FAST:透過預訓練與快速適應實現通用人形機械人全身控制
General Humanoid Whole-Body Control via Pretraining and Fast Adaptation
FAST 是一套通用人形機械人全身控制框架,可零樣本追蹤高動態動作並即時遙距操控,在模擬環境以外維持穩健性與平衡。框架採用 Mixture-of-Experts 架構及 CoM-Aware 控制,並以 Parseval-guided residual policy learning,為低質素或分佈外的動作參考進行輕量化快速適應。
Feb 12, 2026
(Accepted: CoRL 2026)
Zepeng Wang / Jiangxing Wang / Shiqing Yao / Yu Zhang / Ziluo Ding / Ming Yang / Yuxuan Wang / Haobin Jiang / Chao Ma / Xiaochuan Shi / Zongqing Lu‡
WHU/BeingBeyond/PKU
† / ‡ Corresponding author
Contents
Real-World DemoMethod OverviewExperimentCitation
We present FAST, a general humanoid whole-body control framework that enables Fast Adaptation and Stable Motion Tracking.

FAST enables zero-shot high-dynamic motion tracking and real-time teleoperation with strong robustness and balance beyond simulation. For low-quality or out-of-distribution motion references, FAST further supports lightweight residual adaptation for rapid and stable specialization.
Real-World Demo
FAST tracks a high-dynamic kick motion.
Text: "person hops on one leg"
Text: "a person does a spinning kick"
FAST tracks a long-horizon dance motion generated from a video.
Walk and kick football
High dynamic motion
Long-horizon motion "Taichi"
Play with human
Push chair
Grab toys and hold basket
Method Overview

An overview of FAST. The framework consists of three stages: (1) We construct a curated humanoid motion dataset via human-to-humanoid retargeting with auxiliary physical signals. (2) We train a general whole-body controller with a Mixture-of-Experts architecture and Center-of-Mass-Aware control. (3) We perform fast adaptation via Parseval-guided residual policy learning.
Experiment

Abaltions on fast adaptation results on target adaptation dataset (LaFan1) and source dataset (AMASS). The result demonstrates that combining Parseval and KL regularization enables fast adaptation with a favorable balance between target-domain performance and source-domain preservation.

Visualizations of representative motions comparing CoM-Aware Control with the baseline. CoM-Aware Control consistently maintains balance and reduces instability, while the baseline exhibits drift or falls.
Citation
@article{wang2026fast,
title={General Humanoid Whole-Body Control via Pretraining and Fast Adaptation},
author={Zepeng Wang and Jiangxing Wang and Shiqing Yao and Yu Zhang and Ziluo Ding and Ming Yang and Yuxuan Wang and Haobin Jiang and Chao Ma and Xiaochuan Shi and Zongqing Lu},
journal={arXiv preprint arXiv:2602.11929},
year={2026}
}來源:BeingBeyond:具身智能研究 · research.beingbeyond.com