Long-Term Simulator
The
long-term (LT) simulator of SimMobility models the behaviors of agents in the housing market, and
ultimately the commercial real estate market and the job market, in order to
simulate the yearly and longer term impacts of alternative future mobility
scenarios on residential and workplace locations; vehicle ownership; the
density, land use distribution, and value of the built environment. Figure 3
shows the framework of the LT simulator. In general, the LT simulator is
responsible for the generation and updating of a population of agents and their
corresponding demographic and locational attributes. In the beginning, a two-stage
data synthesis methodology is employed for construction of a synthetic
population of households and firm establishments at building scale. The
approach is designed to accommodate the need for spatially disaggregated
details in a manner that can be readily adjusted and rerun to incorporate new
data sources, changed time frames, and updated relationships and hierarchies
across overlapping datasets. Long-term behaviors of agents and their
effects on urban form, markets and other agents are implemented by a group of
behavioral models that are connected in a sequential/event-based framework.
Figure . Framework of the
Long-Term Simulator
These
behavioral models take account of demographic and economic factors of agents,
locational amenities and the regulatory variables translated from exogenously
specified policies. The LT simulator centers on a real estate market module,
which emulates the dynamic interaction process between demand and supply in the
market. The market module include a series of models that simulate (a)
‘awakening’ of households who begin searching for new housing, (b) eligibility,
affordability, and screening constraints, (c) daily housing market bidding, and
(d) modeling developer behavior regarding when, where, what type, and how
much built space to construct by taking into account market cycle and
uncertainty. Changes in residential location then trigger a household’s
re-assessment of private vehicle ownership and possible re-assignment of
workers (students) to jobs (schools).
The long-term simulator is
integrated with the MT simulator via built-in functions facilitating the
exchanges of data that characterize the status-quo of land use and
transportation performance. One set of functions computes accessibility
measures for individuals considering alternative residential, work, or school
locations, and alternative vehicle ownership conditions. These measures can be
computed quickly since they vary the circumstances of only the one individual.
Another set of functions allows the LT simulator to pass population (and firm)
information with updated residential and job locations as well as vehicle
ownership. This information is sent periodically so that the MT simulator can reassess
overall activity patterns and accessibility conditions, and the LT simulator
can then make choices based on adjusted expectations accessibility. Currently
this exchange is done annually. Information on the performance of
transportation services and activity-travel participation of agents is fed to
the land use module of SimMobility (i.e. the LT simulator) through a
utility-based, behaviorally rigorous accessibility measure, the logsum. It is the expected maximum
utility of a person in a series of activity related choice situations.
In SimMobility, the logsum measure reflects the range of choices in destinations and
modes, the scarcity of time and money, and accounts for the heterogeneous
preferences among agents. Therefore, it is a link between the MT simulator and
the LT simulator which ensures the behavioral consistency of agents by
encapsulating agents’ day-to-day activity and travel considerations into their
long-term location and vehicle ownership choices. However, because the logsum measure is individual specific
and not directly comparable across agents, it is first converted to cost
(dollars) before being aggregated in the LT simulator to model household-level
choices.