AdaptNet: Policy adaptation for physics-based character control

Xu, Pei, Xie, Kaixiang, Andrews, Sheldon, Kry, Paul G., Neff, Michael, McGuire, Morgan, Karamouzas, Ioannis et Zordan, Victor. 2023. « AdaptNet: Policy adaptation for physics-based character control ». ACM Transactions on Graphics, vol. 42, nº 6.

Prévisualisation

PDF Andrews-S-2023-28256.pdf - Version publiée Licence d'utilisation : Creative Commons CC BY. Télécharger (46MB) | Prévisualisation

Résumé

Motivated by humans' ability to adapt skills in the learning of new ones, this paper presents AdaptNet, an approach for modifying the latent space of existing policies to allow new behaviors to be quickly learned from like tasks in comparison to learning from scratch. Building on top of a given reinforcement learning controller, AdaptNet uses a two-tier hierarchy that augments the original state embedding to support modest changes in a behavior and further modifies the policy network layers to make more substantive changes. The technique is shown to be effective for adapting existing physics-based controllers to a wide range of new styles for locomotion, new task targets, changes in character morphology and extensive changes in environment. Furthermore, it exhibits significant increase in learning efficiency, as indicated by greatly reduced training times when compared to training from scratch or using other approaches that modify existing policies. Code is available at https://motion-lab.github.io/AdaptNet.

Type de document:	Article publié dans une revue, révisé par les pairs
Professeur:	Professeur Andrews, Sheldon
Affiliation:	Génie logiciel et des technologies de l'information
Date de dépôt:	04 janv. 2024 20:36
Dernière modification:	08 janv. 2024 20:12
URI:	https://espace2.etsmtl.ca/id/eprint/28256

Actions (Authentification requise)

Dernière vérification avant le dépôt