MASTERS THESIS # CODE PARTY # APP STORE APPS
.: M. Sc. Research & Thesis :.
Current Computer Science Master's Research @ WSU (2009), titled:
"L-CAPS: Integrating Cultural Algorithms into a 3D
Virtual World"
ABSTRACT:
Cultural algorithms (CA) provide a
vehicle for modeling social evolution and learning. It does
so by evaluating generations of agent populations within an
environment suitable for learning over a temporal span. The
agent Population Space maintains a Belief Space. The
current generation contributes to the Belief Space, and
future generations are influenced by it. The Belief Space
represents the culture of the population, and is formed
through information in five different knowledge sources.
This paper presents the “ Land Bridge – Cultural Algorithms
Program Simulation”
(L-CAPS), a system utilizing a population of program
instances running in the Cultural Algorithm paradigm. Each
program instance is considered an agent within the CA
paradigm, and each program instance executes a
self-contained virtual world initialized with normative
parameters maintained in the population's Belief Space. The
virtual world belonging to each resident program instance
simulates the same real world location: the Lake Stanley
Land Bridge, an underwater area anthropologically surveyed
for the first time. The L-CAPS design provides an interface
to evaluate the fitness over the lifetime of any program
instance belonging to a CA. The author suggests the L-CAPS
framework represents a crucial first step to theorizing
formal methodologies for using CA for more general game
designs.
Cultural algorithms operate within a segment of artificial
intelligence known as Evolutionary Computation (EC).
Research in applying EC to game agents has seen a
significant increase in the last two decades. EC research
efforts contribute to understanding ways game agents can
evolve, learn and adapt to the presence of hidden
information in games. The evolving agents in our game are
program instances, grouped into populations, over
successive generations. The CA component used by the L-CAPS
framework represents an EC design mechanism not yet seen
within a modern game simulation.
