HuggingFace Daily Papers·· 1 天前
DONIC：从人类演示学习通用双臂灵巧跟踪控制
DONIC: Generalist Dexterous Bimanual Tracking Control from Human Demonstrations
AI 导读
DONIC 是首个跨操作类别与人类动作来源联合训练的通用双臂灵巧跟踪器，配套发布 BiHOI 多视角数据集，含 1575 段序列、229 个物体。它结合 PPO 与行为克隆形成反馈循环，将验证成功的执行回收为训练演示；单一 checkpoint 在训练物体未见轨迹上成功率达 86.79%，在留出物体上为 57.32%，分别超过 SOTA 双臂跟踪基线 77.36 和 45.73 个百分点。
正文 · 原文
Authors:Mingze Sun, Ziang Cao, Yuan Fei, Qiankuan Tao, Zihan Wang, Tianqi Liu, Junhao Chen, Haiwen Diao, Fangzhou Hong, Zhaoxi Chen, Ziwei Liu
Abstract:Despite recent progress in dexterous reference tracking, existing models focus primarily on single-hand manipulation and a limited range of interactions, leaving the bimanual dexterous behaviors common in real life unaddressed. The scarcity of contact-rich bimanual manipulation data and the increased complexity of coordinating contacts make developing a generalist bimanual tracker particularly challenging. Therefore, we present 1) DONIC, the first generalist bimanual tracker jointly trained across operation categories and human-motion sources, together with 2) BiHOI, a multi-view dataset of 1,575 sequences spanning 229 objects and emphasizing sustained bimanual interaction, where reconstructed interactions are retargeted into robot references. DONIC combines proximal policy optimization with behavioral cloning in a feedback loop that recycles verified successful executions into training demonstrations. An object-centered per-hand reward maintains object-motion feedback while supervising each hand according to its reference contact role. Clip-Centered Advantages remove reference-dependent advantage offsets while preserving within-trajectory differences. Together, these designs enable one single policy to learn coordinated manipulation across heterogeneous references. Without finetuning, a single checkpoint achieves 86.79% success on unseen trajectories of training objects and 57.32% on held-out objects, surpassing the state-of-the-art bimanual tracking baselines by 77.36 and 45.73 percentage points.
| Subjects: | Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.29997 [cs.RO] |
| (or arXiv:2610.29997v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.29997 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Mingze Sun [view email]
[v1]
Wed, 21 Oct 2026 16:36:02 UTC (5,726 KB)
来源：HuggingFace Daily Papers · arxiv.org