外观
论文arXiv 2611.03779今天
YUBI软指UMI采集99.7小时数据,多任务策略微调成功率升至57%
本文由 AI 根据公开内容整理,可能有误,请以原文为准:Robot-Matched Soft Fingers for UMI Data Collection: Lessons from 100 Hours of Pick and Place(arXiv 2611.03779)
核心内容
针对现有通用操作接口(UMI)依赖基站、刚性手指无触觉反馈及平行夹持力受限等痛点,研究团队开发了基于YUBI的电池供电手持采集设备。该设备在99.7小时内完成了53类物体的抓取演示数据采集,并基于其中49类物体训练了π0.5基础策略。实验表明,通过多任务数据预训练后再对剩余4类未见物体进行微调,操作成功率由35%显著提升至57%,验证了高质量UMI数据对具身智能泛化能力的关键作用。
硬件层面,新型UMI采用软质倾斜手指设计,能包裹物体并通过形变提供接触力视觉代理信号,且几何形态与目标机器人夹爪精确匹配以消除Sim-to-Real或Teleoperation误差。数据消融实验显示,即便仅约50%的数据包含顶部视角,增加顶置相机仍能使策略成功率提高13个百分点。目前该软指设计方案、完整数据集及数据处理代码已通过lute-corp.github.io/soft-yubi-site/开源,为低成本高质量具身数据采集提供了标准化参考。
关键看点
- 软指力反馈:采用软质倾斜手指设计,通过负载下的可见形变作为接触力视觉代理,解决刚性手指无力觉信号问题
- 数据规模与泛化:构建99.7小时53物体数据集,π0.5策略经多任务预训练后,未见物体微调成功率从35%升至57%
- 传感器配置收益:消融实验证实顶部相机关键性,即使仅半数数据含该视角,仍能带来13个百分点的成功率提升
- 开源资产:完整开源软指机械结构、99.7小时原始数据集及处理代码,降低UMI数据采集与复现门槛
工程与产业观察
该工作证明了UMI硬件形态与目标机器人本体的一致性对数据有效性至关重要,软指形变作为低成本力觉替代方案具有工程实用价值。同时揭示了多任务预训练结合少量新数据微调是提升具身策略泛化的高效路径,为行业构建标准化、可迁移的操作数据集提供了软硬件协同范式。
英文摘要 (Abstract)
Training robot policies for multi-task pick and place requires demonstrations on many objects, which handheld Universal Manipulation Interfaces (UMIs) can collect quickly and cheaply. However, existing UMIs depend on external base stations, use rigid fingers that give no signal about contact forces, or use parallel-jaw fingers that limit grip strength. We present a battery-powered handheld interface, built on YUBI, whose soft, angled fingers wrap around objects, deform visibly under load as a visual proxy for force, and match the grippers of the robot. With it, we collect a 99.7-hour dataset over 53 objects, train $\pi_{0.5}$ policies on 49 of them, and hold out the other 4 as unseen objects. We find that a top camera raises success by 13 points even when only about half of the data includes it, and that mid-training on our multi-task data raises success after fine-tuning on unseen objects from 35% to 57%. Finger design, dataset, and data processing code: https://lute-corp.github.io/soft-yubi-site/
作者:Kobi Kelemen, Tomasz Niedzia{\l}kowski, Marcin Paszczak, Victoriano Montesinos
来源
本页事实对应的公开页面,以原文为准。