Hidden Variables Behind Real-Robot Data Quality: Lessons from Hundreds of Thousands of Hours
As real-robot data collection scales, collection-system design shapes both operator efficiency and the behaviors learned by robot policies. Workspace limitations, control-interface friction, and latency can introduce pauses, detours, and corrective motions into otherwise successful demonstrations. We investigate these effects across embodiment, teleoperation control, and data consistency, combining a kinematic study of 25 dual-arm robots, controlled teleoperation experiments, and real-robot evaluation of fine-tuned $π_{0.5}$ policies. The embodiment study shows that positional reachability overstates orientation-aware task coverage and that modest shoulder motion can substantially expand the effective workspace. Teleoperation experiments show that parallelizing body-part control reduces task completion time by 43-53%, while compliant control improves success and reduces peak contact forces on constrained and contact-rich tasks. Policy-learning experiments show that demonstrations must capture the visual information underlying action choices: active-search demonstrations improve arm-selection accuracy from chance (50%) to 100%. Calibration experiments reveal that mixed-calibration training data remain usable in our tested configurations when an inference-aligned subset is retained, whereas inference-time miscalibration causes all 30 trials to fail. Policies also retain approximately 90% of the extra path length in detour-contaminated demonstrations, showing how inefficient operator motions persist in learned behavior. Together, these findings connect collection-system design to demonstration quality and downstream policy execution. They motivate evaluating effective workspace, control efficiency, observation completeness, calibration alignment, and motion quality before scaling collection, and addressing system-induced compensations at their source.
Authors
- Haomiao Zhang (ORCID: https://orcid.org/0000-0002-5897-2086)
- Guanghui Ren (ORCID: https://orcid.org/0000-0002-9867-8279)
- Cai Xiongfeng
- Naixiang Gao (ORCID: https://orcid.org/0000-0003-3208-5849)
- Pengxin Zha (ORCID: https://orcid.org/0009-0000-5103-8621)
- Xindong He
- Jianheng Song
- Wenjun Sun
- Chiming Liu
- Ruiying Li
- Keru Zhou
- Enlin Gu
- Wei Shan
- Maoqing Yao
- Zhiwen Hou
- Jingmei Jin
- Haojun Ma
- Jiayao Ma
- Minhui Yu
- Jun Zhu
- Yang Yang
- Chang Zhou
- Wenhao Wang
- Xin He
- Jun Zhao
- Peize Li
- Chen Wang
- Liyan Zhang
- Zihao Zhu
- Yang Chen
- World Manipulation Team
- Che Tsang Calvin
- Fan Yang
- Zhaosheng Wang
- Qian Wang
- Xingyu Chen
- Mianrui Zhu
- Xin Zhao
- Bowen Jiang
- Li Liu
- Yi Lu
- Di Yang
Publication Details
- Journal
- arXiv (Cornell University)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.48550/arxiv.2611.22388
- Primary Topic
- Robot Manipulation and Learning
- Type
- preprint