From Fixed to Free Cameras: Calibration-Free View-Robust Vision-Language-Action Model

Jul 16, 2026

Upload by Aria

Real-world robot deployment often breaks the training-stage camera assumption: cameras get repositioned, remounted, or handheld, while many existing view-robust Vision-Language-Action (VLA) policies still assume accurate camera extrinsics at deployment. CamVLA introduces a camera-centric action representation and predicts both local camera-frame end-effector actions and a 6-DoF hand-eye matrix. A deterministic geometric transform converts these into robot-base actions, yielding calibration-free, depth-free, single-view performance from one monocular RGB image plus task instructions, improving success across unseen viewpoints in simulation and real-world data.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages21

Reading Time53 min

Introduction

The Viewpoint Trap in VLA Policies

Structural Cause: Hidden Hand-Eye Transformation

Proposed Solution: Camera-Centric Factorization

arXiv :2607 .05396v 1 [ cs .CV] 6 Jul 2026 From Fixed to Free Cameras: Calibration-Free View-Robust Vision-Language-Action Model Wenhao Li1 , Xueying Jiang1 , Quanhao Qian2,3 , Deli Zhao2,3 , Shijian Lu 1✉ , Gongjie Zhang4✉ , Ran Xu2,3✉ 1Nanyang Technological University 2DAMO Academy, Alibaba Group 3HuPan Lab 4Alibaba Group Abstract: Real-world robot deployment rarely maintains the training-stage camera setup, where cameras often experience repositioning or remounting depending on actual scenarios. Existing view-robust Vision-Language-Action (VLA) policies tolerate such camera variations only when the camera extrinsics are explicitly provided, making them fragile and hard to use especially when view robustness is critical. We argue that the policy should not be told where the camera is, but rather figure it out by itself. To this end, we introduce Camera-Centric VLA (CamVLA), a new VLA model that decouples manipulation controls from camera geometry by predicting (i) a camera-centric end-effector action expressed in the local camera frame, and (ii) a 6-DoF hand-eye matrix relating cameras to the robot base. A deterministic geometric transformation composes the two predictions into a robot base-frame action. This disentangles how I should move in pose-independent camera-centric action generation from where I am looking from in camera-perspective geometric grounding. The resulting policy is calibration-free, depth-free, and single-view, requiring only a single monocular RGB image as the visual observation and task instruction at deployment. Evaluations in both simulation and real-world robot data show that CamVLA consistently improves success rates across diverse unseen viewpoints. Project page: [https://alibaba-damo-academy.github.io/CamVLA/](https://alibaba-damo-academy.github.io/CamVLA/) . Keywords: Vision-Language-Action Models, Viewpoint Robustness, CalibrationFree Manipulation 1 Introduction Vision-Language-Action (VLA) models [1, 2, 3, 4] have rapidly progressed toward generalist robot policies, leveraging internet-scale vision-language data and diverse robotic demonstrations to ground broad semantic knowledge into directly executable manipulation. Yet despite their semantic competence, state-of-the-art VLAs exhibit a sharp and unexpected brittleness to camera viewpoint shifts. As illustrated in Figure 1, π0 [5] trained on a single canonical perspective achieves a ∼65.3% success rate under its training view on RLBench [6], yet collapses to a mere 6.3% under a 15◦ camera rotation. This failure persists even when the scene remains fully observable and the semantic goal is unchanged. Although large-scale multi-view training could mitigate this, acquiring such data is prohibitively expensive and hard to scale. In practice, real-world robot deployment rarely matches the controlled camera setup of training time: sensors get bumped during operation, mounted on different platforms, hand held by operators, or affixed to mobile bases whose pose drifts continuously. Consequently, without inherent view robustness, VLAs remain tethered to static laboratory setups, failing to generalize to the dynamic and unconstrained configurations of real-world deployment. This brittleness has a structural origin. Standard VLAs [7, 5, 4] rigidly predict actions in the robot base frame from camera-perspective visual observations. However, this base-frame parameterization misaligns action outputs with camera-frame inputs, requiring the network to implicitly resolve the spatial transformation from the camera to the robot base (hand-eye transformation) . Without explicit geometric constraints, this hand-eye transformation remains a hidden variable, forcing the policy to memorize coordinate mappings by coupling manipulation control with camera geometry, which easily collapses under minor viewpoint shifts. Recent works converge on a single recipe to fix this by telling the policy where the camera is. For instance, OC-VLA [8] bypasses

FAQs

What inputs and capabilities does CamVLA require at deployment?

Deployment uses only a single monocular RGB image together with the task instruction, without needing explicit camera extrinsics calibration or depth information. The resulting policy is calibration-free, depth-free, and single-view.

Jul 16, 2026

Upload by Aria