Geoscience Reference
In-Depth Information
My view is that AB modelling changes the view on
what is a model
. The tradition of physical
science demands establishing a model's structure by means of equations or a computer program and
then studying the model's dynamics as dependent on parameters. In terms of physical science, dif-
ferent sets of behavioural rules of the Schelling residential agents define different models. Each of
these models has its own set of parameters, and until they are fully investigated, it is impossible to
know whether these models generate similar or different dynamics. However, from the geographic
point of view, different formalisations of the Schelling model presented in Section 9.3.4 express the
same idea of
staying among friends
. In this respect, all our examples are different versions of the
Schelling model. The situation may be compared to the models of species interaction in popula-
tion ecology. Just as in the Lotka-Volterra predator-prey model (Benenson and Torrens, 2004,
Chapter 3), the original assumptions of the Schelling model serve as a starting point and there are
numerous ways to formulate the model's assumptions and demonstrate non-trivial consequences of
the basic idea. How productive is this view? A search for the
predator-prey model
within scirus.
com, conducted on 7 January 2013, resulted in 13,268 hits, of which 3,585 are research papers,
while the search for the
Schelling model
is far behind - 842 hits and 69 papers. To me, this just
demonstrates the amount of potential work for future students of GC!
An advantage of AB modelling is the possibility of representing the behaviour of householders
explicitly; the latter cannot be done with models that are formulated in terms of aggregate param-
eters, for example, with the total numbers of households in different ethnic groups for the Schelling
model. This is critically advantageous for GC since AB modelling breaks the conceptual barrier
between informal and formal geographic knowledge. As far as a new view of possible household
behaviour is established, a new AB model can be constructed or even better, an existing model
can be modified to study whether this new view brings new consequences. I would thus conclude
optimistically: the Schelling model is a forerunner in the development of abstract AB modelling as
a long-awaited framework for quantitative social geography.
The direct interpretation of human and institutional behaviour makes AB modelling especially
attractive for applied simulation. I illustrate this point with the PARKAGENT model, which was
initially constructed as a tool for spatially explicit policymaking. I have exploited the PARKAGENT
model for presenting the basic ideas of geosimulation in Sections 9.2.1 and 9.2.2 of this chapter,
and in what follows, I exploit it for illustrating the applied abilities of AB modelling. As should be
expected, the presentation of a real-world application of an AB model demands essentially more
details than the abstract Schelling model.
9.4 EXAMPLE OF PREDICTIVE AB MODELLING: THE PARKAGENT
MODEL OF PARKING SEARCH AND PARKING IN THE CITY
Citizens of every large city in the world suffer from a lack of parking spaces. Municipalities are
aware of numerous ways to improve the situation, for example, restrict car entrance to the city
centre, increase parking fees and fines, forbid parking along the main streets and restrain parking
within residential areas. However, drivers comprise the majority of the voters and thus municipali-
ties implement parking policies with great care. A tool that would serve to assess the consequences
of parking policies would really help the decision maker. PARKAGENT, an AB model of parking
search and parking in the city, was developed for this purpose. Let us present the features of the
PARKAGENT model according to Benenson et al. (2008) and Levy et al. (2012).
9.4.1 a
gentS
, o
BjectS
and
r
elationShiPS
PARKAGENT model agents and objects were already introduced in Section 9.2.1 and used as an
example of AB model development. These were
driver agents
, and objects of different kinds -
destinations
,
parking places
, and
the
street network
. The full version of the PARKAGENT model
