自变量机器人WALL-B物流分拣实测超Figure AI 45%
自变量机器人用低成本双臂夹爪方案，靠WALL-B模型在真实分拣场景中超越Figure AI效率，证明AI大脑比堆硬件更关键。
自变量机器人公开直播测试：双机械臂加标准夹爪（非人形、非灵巧手）的配置，在随机包裹、无人工干预的情况下连续分拣1小时，效率达1816件/小时，比Figure AI此前公布的1248件/小时高45%，硬件成本下降70%。其自研WALL-B模型融合视觉、语言、触觉与动作，能实时感知并调整策略，展示了以模型能力替代复杂硬件的落地路径。
正文摘录
- title: Chinese Embodied Intelligence Sets Global Record! Runs Ahead of Figure AI with 30% Cost and 45% Efficiency; Smart Embodied Brain Is Key - sourcecompany: QbitAI - bodymarkdown: Without human fallback, no fixed workflow, and even parcels arriving at random, it worked continuously and stably for a full hour, achieving a sorting efficiency of 1,816 items/hour!? What's more, this real-world result surpassed the 1,248 items/hour average efficiency record previously published by US-based Figure AI, with sorting efficiency improving by over 45%. The hardware setup used in this logistics sorting livestream—a dual robotic arm + gripper configuration—cut hardware costs by up to 70% compared to Figure AI's previous humanoid robot + five-finger dexterous hand solution. It's worth noting that F…