Geek+ Unveils Embodied AI Solutions for Autonomous Warehouse Operations at CeMAT 2026
具身智能的下一个主战场，为什么是「仓储物流」?
At CeMAT 2026 in Shanghai, Geek+ demonstrated its Gravity embodied AI model integrated with Gino 1 humanoid robots and existing AMR fleets to showcase end-to-end autonomous warehouse operations. The system utilizes a mixed Transformer architecture for task planning and action execution, natural language correction capabilities that improved complex task success rates from 17.5% to 87.5%, and a Warehouse Intelligence core enabling dynamic scheduling across its fleet of over 81,000 deployed robots serving more than 1,000 global clients. Geek+ announced its '24/7 Autonomous Smart Warehouse Plan' aiming for fully autonomous, self-evolving warehouse operations without human shifts, leveraging accumulated operational data to handle long-tail scenarios and continuous optimization.
Source: Chinese robotics — Deployment 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 success rate
87.5%
Reported trials
Reported success rate
17.5%
Reported trials
pick action duration
≈2 seconds
View original evidence
从陌生物品泛化，到复杂任务连续执行，再到自然语言纠偏，Gravity 试图建立一套更完整的具身智能能力体系，「用同一模型解答三类效率考卷」： 动作效率：通过异步推理机制，让动作执行与后续计算在时间上重叠，单次拣选动作稳定在约 2 秒，避免计算等待拖慢整体节拍； 适配效率：面对未曾采样的透明袋装、易滚圆柱与软包，模型无需追加专项数据采集与单独微调，直接理解空间几何与抓取规律，大幅压缩新增 SKU 的适配成本； 异常恢复效率：面对商品滑动或翻滚，机器人无需停机，而是自主调整构型或重规划抓取，将单次失败降维为中间状态，极大保障了产线的连续运行。
Open source S15
- paper: 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.
实验数据显示，复杂操作任务的整体成功率从 17.5% 提升至 87.5%。
Open source S14
从陌生物品泛化，到复杂任务连续执行，再到自然语言纠偏，Gravity 试图建立一套更完整的具身智能能力体系，「用同一模型解答三类效率考卷」： 动作效率：通过异步推理机制，让动作执行与后续计算在时间上重叠，单次拣选动作稳定在约 2 秒，避免计算等待拖慢整体节拍； 适配效率：面对未曾采样的透明袋装、易滚圆柱与软包，模型无需追加专项数据采集与单独微调，直接理解空间几何与抓取规律，大幅压缩新增 SKU 的适配成本； 异常恢复效率：面对商品滑动或翻滚，机器人无需停机，而是自主调整构型或重规划抓取，将单次失败降维为中间状态，极大保障了产线的连续运行。
Open source S15
从单机操作到与仓储机器人协同作业，极智嘉的野心显然不只是制造一台高效具身机器人，更重要的是，如何让具身智能融入整座仓库，与其他机器人协同工作，释放系统级整仓生产力。
Open source S8
例如，当一个订单需要从货架中取出商品，再经过搬运、拣选和归并，系统需要决定由哪台机器人执行任务、何时调用工作站、如何规划路径，以及不同设备之间如何衔接。
Open source S23
极智嘉通过 RMS（机器人管理系统）、WES（仓储执行系统）和多机器人协同调度算法，完成任务分配、路径规划、交通管理与作业流程编排，让不同机器人能够在同一套系统中协同工作。
Open source S24
这也是极智嘉与许多主要聚焦单机能力展示的具身企业之间的差异：将具身机器人融入已有仓储机器人体系，通过多形态机器人和统一软件系统协同，形成端到端整仓生产力。
Open source S19
但这不是简单地把更多机器人摆进同一个仓库，端到端机器人解决方案不断拓展仓库的自主作业范围，AI 驱动的仓库大脑则让整个机器人系统实现更高效的协同、智能决策与运营优化。
Open source S22
Source：Chinese robotics — Deployment discovery · msn.cn