arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2610.36859cs.ROcs.AIcs.CV
收藏

ActiveAlign:通过主动感知实现自我中心操作的对齐

ActiveAlign: Aligning Egocentric Manipulation via Active Perception

NVIDIA(英伟达);UC San Diego(加州大学圣迭戈分校)

Ri-Zhao Qiu, Bowen Wen, Yifeng Jiang, Chen Tessler, Eric Heiden, Xue Bin Peng, Xiaolong Wang, Stan Birchfield, Erwin Coumans

AI总结:

提出ActiveAlign框架,通过主动感知的两个互补模块缩小真实-仿真差距,在Unitree G1机器人上实现毫米级物体精细操作,还可用于整理机器人操作数据集的仿真环境。

摘要:

真实到仿真再到真实(Real2Sim2Real)为规划、可解释性与安全执行提供结构化世界表征。现有多数方法依赖被动感知,因观测不完善产生真实-仿真差距。受人类通过动作主动感知的启发,我们提出框架ActiveAlign,利用主动感知从交互中缩小真实-仿真差距。该框架包含两个互补模块:(i)ActiveAlign-Waving,通过利用主动手部运动与可微分渲染执行手眼校准;(ii)ActiveAlign-Touching,通过触觉反馈优化物体对齐,实现毫米级精度。这些能力共同实现强真实-仿真对齐。我们证明ActiveAlign-Waving可用于为开源未校准机器人操作数据集整理仿真环境,并在真实Unitree G1机器人上验证,ActiveAlign可实现毫米级物体的精细操作,如按压按钮、移动滑块。

英文摘要:

Real-to-Sim-to-Real (Real2Sim2Real) enables structured world representations for planning, interpretability, and safe execution. Most existing approaches rely on passive perception, which creates real2sim gaps due to imperfect observation. Inspired by how humans perceive actively through actions, we propose ActiveAlign, a framework that leverages active perception to align the Real2Sim2Real gap from interactions. ActiveAlign introduces two complementary modules: (i) ActiveAlign-Waving, which performs hand-eye calibration by exploiting active hand motions and differentiable rendering, and (ii) ActiveAlign-Touching, which refines object alignment through tactile feedback to achieve millimeter-level precision. Together, these capabilities enable strong real2sim alignment. We demonstrate that ActiveAlign-Waving can be used to curate simulation environments for open-sourced uncalibrated robot manipulation datasets, and demonstrate on a real Unitree G1 robot that ActiveAlign enables fine manipulation of millimeter-scale objects such as pressing buttons and moving sliders.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑