Evolvability of the gentotype-phenotype relation in populations of self-replicating digital organisms in a Tierra-like system

Abstract

In other Tierra-like systems the genotype is a sequence of instructions and the phenotype is the corresponding executed algorithm. This way the genotype-phenotype mapping is constrained by the structure of a creature’s processor, and this structure was fixed for an evolutionary scenario in previous systems. Our approach here is to put the mapping under evolutionary control. We use a universal processor (analogous to a universal Turing-machine) and put the structural description of the creature’s processor as well as the instruction set of the actual processor into the organism’s genome. The life-cycle of an organism begins with building its actual processor, then the organism can start executing instructions in the rest of its genome with the newly built processor. Since the definitions of the processors and instruction sets are in the genome, they are subject to mutations and heritable variation enabling their evolution. In this work we investigate the evolutionary development of the processor structures. In evolving populations, changes in the components (registers, stacks, queues), variations in instruction-set size and the redefinition of the instructions can be observed during experiments.