Embodied Computational Evolution: A Model for Investigating Randomness and the Evolution of Morphological Complexity

For an integrated understanding of how evolutionary dynamics operate in parallel on multiple levels, computational models can enable investigations that would be otherwise infeasible or impossible. We present one modeling framework, Embodied Computational Evolution (ECE), and employ it to investigate how two types of randomness-genetic and developmental-drive the evolution of morphological complexity. With these two types of randomness implemented as germline mutation and transcription error, with rates varied in an [Formula: see text] factorial experimental design, we tested two related hypotheses: ( ) Randomness in the gene transcription process alters the direct impact of selection on populations; and ( ) Selection on locomotor performance targets morphological complexity. The experiment consisted of 121 conditions; in each condition, nine starting phenotypic populations developed from different randomly generated genomic populations of 60 individuals. Each of the resulting 1089 phenotypic populations evolved over 100 generations, with the autonomous, self-propelled individuals under directional selection for enhanced locomotor performance. As encoded by their genome, individuals had heritable morphological traits, including the numbers of segments, sensors, neurons, and connections between sensors and motorized joints that they activated. An individual's morphological complexity was measured by three different metrics derived from counts of the body parts. In support of , variations in the rate of randomness in the gene transcription process varied the dynamics of selection. In support of , the morphological complexity of populations evolved adaptively.