In Vivo Veritas: Towards the Evolution of Things

Agoston Endre Eiben

Abstract. The main thesis of the position paper is that in the near future it will be possible to create populations of animate physical objects that undergo evolution in real space and real time. The resulting systems will differ from Evolutionary Computing in two crucial aspects. First, the individuals will be physical rather than digital. This requires reproduction operators for physical objects, which forms an engineering challenge. Second, the evolutionary process will be induced by the autonomous behavior of the individuals themselves, not by some central evolutionary agency that orchestrates selection and reproduction. These differences imply severe challenges for evolutionary algorithm designers because ‘tricks’ that work in in silico may not work in vivo. However, overcoming these challenges will ignite the development of a new field that combines Evolutionary Computing, Robotics, Artificial Life, and Embodied AI with a great potential for engineering as well as scientific research.

Keywords: Embodied Evolution, Evolutionary Computing, Evolutionary Robotics, Self-reproducing Robots

LNCS 8672, p. 24 ff.

Full article in PDF | BibTeX

lncs@springer.com
© Springer International Publishing Switzerland 2014