Physics-based motion control through hierarchical neuroevolution - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Physics-based motion control through hierarchical neuroevolution

Mart Hagenaars
  • Fonction : Auteur
  • PersonId : 1124206
Arjan Egges
  • Fonction : Auteur
  • PersonId : 1124207

Résumé

In this paper, we propose a hierarchical neuroevolution technique for physics-based character animation control. The artificial neural network that makes up the controller is composed of a number of interdependent control modules. As a proof-of-concept, modules for posing, standing, and reaching motions are demonstrated. We show that evolving these modules one-by-one, with each of them dependent on its predecessors, allows evolution to converge faster, and possibly deliver better and more stable results than common, i.e. non-hierarchical, controllers.
Fichier principal
Vignette du fichier
Physics-based motion control through hierarchical neuroevolution CASA2014.pdf (726.05 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03542708 , version 1 (25-01-2022)

Identifiants

  • HAL Id : hal-03542708 , version 1

Citer

Mart Hagenaars, Nicolas Pronost, Arjan Egges. Physics-based motion control through hierarchical neuroevolution. 27th Conference on Computer Animation and Social Agents (CASA), May 2014, Houston, United States. ⟨hal-03542708⟩
25 Consultations
28 Téléchargements

Partager

Gmail Facebook X LinkedIn More