# MobileVISTA框架解决移动操作位姿泛化难题，Galaxea R1验证有效

> 原文链接：https://robotics.hellobike.com/papers/2026-10-05-2610-07511
> 频道：papers | 更新日期：2026-10-05 | 标签：auto, arxiv
> 核心摘要：针对移动机器人单点位姿训练策略脆弱问题，提出MobileVISTA数据增强框架。通过联合增强自中心视觉与重定向动作补偿基座偏差，在Galaxea R1 Pro实机验证中显著提升测试时分布外位姿鲁棒性，无需额外采集或生成模型。

# MobileVISTA框架解决移动操作位姿泛化难题，Galaxea R1验证有效

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[MobileVISTA: Generative Data Augmentation for Pose Generalization in Mobile Manipulation](https://arxiv.org/abs/2610.07511)（arXiv 2610.07511）

## 核心内容

移动操作机器人在非结构化环境中常因部署位姿厘米级偏差导致端到端策略失效。MobileVISTA框架通过将标准位姿演示数据转换为多样化位姿扰动数据，解决自中心视角下相机随运动链移动导致的观测分布偏移问题。该方法在仿真及Galaxea R1 Pro真机上验证，无需新数据采集即可提升策略对未见位姿的适应性，尤其适用于相机安装在执行链上的人形机器人平台。

技术核心在于联合处理视觉观测增强与动作重定向：一方面合成位姿变化后的自中心图像，另一方面根据基座位姿偏移重新计算末端执行器轨迹以维持任务语义。不同于假设相机刚性固定或机器人主体出帧的传统方法，该方案专门适配人形等自中心平台。实验表明，在双臂及人形仿真任务与真机测试中，增强后策略在分布外位姿下的成功率显著优于基线，且收益在人形本体上最为明显。

## 关键看点

- **数据增强机制**：联合增强自中心视觉观测并重定向动作序列，补偿基座位姿变化对末端轨迹的影响
- **真机验证平台**：在Galaxea R1 Pro移动操作机器人上完成实机测试，证实无需额外演示采集即可提升泛化能力
- **适用场景特征**：专为相机随动且占据画面主体的自中心平台设计，解决传统方法在人形机器人上的兼容性问题
- **性能提升结论**：策略在测试阶段遭遇分布外位姿时鲁棒性显著增强，且在相机绑定执行链的人形构型上收益最大

## 工程与产业观察

该工作揭示了移动操作领域“位姿-观测”耦合导致的泛化瓶颈，提出的纯几何数据增强范式为具身智能提供了低成本Sim-to-Real桥梁。对于采用自中心视觉的人形机器人研发，此方法可在不依赖大规模真实数据或昂贵生成模型前提下，快速提升端到端策略在非结构化环境中的部署可靠性，具有明确工程复用价值。

## 英文摘要 (Abstract)

Mobile manipulators such as humanoid robots are increasingly deployed in dynamic, unstructured environments to perform dexterous manipulation tasks. However, end-to-end manipulation policies trained to imitate demonstration data collected from a single robot pose are brittle: even centimeter-scale deviations in robot pose at deployment can drive ego-centric observations and end-effector trajectories out of the training distribution, leading to sharp drops in performance. We introduce MobileVISTA, a data generation framework that transforms demonstrations captured at canonical poses into diverse, pose-perturbed training data by jointly (1) augmenting egocentric visual observations and (2) retargeting actions to compensate for base pose changes. Unlike prior methods, which assume a camera rigidly mounted off the actuated chain or non-trivial articulated robot geometry largely out of frame, MobileVISTA targets compatibility with egocentric platforms (e.g., humanoids) where the camera is both influenced by and must observe the robot's kinematic chain as it moves. We study MobileVISTA in simulated tasks spanning humanoid and bimanual embodiments, and on a real Galaxea R1 Pro. We find policies trained on MobileVISTA-augmented data demonstrate improved robustness to previously out-of-distribution poses encountered at test time, without additional demonstration collection or a trained generative model. Additionally, we find MobileVISTA's benefit is largest on tested humanoids, where the camera rides the actuated chain and the robot fills much of the frame. Additional videos and appendix can be found on our website: https://mobilevista.github.io

作者：Suzannah Wistreich, Stephen Tian, Isabella Huang, Vitor Campagnolo Guizilini, Sergey Zakharov, Katherine Liu, Jiajun Wu
:::

## 原始参考来源 (Sources)

- [arXiv 2610.07511：MobileVISTA: Generative Data Augmentation for Pose Generalization in Mobile Manipulation](https://arxiv.org/abs/2610.07511)