Information Technology Reference
In-Depth Information
in ecosystems, vehicles in tra
2
c, people in crowds, or autonomous characters in
animation and games. Unlike CA-based simulation, which is based on a dense
and uniform dissection of the space where the execution control is centralized
and the simulation inner loop proceeds cell by cell [5], in MABS applications
the system simulation is based on autonomous and distributed agents (i.e. it
proceeds agent by agent each one with its own thread of control).
Recent results in complexity science [6] suggest that the topology of agent
interaction is critical to the nature of the emergent behavior of the MAS. With
the exception of some preliminary proposals [7], none of the MAS models pre-
sented in the literature explicitly takes into account the spatial structure (i.e.
topology) of the environment where agents are located. This happens despite
of the fact that a large class of problems is characterized by unavoidable spa-
tial features requiring the related modelling approach to incorporate an explicit
or implicit model of the space. In fact, several domains deal with space itself
(e.g., geographical location) or a model of it (e.g., information flow in an or-
ganizational structure) or both of them (e.g., to conquer some favorable selling
location in a region or to play a new role implying a move in the physical and
organizational structure of a company). In traditional MAS approach, agents
can be associated with a location in their environment, but no explicit structure
of the environment is given. For instance, mobile information agents that are
located on MAS models of networked computers do not refer to the network
structure as an explicitly defined geometrical space [8].
On the contrary, CAs have offered a very interesting framework to model
and simulate natural and artificial phenomena involving space, due to their ba-
sic definition and structure [9]. CA have been profitably used in various cases
of simulation where space has a crucial role. In this respect, we can distinguish
between CA models that intrinsically allow parametric spatial conditions to be
represented (e.g. fluid-dynamics in porous media, cellular geography and so on),
and CA models that allow a spatial representation to be created (e.g., in compet-
itive behavior, population dynamics, financial data clustering). CA modelling is
designed to simulate the dynamics of spatial interaction and, as a consequence,
CA have been employed in the exploration of various urban phenomena (e.g.
tra
2
c simulation, regional urbanization, land use dynamics) explicitly dealing
with the spatial relation and interaction among locations [10]. Moreover, in some
cases, CA offered the possibility to conceptualize and visualize abstract and in-
trinsically not spatial problems [11].
Interesting results have been shown by the combination of MAS and CA [12].
The approach basically consists in the positioning of a MAS on a cellular space
and has been mainly applied to analyze urban system dynamics and pedestrian
activity [13]. Two sets of example scenarios where agents have been combined
with cellular spaces to model environmental and urban systems can be found
in [10] and [14]. In these examples, the cellular space simulates the dynamics
of the urban infrastructure (e.g. land-use transition, real estate development
and redevelopment, urban growth) while a MAS simulates the dynamic of the
