Skip to content

论文arXiv 2610.0911722 天前

Workhorse框架实现Unitree G1无机器人数据全身操作 ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:Workhorse: Learning Robust Whole-Body Humanoid Loco-Manipulation from Human Data(arXiv 2610.09117)

核心内容 ​

针对人形机器人缺乏接触丰富操作数据难题,研究提出Workhorse框架,仅使用第一人称RGB与本体感知的人类演示数据训练Unitree G1。系统解耦为视觉规划器与RL全身跟踪器,两者独立训练且无需运动重定向。在真机验证中,G1成功执行双手搬运、踢箱、接抛掷物及推倒攀爬行李箱等复杂任务,并能在人为推拉干扰或物体被夺走后自主恢复,验证了从人到机器人的直接策略迁移能力。

技术架构上,视觉规划器输出躯干、双腕及双脚的五连杆位姿目标,RL跟踪器负责底层执行。核心创新在于交叉误差数据增强:将一方部署时的错误注入另一方训练集以模拟真实偏差。仿真测试显示,该系统在标准场景下搬运成功率达77%,在40 N.s外力扰动下保持64%成功率;若基于相同演示数据重训第二台人形机器人,无扰动搬运成功率提升至83%,证实了方案对硬件差异的泛化潜力。

关键看点 ​

  • 无重定向训练:视觉规划与RL跟踪策略均直接从原始人类姿态数据学习,省去传统运动重定向环节,降低数据预处理门槛。
  • 交叉误差增强:通过将对方策略的部署误差注入自身训练集,使两模块在训练阶段即适应彼此的真实偏差,显著提升闭环鲁棒性。
  • 五连杆视觉规划:视觉规划器直接预测躯干、双腕及双脚五个关键端点的6-DoF位姿,为全身控制器提供稀疏但语义明确的目标指引。
  • 量化鲁棒指标:仿真环境中系统在承受40 N.s外部冲击时仍维持64%的任务完成率,真机演示了受推挤及物体被抢夺后的动态恢复能力。

工程与产业观察 ​

该工作验证了“纯人类视频/动捕数据+交叉误差增强”可替代昂贵的机器人遥操作数据,为人形机器人全身操作提供了低成本冷启动路径。其解耦架构与误差对齐机制解决了模仿学习中常见的分布偏移问题,对构建通用具身基础模型的数据飞轮具有重要工程参考价值。

英文摘要 (Abstract) ​

Humanoid robots still struggle to plan contact-rich whole-body manipulation from egocentric RGB and proprioception. Workhorse learns such manipulation from robot-free human demonstrations. A visual planner predicts five-link targets: the poses of the torso, both wrists, and both feet. A reinforcement-learning whole-body tracker follows them on the robot. Both policies train separately on the same recorded human poses, without retargeting. We augment the training data of each policy to imitate the errors that the other makes at deployment. On a real Unitree G1, Workhorse transport boxes with its hands and a kick, catches a thrown box, and topples and climbs a suitcase. During box transport, we show recoveries after a person pushes the robot or takes the box away. In a simulated copy of the demonstration room, the system completes box transport in 77% of episodes, and in 64% under 40 N.s pushes. With both policies retrained from the same demonstrations, a simulated second humanoid completes box transport in 83% of episodes without pushes.

作者:Songbo Hu, Qiayuan Liao, Yufeng Chi, Kevin Zakka, Yakun Sophia Shao, Pieter Abbeel, Koushil Sreenath

来源

本页事实对应的公开页面,以原文为准。

  1. arXiv 2610.09117:Workhorse: Learning Robust Whole-Body Humanoid Loco-Manipulation from Human Data访问 2026-10-12