Skip to content

论文arXiv 2611.1013712 天前

统一惯性编码器实现人形机器人零样本状态估计与语义理解 ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:A Unified Inertial Encoder for Human and Humanoid Motion(arXiv 2611.10137)

核心内容 ​

针对人形机器人训练数据覆盖受限问题,研究提出统一惯性编码器(Unified Inertial Encoder),利用运动学IMU合成技术构建人与机器人共享的惯性输入空间。该方法联合训练人类数据与重定向后的人形数据,结合连续时间编码及传感器/运动随机化策略,实现了速度估计与文本-动作对齐。实验表明,模型无需物理仿真、真机轨迹或微调即可零样本迁移至Unitree G1,在五个数据集上取得最低速度误差,实机测试达0.099m/s。

技术方案核心在于通过运动学IMU合成消除人体与人形机器人在几何结构、传感配置及物理扰动上的差异,建立统一表征接口。在共享输入空间内,模型同步优化惯性里程计精度与语言接地能力。除速度估计外,该编码器还展现出跨域语义泛化能力,能够检索仅在人类训练集中出现但经重定向映射到G1的动作类别,证明了利用大规模人类运动数据增强人形机器人状态估计与语义理解的可行性,摆脱了对高成本真机采集数据的依赖。

关键看点 ​

  • 零样本迁移:无需物理仿真、真机训练轨迹或微调,模型直接部署于Unitree G1,实测速度误差低至0.099m/s
  • 统一惯性接口:采用运动学IMU合成技术对齐人与机器人身体部位数据,克服几何与传感差异,构建共享输入空间
  • 语义泛化能力:支持检索仅在人类训练集中出现的动作描述,验证了人类运动知识向人形机器人语义理解的跨域迁移
  • 联合训练策略:融合连续时间编码与传感器/运动随机化,在同一框架下协同优化速度估计精度与文本-动作对齐效果

工程与产业观察 ​

该工作为具身智能提供了低成本数据复用范式,通过运动学IMU合成打通人类与机器人惯性数据壁垒。工程上证明了无需昂贵真机采集与仿真微调,仅凭人类数据即可实现高精度零样本状态估计与语义理解,显著降低人形机器人感知算法落地门槛,对解决具身数据稀缺问题具有重要参考价值。

英文摘要 (Abstract) ​

Learning humanoid state estimation and language-grounded motion understanding from robot recordings ties training coverage to the robot's motion repertoire. Human motion datasets offer broader coverage, yet transferring this knowledge requires overcoming differences in body geometry, sensing, and physical disturbances. To enable this transfer, we propose a unified inertial encoder whose shared inertial interface uses kinematic IMU synthesis to represent human and humanoid motion at corresponding body sites. Within this common input space, joint training on human and kinematically retargeted humanoid motion combines velocity estimation and motion--text alignment with continuous-time encoding and sensor and motion randomization. The resulting model transfers zero-shot to a physical Unitree G1 without physics simulation, real-robot training trajectories, or fine-tuning, achieving the lowest velocity error among evaluated methods across five datasets, including 0.099m/s on real G1. Its motion-description retrieval also extends to retargeted G1 motion categories observed only in human training, demonstrating how human motion knowledge can support the generalization of humanoid inertial odometry and language-grounded motion understanding.

作者:Yutian Chen, Shibo Zhao, Rocky Duan, Karen Liu, Carmelo Sferrazza, Guanya Shi, Yuheng Qiu, Koushil Sreenath

基于本页技术方向关联的整机参数、核心零部件、实测报告与同赛道选型横评。

外部来源

本页事实对应的公开页面,以原文为准。(点击将离开本站)

  1. arXiv 2611.10137:A Unified Inertial Encoder for Human and Humanoid Motion ↗访问 2026-11-11