Enregistré dans:
Détails bibliographiques
Auteurs principaux: Escoriza, Adrià López, Hansen, Nicklas, Tao, Stone, Mu, Tongzhou, Su, Hao
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2503.01837
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866916986187939840
author Escoriza, Adrià López
Hansen, Nicklas
Tao, Stone
Mu, Tongzhou
Su, Hao
author_facet Escoriza, Adrià López
Hansen, Nicklas
Tao, Stone
Mu, Tongzhou
Su, Hao
contents Long-horizon tasks in robotic manipulation present significant challenges in reinforcement learning (RL) due to the difficulty of designing dense reward functions and effectively exploring the expansive state-action space. However, despite a lack of dense rewards, these tasks often have a multi-stage structure, which can be leveraged to decompose the overall objective into manageable subgoals. In this work, we propose DEMO3, a framework that exploits this structure for efficient learning from visual inputs. Specifically, our approach incorporates multi-stage dense reward learning, a bi-phasic training scheme, and world model learning into a carefully designed demonstration-augmented RL framework that strongly mitigates the challenge of exploration in long-horizon tasks. Our evaluations demonstrate that our method improves data-efficiency by an average of 40% and by 70% on particularly difficult tasks compared to state-of-the-art approaches. We validate this across 16 sparse-reward tasks spanning four domains, including challenging humanoid visual control tasks using as few as five demonstrations.
format Preprint
id arxiv_https___arxiv_org_abs_2503_01837
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multi-Stage Manipulation with Demonstration-Augmented Reward, Policy, and World Model Learning
Escoriza, Adrià López
Hansen, Nicklas
Tao, Stone
Mu, Tongzhou
Su, Hao
Machine Learning
Computer Vision and Pattern Recognition
Robotics
Long-horizon tasks in robotic manipulation present significant challenges in reinforcement learning (RL) due to the difficulty of designing dense reward functions and effectively exploring the expansive state-action space. However, despite a lack of dense rewards, these tasks often have a multi-stage structure, which can be leveraged to decompose the overall objective into manageable subgoals. In this work, we propose DEMO3, a framework that exploits this structure for efficient learning from visual inputs. Specifically, our approach incorporates multi-stage dense reward learning, a bi-phasic training scheme, and world model learning into a carefully designed demonstration-augmented RL framework that strongly mitigates the challenge of exploration in long-horizon tasks. Our evaluations demonstrate that our method improves data-efficiency by an average of 40% and by 70% on particularly difficult tasks compared to state-of-the-art approaches. We validate this across 16 sparse-reward tasks spanning four domains, including challenging humanoid visual control tasks using as few as five demonstrations.
title Multi-Stage Manipulation with Demonstration-Augmented Reward, Policy, and World Model Learning
topic Machine Learning
Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2503.01837