This paper introduces a method through which, using genetic
algorithms
on two dimensional cellular
automata
, we obtain emergent phenomena of
self-replication
. Three indices of
complexity
, based on input
entropy
have been developed and used as fitness
functions in the evolutionary experiments. The genetic algorithm,
realized by a special design of the genome, is efficient and the
research in the CA rules space has given appreciable results, both
for the quantity and for the quality of the emergent phenomena. We
found that each of these indices is strictly connected to the
complexity of the rules and to the self-reproducers behavior
contained in them. We noticed that self-reproduction is a widespread
process also in artificial life simulations. Almost all the evolved
rules manifest self-reproducers, as if this process were an embedded
characteristic of artificial/living matter. The self-reproducers,
different in shape, function and behavior, reveal an algorithmic
logic in self-replication, which follows different but synchronized
rhythms, evidencing variation, increasing structural complexity and
some of them general constructive capacity.