Geoscience Reference
In-Depth Information
9.2.1 a
gentS
and
a
gent
P
roPertieS
r
eflect
a
M
odeller
'
S
V
iew
of
r
eality
The overall goal of the PARKAGENT model is to assess the maximum number of cars that can be
permitted to enter the city area for parking search and parking. The policy maker wants to permit
as many cars to enter as possible while minimising the need to cruise an area for a free parking
space, which essentially decreases the effectiveness of the urban transportation system. For the
purpose of this chapter, let us consider minimal possible settings, in which parking is only available
along the streets and is free of charge. The specific goal of a model is to estimate the conditions
after which a significant number of drivers would cruise the area in search of a parking space for
too long. A rule of thumb from civil engineering is that in order to guarantee a short search time
and parking that is close to the destination for the majority of drivers, the fraction of occupied
places should be kept below 85%, that is, one in seven parking places should always remain free
(Shoup, 2006). Let us test this assertion and investigate the system in the situation when higher
parking occupation rates occur.
It is always worth starting with the simplest model and then extending it in case the existing ver-
sion becomes insufficient for investigating the problem. From this point of view, the minimum set
of geographic features that are necessary for representing parking dynamics in the city consists of
drivers
,
destinations
,
parking places
and
the street network
. The street network, the destinations
and the parking places are passive geographic objects, while the drivers are agents who make deci-
sions regarding parking search and parking.
To model parking search, each driver should be assigned a destination - let us thus assume that
the driver agent knows it. The distance at which a driver agent decides to start searching for a park-
ing space can be different for different kinds of drivers. Residents of densely populated areas will
start their search well before they reach their homes based on previous negative experiences with
finding parking spaces in the past, while the first-time visitor will drive straight to their destination.
To combine these, let us assume that the driver agents start their parking search 500 m before the
destination. To search for parking, a driver should see free and occupied parking places and, for this
purpose, driver agents should be exactly located on the street. The planned duration of the parking
is another property of the driver agent.
The street network is represented in the PARKAGENT model in a standard GIS network topo-
logical format of nodes (junctions) connected by links (street segments) (Kresse et al., 2012). The
links of the street network are characterised by traffic direction - one or two way. Parking places
are represented by the points located along the street segments at a distance between each other that
equals the average car length plus an average gap between two parked cars.
Important for the model, parking places can have a common or limited use, for example, only
for use by residents in the area and can differ in price; these are two parking place attributes. The
destinations of drivers are buildings. The main property of the destination is the number of drivers
that are heading towards this location (Figure 9.2).
After all the objects, agents and properties are defined, the relationships between them should
then be specified. In geographic systems, relationships are also used for locating agents and objects
in space and the PARKAGENT model employs this approach.
9.2.2 r
elationShiPS
Between
a
gentS
and
o
BjectS
During the search for a parking place, a driver agent must know which parking places it passes, and of
these, how many are free. The real-world driver would simply see this so the driver agent must imitate
this ability with the help of a calculation. Relationships are very useful at this stage. Database theory
defines a relationship as an 'association between two or more entities' (Howe, 2001). According to
this theory, relationships and their properties are stored in tables, just as entities are.
The use of relationships is very helpful for the PARKAGENT model too. As we mentioned in
the previous section, the first set of relationships between the street segments and the junctions is
