Optimización en sistemas multi-robot mediante embodied evolution
- Trueba Martínez, Pedro José
- Francisco Bellas Bouza Co-director
- A. Prieto Co-director
Defence university: Universidade da Coruña
Fecha de defensa: 23 June 2017
- Javier de Lope Asiaín Chair
- José Santos Reyes Secretary
- Pilar Caamaño Sobrino Committee member
Type: Thesis
Abstract
In this PhD thesis, a version of the Embodied Evolution (EE) algorithm has been developed that generalizes the existing version on the field, with the goal of advancing in the standardization of this paradigm such that it could be studied in a formal way and discover its strengths and limitations. The developed algorithm has been named “canonical” because it was designed after analyzing in detail the basic common processes of the di↵erent variants in Embodied Evolution, in a way that only contains the intrinsic processes and a parametrization as general as possible. The operation of this Embodied Evolution canonical algorithm has been analyzed in a set of theoretic functions that represent the di↵erent search landscapes in collective optimization, in which a sensibility analysis has been performed. Finally, the conclusions of the theoretic analysis has been validated in a real task in which the validity of the approximation has been confirmed through the successful optimization of the emergent coordination of the multi-robot system, both in performance and in automatic organization in species, a fundamental property in this paradigm.