ABSTRACT
Thomas Schmickl,a Karl Crailsheim,b Jean-Louis Deneubourg,c
In collective systems, aggregation is among the essential behaviors
that have to be performed before being able to accomplish collective
tasks. Obviously, agents have to converge to some interesting
spots where collective action has to take place. However, the
biological models that are presented here show that aggregation
can also be the starting point of elaborate decision making.
Aggregation phenomena go beyond the mere gathering of agents
to a specific place. Indeed, self-organized aggregation, as defined
later, has also many interesting properties leading, for example,
to collective choices. The examples discussed here show that
aggregation and collective decision making are often tightly linked
in natural systems, as well as in bioinspired distributed technical
systems. This finding implies that-when primarily designing
algorithms for aggregation-one can get additional interesting
collective intelligence capabilities as a consequence of underlying
self-organizing mechanisms. For such systems it is an important
prerequisite that the focal collective system’s microscopic behaviors
are linked and interconnected. These characteristics make designing
for emergence a very difficult and tricky problem as many features
have to be integrated at the same time and in real time. Thus,
it is interesting to look for biological examples as sources of
inspiration that are already offering important insights into how to
solve these kinds of issues. Two biological examples are presented
here, cockroaches and honeybees, both naturally distributed and
self-organizing animal societies. These societies offer solutions
to questions like, How can we integrate many individual and
environmental features/properties in an attempt to obtain collective
intelligence at the same time as aggregation of individuals? With
relatively simple algorithms leading to aggregation, one can also get,
“for free,” many different collective patterns and complex dynamics.
These two biological examples have been successfully implemented
in robots. The case studies presented here also illustrate the
interplay between individual behavior and the perception of other
agents and the environment. Finally, we also discuss the difficulties
arising from translating biological macroscopic models of animal
behavior to microscopic robotic implementation.
