Information Technology Reference
In-Depth Information
we took them as they have been proposed by their authors. Consequently, we have
developed a common scenario to test both reproduction strategies, so that the
robustness properties are exactly the same in all different experiments.
Earlier attempts were done by James Reggia
et al.
[17] to achieve the spontaneous
emergency of cellular reproducers from a random soup of cell-states. In 99% of the
cases emergence of self-replicating loops (2x2-cell patterns) is achieved, and the
interaction among these minimal replicators originate the construction of bigger
reproducing structures. Robustness of emerging patterns is obtained through a variety
of mechanisms. Due to the small size of the reproducers (all their cells are neighbor to
each other) the local rules are able to detect whenever one of these small replicating
agents is produced. Therefore its cohesion is protected through a shift on the state set
(from unbound to bound) and, accordingly, of the rules that guide the dynamic
behavior of the system. For instance, the newborn reproducers are able to clean the
surrounding area to allow their reproduction dynamics, and there are special failure-
detection mechanisms to recycle the cells belonging to an unsuccessful reproducer.
To summarize, our aim is to design a cellular model in which reproducing patterns
can be defined and operate in an uncertain environment, and to test the possibility of
emergence of such patterns from a random distribution of cell-states. We will define a
class of such reproducers and don't want to favor the appearance of any special or
canonical reproducing pattern. Moreover, we want to preserve the rule set along all
the phases of the system (emergence, reproduction and interaction between patterns),
so that the conditions that allow the appearance of replicators are the same as those in
which these replicators will operate. Finally, we want to design a cellular transition
function that encompasses the features of both self-inspection and genetic models in
order to test their relative suitability for the emergence of patterns.
3
Experimental Framework
To implement the approach sketched in the previous section we need to meet three
consecutive goals. First of all, a cellular model capable of encompassing two different
modes of reproduction. Once a unified framework is designed, it is necessary to
enhance the rule set, typically defined just for a quiescent environment, with
additional rules that allow a robust behavior of the reproducing patterns in presence of
a nonquiescent environment [18]. Finally, the experimental benchmark should be
designed, in which the method of generation of the initial configurations (primeval
soups) to test the emergence of replicators will be the clue.
3.1
Pattern Choice
Our unified model of replicating patterns is constituted by closed paths through which
construction signals can circulate, with a small constructing arm that serves to execute
the signals outside the pattern to produce structure. Each constructing signal codes a
particular direction (ahead, right or left), and when it arrives to the open end of a path,
it makes it one cell longer, following the appropriate direction. This general schema
can be found in many of the models introduced by the previous works quoted above.
