Repeated structure and dissociation of genotypic and phenotypic complexity in artificial ontogeny
Josh Bongard, Rolf Pfeifer · 2001
In this paper, a minimal model of ontogenetic development, combined with differential gene expression and a genetic algorithm, is used to evolve both the morphology and neural control of agents that perform a block-pushing task in a physically-realistic, virtual environment. We refer to this methodology as artificial ontogeny (AO). It is demonstrated that evolved genetic regulatory networks in AO give rise to hierarchical, repeated phenotypic structures. Moreover, it is shown that the indirect genotype to phenotype mapping results in a dissociation between the information content in the genome, and the complexity of the evolved agent. It is argued that these findings support the claim that artificial ontogeny is a useful design tool for the evolutionary design of virtual agents and real-world robots.