254 Agent-Based Modeling and Simulation with Swarm
Sugarsca pe, which was proposed by Joshua Epstein and Robert Axtell [32],
is a model that for ms an artificial society (see Section 7.10 for details). This
is basically a simulation where ants move ar ound looking for suga r, but its
objective is to understand mechanisms behind social behavior by taking into
account concepts such as mating , warfare, and dealing. Understanding the
environments around each agent and under what conditions agents engage
in warfare or make deals would help us understand the behavior of actual
economies. For example, Kiyoshi Izumi [62] attempted to explain ma c roscopic
phenomena such as the frequency distribution of rate fluctuation or contrary
opinion by laws of cause and effect at the microscopic level through interviews
with dea lers (see [59]). This corresponds to stage 5 research that aims to
connect emergent phenomena to the actual world.
Related research 3: Acoustic consonance and dissonance perception
model
Humans perceive two or more combined sounds as a pleasant consonance
or an unpleasant dissonance. Perception models of consonance and dissonance
are based on Helmholtz’s theory of beats where the extent of consona nce de-
pends on the closeness of harmonics [48]. T his theory was formulated math-
ematically by Kameoka and Kuriyagawa [63, 64], and its internal parameters
were determined by psychological experiments. The mechanism of the inner
ear that picks up frequency, which corresponds to the input in the model, has
been clar ified; however, the r e lation to the internal mechanism of the bra in is
still not under stood. This connection would correspond to stage 5.
Related research 4: Biolo gical speciation
Clement researched biological speciation through an investigation of the
ecology of fish [17]. Artificial life that simulated fish was created to research
what kind of clustering is effective in speciation. Metivier et al. investigated
how stress affects speciation of individuals through a simulator called Life
Drop [84]. The simulator developed showed that stress fro m the environment
increases the possibility of crossover between species, which in turn affects
sp e c iation. The conclusion was that stress strongly influences evolution. The
simulation results agreed with the results of biological experiments using bac-
teria, and hence this corresponds to stage 5 in the constructive approach.
Related research 5: Foraging of animals
There has been much research on optimization of the foraging strategy
of animals in mathematical ecology [61] and behavioral ecology [26]. For
instance, assume the nutrition values (g
i
) and the cost necessary for intake
(h
i
) are given for multiple type s o f food. The hypotheses derived from the
theory to optimize foraging are the following.
Claim 1 If every type of food exists with the same distribution, the food
with higher
g
h
has higher preference.
Conclusion 255
Claim 2 If there is an abundance of a preferred food (food with large
g
h
),
only that food should be eaten.
Claim 3 The decision to eat o r not to eat a pre ferred food does not depend
on the amount of les s-preferred fo ods.
The author used simulations, for instance to learn optimization strategies
using classifier systems, and confirmed that rules that are consistent with
these hypotheses could be o btained [56]. T here is also an investigation on
the searching pattern that results in optimized foraging (cognitive model for
efficient fo raging). Erichsen et al. experimentally verified cognitive models
using paridae [36].
Related research 6: Evolutionary robotics
Karl Sims [107] created evolved virtual crea tures, or creatures made of
directional graphs that have actions to mimic creatures (see Section 4.4.2 for
details). These c reatures may show shapes and actions that people cannot
even conceive of, and one study applied these creatures to morphogenes is of
robots [117]. A project, RobotCub, aims to crea te robots with the learning
ability of human toddlers through a constructive approach [11, 27, 39]. The
objective of this rese arch is to clarify how toddlers learn specific task s and what
is necessary in the learning pr ocess. Mitsuo Kawato created walking robots
that reproduce information on brain activity in walking monkeys to gain a
deeper understanding of the brain [68]. Cre ating ac tual models to understand
actual subjects and to make connections corresponds to stage 5. Takashima
et al. carried out research to understand the mechanisms in the brain by
connecting silkworms or their brains to robots to make silkworms intelligence
control robots [113].
Related research 7: Emergence in army ants
The research introduced in Section 5.8 aims to reproduce and understand
the cooperative behavior of ar my ants. Exper iments using simulators designed
to reproduce cooperative behavior in army ants s howed that agents built
bridges in appropriate places to ma ke shortcuts. This is an accomplishment of
stage 2. Incorporating different co llaborative behavior and hypotheses c onsid-
ering the ecology o f ants in the simulator resulted in agents initially attempt-
ing to build bridges in various positions and later concentrating on building
existing bridges located in vario us positions. This is similar to the collective
decision-making of ants in nature, and corresp onds to stage 3. However, this
simulation does not connect directly to the actual world; therefore, the current
goal is to repeat hypothesis verification to connect more closely to the results
of experiments invo lv ing animals (stage 4). The environments in a simulator
in which agents move around are currently limited to simple environments.
Connection to the actual world will be attempted by adjusting factors such
as the field in which the agents move, the size of the agents, or the variation
in velocity of agents to make the e nvironment in the simulator similar to the
256 Agent-Based Modeling and Simulation with Swarm
real environment (examples include the size of the exper imental equipment,
number of ants, and s peed of ants [80, 81]) in experiments involving animals.
In this book, both the constructive approach and actual examples have
been introduced to help clarify complex systems and artificia l life. Past re-
search studies in this field show that the investigation of cause and effect is
of paramount importance. Therefore, current research has been presented as
actual examples for showing how to perform research using the bottom-up ap-
proach. Various multi-agent simulation tests implemented in this book have
been categorized and discussed in relevant sections.
The approach explained in this book may not be applicable for research
studies dealing with some types of complex systems. However, structural mod-
els that can be imagined and constructed are necessary for simulations in
complex systems science, and the models become closer to r e ality by “iden-
tification of simulations.” The multi-agent s imulations discussed in this book
will increase the understanding of target pheno mena when actual models ex-
ist. This approach aims to investiga te the be havior of actual targets, i.e., the
cause and effect phenomenon, and then proceed to resear ching the systems in
a step-by-step manner . This book would immensely contribute to research in
the fields of complex systems and artificial life.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.