Homeward: No Place Like the Training View
Vision-language-action (VLA) policies implicitly assume that third-person cameras remain fixed at their data-collection poses; once a camera is bumped, remounted, or swapped at deployment, task success degrades precipitously even though the underlying task and robot remain unchanged. We present Homeward, a training-free, plug-and-play observation adapter that canonicalizes observations from an uncalibrated deployment camera back into the training view for frozen VLA policies. Our design stems from a closed-loop diagnostic paradox in simulation: while re-rendering perturbed views into the training camera with oracle geometry recovers much of the lost performance, substituting an off-the-shelf feed-forward 3D foundation model can leave the policy worse off than applying no correction at all. This degradation concentrates on the thin, articulated robot arm, which feed-forward depth prediction tears and scatters - yet the arm's exact 3D state is already known from robot kinematics. Homeward exploits this asymmetry: it estimates what is unknown - camera parameters and static scene depth - from the live RGB frame and a single training-view reference, and renders what is known - the moving robot links - directly from the kinematic model. Evaluated across $\sim$300,000 closed-loop episodes spanning four frozen VLAs, Homeward boosts success across all 1,599 camera variants of LIBERO-Plus from 62.5%, 76.5%, 68.0%, and 3.6% to 77.6%, 95.1%, 91.2%, and 26.4% for OpenVLA-OFT, $π_{0.5}$, $π_0$, and OpenVLA, respectively. Without any scene fine-tuning or deployment camera calibration, Homeward extends to 6-DoF random placements, focal-length changes, and, with per-step pose estimation, continuous intra-episode camera drift, while matching or surpassing a calibrated, scene-trained view-synthesis baseline on three of four policies on LIBERO-Plus.
Authors
- Yinqian Sun (ORCID: https://orcid.org/0000-0003-3879-2624)
- Gao Chen (ORCID: https://orcid.org/0000-0001-9905-5404)
- Mingyang Lyu (ORCID: https://orcid.org/0009-0005-9439-2993)
- Gongce Cao
- Chenfei Yan (ORCID: https://orcid.org/0009-0009-8003-617X)
- Lu Jia
- Qian Zhang
- Bo Wang
- Yuhao Zhang
- Guobin Shen
- Pin Lyu
Institutions
- University of Illinois Urbana-Champaign (US)
- Chinese Academy of Sciences (CN)
- Institute of Automation (CN)
- University of Chinese Academy of Sciences (CN)
- ByteDance (China) (CN)
- Beihang University (CN)
- University of Science and Technology Beijing (CN)
- Tsinghua University (CN)
Publication Details
- Journal
- arXiv (Cornell University)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.48550/arxiv.2611.24455
- Primary Topic
- Robot Manipulation and Learning
- Type
- preprint