MetaLoco: Universal Quadrupedal Locomotion with Meta-Reinforcement Learning and Motion Imitation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zargarbashi, Fatemeh, Di Giuro, Fabrizio, Cheng, Jin, Kang, Dongho, Sukhija, Bhavya, Coros, Stelian
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913571061891072
author Zargarbashi, Fatemeh
Di Giuro, Fabrizio
Cheng, Jin
Kang, Dongho
Sukhija, Bhavya
Coros, Stelian
author_facet Zargarbashi, Fatemeh
Di Giuro, Fabrizio
Cheng, Jin
Kang, Dongho
Sukhija, Bhavya
Coros, Stelian
contents This work presents a meta-reinforcement learning approach to develop a universal locomotion control policy capable of zero-shot generalization across diverse quadrupedal platforms. The proposed method trains an RL agent equipped with a memory unit to imitate reference motions using a small set of procedurally generated quadruped robots. Through comprehensive simulation and real-world hardware experiments, we demonstrate the efficacy of our approach in achieving locomotion across various robots without requiring robot-specific fine-tuning. Furthermore, we highlight the critical role of the memory unit in enabling generalization, facilitating rapid adaptation to changes in the robot properties, and improving sample efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2407_17502
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MetaLoco: Universal Quadrupedal Locomotion with Meta-Reinforcement Learning and Motion Imitation
Zargarbashi, Fatemeh
Di Giuro, Fabrizio
Cheng, Jin
Kang, Dongho
Sukhija, Bhavya
Coros, Stelian
Robotics
This work presents a meta-reinforcement learning approach to develop a universal locomotion control policy capable of zero-shot generalization across diverse quadrupedal platforms. The proposed method trains an RL agent equipped with a memory unit to imitate reference motions using a small set of procedurally generated quadruped robots. Through comprehensive simulation and real-world hardware experiments, we demonstrate the efficacy of our approach in achieving locomotion across various robots without requiring robot-specific fine-tuning. Furthermore, we highlight the critical role of the memory unit in enabling generalization, facilitating rapid adaptation to changes in the robot properties, and improving sample efficiency.
title MetaLoco: Universal Quadrupedal Locomotion with Meta-Reinforcement Learning and Motion Imitation
topic Robotics
url https://arxiv.org/abs/2407.17502