traditional 4-step model setup, several feedback iterations of the demand-generation components and network procedures may be able to be run in a relatively short time frame, and many different projects or scenarios may be evaluated under fully equilibrated conditions. As transportation policy questions grow in complexity and dependency on individual characteristics, Activity Based Models (ABMs) are increasingly being used by a significant and growing number of planning agencies. With the growth in the use of ABMs, the need for efficient and robust equilibration procedures becomes increasingly evident. However, equilibration of an ABM with network assignments is in general more complex than equilibration of 4-step models, due to the individual microsimulation structure of an ABM.
The authors have recently developed an advanced ABM system for three cities in Ohio (Columbus, Cincinnati-Dayton, and Cleveland-Akron) and explored several strategies to increase the efficiency of the equilibration process and reduce total runtimes. These strategies include: various forms of “warm starts” by ramping up ABM demand over feedback iterations through sampling of individual households, selective “freezing” of certain ABM sub-models (such as long- and mid-term sub-models) in feedback iterations, averaging flows and/or skims over iterations, selectively running transit procedures, and many others. One of the important conclusions of this effort is that the equilibration schema should be flexible and adjusted to each particular project or policy as well as to the nature of compared transportation or land-use alternatives.
The presentation will discuss the results of utilizing each of these options, including runtime savings, convergence statistics, and impact on model outputs. These strategies help to make the ABM a stronger tool in the planner’s arsenal, allowing for quicker global convergence without compromising model sensitivity and objectiveness of the results.