We propose a method for evolving neural network controllers robust with respect to variations of the environmental conditions (i.e. that can operate effectively in new conditions immediately, without the need to adapt to variations). The method specifies how the fitness of candidate solutions can be evaluated, how the environmental conditions should vary during the course of the evolutionary process, which algorithm can be used, and how the best solution can be identified. The obtained results show how the method proposed is effective and computational tractable. It allows to improve performance on an extended version of the double-pole balancing problem, outperform the best available human-designed controllers on a car racing problem, and generate effective solutions for a swarm robotic problem. The comparison of different algorithms indicates that the CMA-ES and xNES methods, that operate by optimizing a distribution of parameters, represent the best options for the evolution of robust neural network controllers.
Robust optimization through neuroevolution
Public Library of Science, San Francisco, CA , Stati Uniti d'America
PloS one 14 (2019): 1–27. doi:10.1371/journal.pone.0213193
info:cnr-pdr/source/autori:Pagliuca P.; Nolfi S./titolo:Robust optimization through neuroevolution/doi:10.1371/journal.pone.0213193/rivista:PloS one/anno:2019/pagina_da:1/pagina_a:27/intervallo_pagine:1–27/volume:14