Learning task controllers on a humanoid robot using multi-objective optimization - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Poster Communications Year : 2021

Learning task controllers on a humanoid robot using multi-objective optimization

Abstract

Designing controllers for complex robots is not an easy task. Often, researchers hand-tune controllers for humanoid robots, but this approach requires lots of time for a single controller which cannot generalize accurately to varied tasks. We present a method which uses a multi-objective evolutionary algorithm with various training trajectories and outputs a diverse set of well-functioning controller weights and gains. The results of this optimization in the Talos robot show which weight and gain ranges can be used for a robust controller and prove that the optimization yields a diverse set of controller parameters, many of which can succeed on modified robot models.
Fichier principal
Vignette du fichier
Extended_abstract_ICRA_2021_workshop.pdf (376.6 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03255102 , version 1 (09-06-2021)

Identifiers

  • HAL Id : hal-03255102 , version 1

Cite

Evelyn d'Elia, Jean-Baptiste Mouret, Jens Kober, Serena Ivaldi. Learning task controllers on a humanoid robot using multi-objective optimization. ICRA 2021: 5th Full-Day Workshop on Legged Robots (Virtual), Jun 2021, Xi'an/Virtual, China. ⟨hal-03255102⟩
107 View
134 Download

Share

Gmail Facebook X LinkedIn More