Matteo De Carlo, Daan Zeeuwe, Eliseo Ferrante, Gerben Meynen, Jacintha Ellers, A.E. Eiben

Robotic task affects the resulting morphology and behaviour in evolutionary robotics


In 2020 IEEE Symposium Series on Computational Intelligence (SSCI), 2020

DOI 10.1109/SSCI47803.2020.9308272

Abstract

In evolution, the evolutionary success of individuals is influenced in equal parts by the environment they are living in and by the adapting capability that they possess. In this work, instead of analyzing the adaptability of a population, we investigate how different environments influence the evolution of morphologies and controllers of modular robots. Previous work analyzed how robot evolution is influenced by different environments while keeping fixed the robot task. However, each environment may present different tasks that influence robot success. In this work, we investigate how different tasks in the same physical environment influence the evolutionary outcome. Specifically, we compare morphologies and behaviors under two tasks, undirected locomotion, and rotation in a flat terrain environment. We perform this analysis by analyzing the morphological and behavioral traits of the final population. Results suggest that both morphology and behavior types are strongly influenced by the environment, with the rotation task producing more proportionate and balanced robots compared to the locomotion task.