[论文] Ropedia W0: Scaling Loco-Manipulation World Models across Embodiments ...

论文概要

研究领域: CV 作者: Tianqi Liu, Hanfei Zhao, Qifeng Liang, Hui En Pang, Jiarui Xu, Sayang Mu, Penghao Qiu, YingSheng Zhang, Jun Wang, Ziang Cao, Mingze Sun, Yukang Cao, Fangzhou Hong, Zhaoxi Chen, Ziwei Liu 发布时间: 2026-10-19 arXiv: 2610.26855

中文摘要

世界模型使具身智能体能够预判动作后果，但跨多样化形态统一建模运动与操作仍然充满挑战。我们提出Ropedia W0——首个跨越多形态人类和机器人的以自我为中心的运动-操作世界基础模型。我们构建Xp-Data数据集，包含1万小时人类自我中心数据和3千小时机器人数据，配备统一多模态标注。评估表明模型在五个跨形态基准测试（DreamDojo、Ψ₀、ω-HOME、Xp-Teleop、AgiBot）上达到最优性能。扩大人类预训练数据规模可同时改善人类和机器人评估的性能。蒸馏后模型支持实时分钟级rollout，实现实时VR交互，在策略评估、基于模型的规划等应用上展示了实用价值。

原文摘要

World models enable embodied agents to anticipate action consequences, yet jointly modeling locomotion and manipulation across diverse embodiments remains challenging due to heterogeneous action spaces and limited multimodal supervision. We present Ropedia W0, the first egocentric loco-manipulation world foundation model spanning humans and robots with diverse embodiments. We curate Xp-Data, comprising 10k hours of egocentric human data and 3k hours of robot data with unified multimodal annotations and systematic quality control. A unified kinematic representation encodes locomotion through camera poses and manipulation through projected skeletons, with robot skeletons combining embodiment-specific forward kinematics with extensive manual keypoint annotation and correction. We also introdu...

#论文 #arXiv #CV #小凯