SENTINEL: A Fully End-to-End Language-Action Model for Humanoid Whole Body Control

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Yuxuan, Jiang, Haobin, Yao, Shiqing, Ding, Ziluo, Lu, Zongqing
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915635262390272
author Wang, Yuxuan
Jiang, Haobin
Yao, Shiqing
Ding, Ziluo
Lu, Zongqing
author_facet Wang, Yuxuan
Jiang, Haobin
Yao, Shiqing
Ding, Ziluo
Lu, Zongqing
contents Existing humanoid control systems often rely on teleoperation or modular generation pipelines that separate language understanding from physical execution. However, the former is entirely human-driven, and the latter lacks tight alignment between language commands and physical behaviors. In this paper, we present SENTINEL, a fully end-to-end language-action model for humanoid whole-body control. We construct a large-scale dataset by tracking human motions in simulation using a pretrained whole body controller, combined with their text annotations. The model directly maps language commands and proprioceptive inputs to low-level actions without any intermediate representation. The model generates action chunks using flow matching, which can be subsequently refined by a residual action head for real-world deployment. Our method exhibits strong semantic understanding and stable execution on humanoid robots in both simulation and real-world deployment, and also supports multi-modal extensions by converting inputs into texts.
format Preprint
id arxiv_https___arxiv_org_abs_2511_19236
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SENTINEL: A Fully End-to-End Language-Action Model for Humanoid Whole Body Control
Wang, Yuxuan
Jiang, Haobin
Yao, Shiqing
Ding, Ziluo
Lu, Zongqing
Robotics
Artificial Intelligence
Existing humanoid control systems often rely on teleoperation or modular generation pipelines that separate language understanding from physical execution. However, the former is entirely human-driven, and the latter lacks tight alignment between language commands and physical behaviors. In this paper, we present SENTINEL, a fully end-to-end language-action model for humanoid whole-body control. We construct a large-scale dataset by tracking human motions in simulation using a pretrained whole body controller, combined with their text annotations. The model directly maps language commands and proprioceptive inputs to low-level actions without any intermediate representation. The model generates action chunks using flow matching, which can be subsequently refined by a residual action head for real-world deployment. Our method exhibits strong semantic understanding and stable execution on humanoid robots in both simulation and real-world deployment, and also supports multi-modal extensions by converting inputs into texts.
title SENTINEL: A Fully End-to-End Language-Action Model for Humanoid Whole Body Control
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2511.19236