Xiaomi's CyberOne 02 Humanoid Robot Completes 8-Hour Shift on Assembly Line
小米机器人首过“具身斩杀线”!
Xiaomi founder Lei Jun announced that the CyberOne 02 humanoid robot has completed stable 8-hour operations in a vehicle manufacturing facility, performing tasks such as sorting center console side covers. Powered by the Xiaomi-Robotics-1 (XR-1) model, which utilizes an end-to-end VLA architecture trained on over 100,000 hours of human data collected via the UMI device, the robot demonstrated adaptability to flexible parts and production rhythms. While this marks a preliminary validation of line-level capabilities, the source notes that full commercial maturity remains unproven due to gaps in real-world deployment data and reliance on open-source datasets for cross-body adaptation.
Source: Robotics — Chinese web discovery · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported duration
8 hours
View original evidence
集微网·集微APP,各大主流应用商店均可下载 小米创始人雷军最新发布具身智能关键进展:人形机器人 CyberOne 02 已在小米汽车工厂完成连续8小时稳定作业,胜任总装物流区中控台侧盖板排序任务,这被业内视为小米首次越过具身智能“产线斩杀线”。
Open source S2Reported duration
30 minutes
Reported success rate
75%
Reported trials
View original evidence
论文显示,单项任务平均不到10小时真机演示,四项复杂任务平均成功率达75%,对比基线 π0.5 的40%近乎翻倍;演示扩到40小时内,成功率可提升至85%。
Open source S11Reported duration
<10 hours
View original evidence
论文显示,单项任务平均不到10小时真机演示,四项复杂任务平均成功率达75%,对比基线 π0.5 的40%近乎翻倍;演示扩到40小时内,成功率可提升至85%。
Open source S11Reported success rate
85%
Reported trials
View original evidence
论文显示,单项任务平均不到10小时真机演示,四项复杂任务平均成功率达75%,对比基线 π0.5 的40%近乎翻倍;演示扩到40小时内,成功率可提升至85%。
Open source S11Reported duration
100,000 hours
View original evidence
训练分两阶段,先用手持采集设备 UMI 搜集超10万小时真人操作轨迹做无本体预训练,再以超1万小时跨本体资料(含小米自采7200小时)做后训练,把动作统一成“相对末端位姿 delta EE pose”,让同一套模型能适配不同机体。
Open source S9Reported duration
7,200 hours
View original evidence
训练分两阶段,先用手持采集设备 UMI 搜集超10万小时真人操作轨迹做无本体预训练,再以超1万小时跨本体资料(含小米自采7200小时)做后训练,把动作统一成“相对末端位姿 delta EE pose”,让同一套模型能适配不同机体。
Open source S9
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
集微网·集微APP,各大主流应用商店均可下载 小米创始人雷军最新发布具身智能关键进展:人形机器人 CyberOne 02 已在小米汽车工厂完成连续8小时稳定作业,胜任总装物流区中控台侧盖板排序任务,这被业内视为小米首次越过具身智能“产线斩杀线”。
Open source S2
除工厂外,小米青年公寓也跑通连续超30分钟、覆盖3个房间、50多项任务的长程操作,包含整理、清洁、分类、巡检与离场全流程。
Open source S6
论文显示,单项任务平均不到10小时真机演示,四项复杂任务平均成功率达75%,对比基线 π0.5 的40%近乎翻倍;演示扩到40小时内,成功率可提升至85%。
Open source S11
训练分两阶段,先用手持采集设备 UMI 搜集超10万小时真人操作轨迹做无本体预训练,再以超1万小时跨本体资料(含小米自采7200小时)做后训练,把动作统一成“相对末端位姿 delta EE pose”,让同一套模型能适配不同机体。
Open source S9
Source:Robotics — Chinese web discovery · msn.cn