A fitness-independent evolvability measure for evolutionary developmental systems

Abstract

Evolvability refers to the organisms ability to create heritable new phenotypes that potentially facilitate the organism's survival and reproduction. In this paper, a general evolvability measure for a computational model of evolutionary development is proposed. The measure is able to quantify individuals' evolvability, including robustness and innovation, independent of the fitness function of the evolutionary system. Empirical studies are performed to check the evolvability of individuals in in silico evolution of oscillatory behavior using the proposed evolvability measure. Our preliminary results suggest that evolvability of the developmental system can evolve without an explicit selection pressure on evolvability, confirming findings revealed in other artificial evolutionary systems.