外观
论文arXiv 2610.1150820 天前
WARP-VLA采用MoE架构解决腕部相机扰动，操作成功率升至78.3%
本文由 AI 根据公开内容整理，可能有误，请以原文为准：WARP-VLA: Wrist-Camera Adaptation for View-Robust Policy Execution in Vision-Language-Action Models（arXiv 2610.11508）
核心内容
针对视觉-语言-动作（VLA）模型在跨平台部署时因腕部相机安装误差导致性能骤降的问题，研究团队提出WARP-VLA框架。该工作直指腕部视角随动特性带来的几何线索漂移痛点，通过算法层面的视图鲁棒性设计替代了对硬件安装精度的苛刻要求。实验数据显示，在LIBERO基准的腕部视角扰动测试中，WARP-VLA使pi-0.5基线模型的平均任务成功率由39.2%跃升至78.3%，验证了其在非结构化相机配置下的策略执行稳定性。
技术上，WARP-VLA采用混合专家（MoE）架构，各专家网络独立学习特定视图的特征变换，路由器基于隐式视图信息动态融合特征，使策略推理完全解耦于相机外参输入。这种特征级适配机制不仅在仿真中有效，更在真机实验中证实了Sim-to-Real迁移能力，解决了多机型复用同一VLA策略时的标定瓶颈。团队已开源腕部视点鲁棒性基准数据集及即插即用代码实现，为具身智能跨本体部署提供了标准化评测工具与算法范式。
关键看点
- MoE视图适配：采用混合专家架构学习视图特异性特征变换，路由器基于隐式信息融合，推理无需相机外参输入
- 性能量化提升：在LIBERO基准腕部视角扰动测试中，pi-0.5模型平均成功率从39.2%提升至78.3%
- Sim-to-Real迁移：仿真环境习得的特征级适配能力成功转移至真实机器人，验证了跨部署设置的泛化性
- 开源评测资产：发布腕部视点鲁棒性基准数据集与即插即用代码实现，支持VLA模型跨相机配置复现研究
工程与产业观察
WARP-VLA证明了通过MoE架构解耦视觉特征与相机外参是解决具身智能跨本体部署难题的有效路径。对于算法工程师，这意味着可在不重新标定硬件的前提下复用预训练VLA策略；对于产业侧，该技术显著降低了机器人换型或量产装配公差对算法性能的敏感度，为具身大模型从单一实验室环境走向多样化工业现场提供了关键的工程化适配方案。
英文摘要 (Abstract)
Despite recent advances in Vision-Language-Action models (VLAs) for robotic manipulation, their performance remains sensitive to changes in camera configuration. The problem becomes more evident in cross-setup deployment, as reproducing the exact camera pose used for training is nearly impossible. Unlike fixed external views, wrist views are more challenging because the camera moves with the robot, causing even small mounting variations to alter fine-grained geometric cues. To address this, we propose WARP-VLA, a camera-view robust VLA for diverse wrist camera configurations. WARP-VLA adopts a Mixture-of-Experts (MoE) architecture where individual experts learn view-specific feature transformations, and a router combines them based on implicit view information. This allows the policy to be deployed without requiring camera extrinsic parameters as additional input. Through experiments on the LIBERO benchmark, WARP-VLA improves the average success rate of pi-0.5 from 39.2% to 78.3% under wrist-view perturbations. The real-robot experiments further show that the feature-level adaptation learned in simulation successfully transfers to diverse deployment settings. To facilitate reproducibility and future research, we release our wrist viewpoint robustness benchmark and a plug-and-play implementation.
作者：Junmyeong Lee, Dongmin Shin, Min-Gyu Park, Wooseok Jeon, Inho Chang, Hae-Gon Jeon
来源
本页事实对应的公开页面，以原文为准。