One Policy, Many Embodiments: Unified Camera-Centric Action Geometry Pre-training for Heterogeneous Embodied Manipulation
2026-08-26 • Robotics
Robotics
AI summaryⓘ
The authors address a challenge in teaching robots and humans to work with many different types of robots and cameras by creating a shared way to describe actions based on what the cameras see, not on the specific robot commands. Their method, UCAG-P, turns various robot movements into common motions seen through cameras, making it easier for a single system to learn how to act across different robots and humans. They then convert these common actions back into robot-specific commands using a geometry-based translator. This approach lets one policy be trained on a large mix of data and perform well on several robot and simulation benchmarks without extra tuning.
vision-language-action (VLA) policiesembodied datarobot morphologycamera configurationaction retargetinggeometry-conditioned action translatormanipulation anchor motionzero-shot generalizationunified policyrobot kinematics
Authors
Xiaomi Embodied Intelligence Team, University of Macau, :, Shaoqing Xu, Fang Li, Guozhi Zhan, Zhixiang Duan, Yuhan Wang, Yuechen Luo, Shengyin Jiang, Hanbing Li, Zhiying Du, Longlong Wang, Longmei Jiang, Weixiang Liang, Ying Gong, Yong Pan, Ziping Zhao, Zhiyuan Chen, Yangwei You, Kun Ma, Qinyuan Liu, Hangjun Ye, Zhi-xin Yang
Abstract
Scaling generalist vision-language-action (VLA) policies is severely bottlenecked by the inherent heterogeneity of embodied data, which spans diverse robot morphologies, camera configurations, and low-level action spaces. Existing paradigms typically address this mismatch through explicit action retargeting, human-to-robot video synthesis, or dataset-specific adaptation branches, fundamentally hindering the joint learning of a unified policy. We introduce UCAG-P, a camera-centric unified action formulation that structurally aligns heterogeneous embodied datasets into a shared geometric action space. Rather than treating robot-specific commands as the shared policy target, UCAG-P represents manipulation through camera-observable anchor motion in image and camera-frame coordinates, treating robot arms, humanoids, and human hands as different embodiments of a common action schema. A geometry-conditioned action translator combines predicted motion with target-embodiment kinematics to produce executable controls. The resulting decoupled architecture allows a shared VLA policy to learn transferable manipulation geometry while retaining embodiment-specific controllability. UCAG-P is trained on 4.03K hours of robot and simulation data and 2.34K hours of human demonstrations. A single checkpoint reaches 98.3% on LIBERO, 88.7% and 89.2% on RoboTwin Easy and Hard, 82.0% zero-shot on LIBERO-Plus, and 62.0% on RoboCasa GR-1, without benchmark-specific fine-tuning.