A Hierarchical Scheme for Adapting Learned Quadruped Locomotion - Archive ouverte HAL
Conference Papers Year : 2023

A Hierarchical Scheme for Adapting Learned Quadruped Locomotion

Abstract

The quadruped locomotion task is generally con- ditioned on a velocity command that is user-defined. However, certain features of the locomotion are not represented in this high-level definition of the task, e.g., swing feet height, step length and expended energy. Using reinforcement learning, many of these features can be determined by the reward function terms and scales that are often fixed. In this work, we propose a deep reinforcement learning (DRL) approach to learn control policies augmented with parameters that modify different aspects of the reward function and control setup which, in turn, result in variations of the locomotion. We can then define a hierarchical architecture where a high level policy infers the suitable parameters to complete a given task. We show that this setup makes it possible to learn more complex behaviours that can be adapted for different terrains and environments to ensure successful and efficient locomotion. We display our results by deploying the low-level parameterized policy on the MIT Mini-Cheetah quadruped.
Fichier principal
Vignette du fichier
HRL_Loco-2.pdf (2.68 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04174932 , version 1 (01-08-2023)
hal-04174932 , version 2 (12-12-2023)

Identifiers

Cite

Michel Aractingi, Pierre-Alexandre Léziart, Thomas Flayols, Julien Perez, Tomi Silander, et al.. A Hierarchical Scheme for Adapting Learned Quadruped Locomotion. IEEE-RAS Humanoids 2023, IEEE-RAS, Dec 2023, Austin, TX, United States. ⟨10.1109/Humanoids57100.2023.10375148⟩. ⟨hal-04174932v2⟩
202 View
316 Download

Altmetric

Share

More