SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Kuang, Yuxuan, Geng, Haoran, Elhafsi, Amine, Do, Tan-Dzung, Abbeel, Pieter, Malik, Jitendra, Pavone, Marco, Wang, Yue
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909645909524480
author Kuang, Yuxuan
Geng, Haoran
Elhafsi, Amine
Do, Tan-Dzung
Abbeel, Pieter
Malik, Jitendra
Pavone, Marco
Wang, Yue
author_facet Kuang, Yuxuan
Geng, Haoran
Elhafsi, Amine
Do, Tan-Dzung
Abbeel, Pieter
Malik, Jitendra
Pavone, Marco
Wang, Yue
contents Humanoid robots hold significant potential in accomplishing daily tasks across diverse environments thanks to their flexibility and human-like morphology. Recent works have made significant progress in humanoid whole-body control and loco-manipulation leveraging optimal control or reinforcement learning. However, these methods require tedious task-specific tuning for each task to achieve satisfactory behaviors, limiting their versatility and scalability to diverse tasks in daily scenarios. To that end, we introduce SkillBlender, a novel hierarchical reinforcement learning framework for versatile humanoid loco-manipulation. SkillBlender first pretrains goal-conditioned task-agnostic primitive skills, and then dynamically blends these skills to accomplish complex loco-manipulation tasks with minimal task-specific reward engineering. We also introduce SkillBench, a parallel, cross-embodiment, and diverse simulated benchmark containing three embodiments, four primitive skills, and eight challenging loco-manipulation tasks, accompanied by a set of scientific evaluation metrics balancing accuracy and feasibility. Extensive simulated experiments show that our method significantly outperforms all baselines, while naturally regularizing behaviors to avoid reward hacking, resulting in more accurate and feasible movements for diverse loco-manipulation tasks in our daily scenarios. Our code and benchmark will be open-sourced to the community to facilitate future research. Project page: https://usc-gvl.github.io/SkillBlender-web/.
format Preprint
id arxiv_https___arxiv_org_abs_2506_09366
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending
Kuang, Yuxuan
Geng, Haoran
Elhafsi, Amine
Do, Tan-Dzung
Abbeel, Pieter
Malik, Jitendra
Pavone, Marco
Wang, Yue
Robotics
Machine Learning
Humanoid robots hold significant potential in accomplishing daily tasks across diverse environments thanks to their flexibility and human-like morphology. Recent works have made significant progress in humanoid whole-body control and loco-manipulation leveraging optimal control or reinforcement learning. However, these methods require tedious task-specific tuning for each task to achieve satisfactory behaviors, limiting their versatility and scalability to diverse tasks in daily scenarios. To that end, we introduce SkillBlender, a novel hierarchical reinforcement learning framework for versatile humanoid loco-manipulation. SkillBlender first pretrains goal-conditioned task-agnostic primitive skills, and then dynamically blends these skills to accomplish complex loco-manipulation tasks with minimal task-specific reward engineering. We also introduce SkillBench, a parallel, cross-embodiment, and diverse simulated benchmark containing three embodiments, four primitive skills, and eight challenging loco-manipulation tasks, accompanied by a set of scientific evaluation metrics balancing accuracy and feasibility. Extensive simulated experiments show that our method significantly outperforms all baselines, while naturally regularizing behaviors to avoid reward hacking, resulting in more accurate and feasible movements for diverse loco-manipulation tasks in our daily scenarios. Our code and benchmark will be open-sourced to the community to facilitate future research. Project page: https://usc-gvl.github.io/SkillBlender-web/.
title SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending
topic Robotics
Machine Learning
url https://arxiv.org/abs/2506.09366