Heterochronic scaling of developmental durations in evolved soft robots

Abstract

In the evolution of Generative and Developmental Systems (GDS), the choice of where along the ontogenic trajectory to stop development in order to measure fitness can have a profound effect upon the emergent solutions. After illustrating the complexities of ontogenic fitness trajectories, we introduce a GDS encoding without an a priori fixed developmental duration, which instead slowly increases the duration over the course of evolution. Applied to a soft robotic locomotion task, we demonstrate how this approach can not only retain the well known advantages of developmental encodings, but also be more efficient and arrive at more parsimonious solutions than approaches with static developmental time frames.