10月12日，第26届中国国际工业博览会在上海开幕。梅卡曼德机器人连续第八年亮相工博会，全场景展示眼脑手技术升级成果以及跨行业、跨形态、跨任务的通用能力。展示场景覆盖汽车、新能源、3C、半导体、AI服务器、机加工等十余个行业，贯穿工件上下料、精密装配、AI质检等智能制造关键环节。通过对眼脑手全链技术升级，机器人能够更稳定地应对透明物体、海量SKU、高精度装配、动态作业等复杂任务，这也意味着，具身智能正从技术验证走向规模化落地。
On October 12, the 26th China International Industry Fair opened in Shanghai. Mech-Mind Robotics appeared at the fair for the eighth consecutive year, showcasing full-scenario "eye-brain-hand" technology upgrade results and cross-industry, cross-form, cross-task general capabilities. Display scenarios cover over ten industries including automotive, new energy, 3C, semiconductors, AI servers, and machining, running through key intelligent manufacturing links including workpiece loading/unloading, precision assembly, and AI quality inspection. Through full-chain "eye-brain-hand" technology upgrades, robots can more stably handle complex tasks including transparent objects, massive SKUs, high-precision assembly, and dynamic operations—meaning embodied intelligence is moving from technical validation to scaled deployment.
人形机器人自主分拣海量随机物体这个展示单元，依托以Mech-GPT多模态大模型为核心的具身大脑，机器人可理解自然语言指令与料筐电子标签，面对随机摆放的食品、日用品、文具、玩具、工具等数百种物体，实时识别物体类别并自主调整分拣策略，双臂与Mech-Hand灵巧手协同操作，实现物体的精准抓取与分拣。该单元展示了机器人对未知物体与开放类别的泛化操作能力，可应用于商超零售、物流分拣等服务领域。
In the humanoid robot autonomous sorting of massive random objects display unit, relying on an embodied brain centered on the Mech-GPT multimodal large model, robots can understand natural language instructions and basket electronic labels, and facing hundreds of randomly placed objects including food, daily necessities, stationery, toys, and tools, real-time recognize object categories and autonomously adjust sorting strategies. Dual arms and Mech-Hand dexterous hands operate collaboratively to achieve precise object grasping and sorting. This unit demonstrates robots' generalization capability for unknown objects and open categories, applicable to service fields including supermarket retail and logistics sorting.
Mech-GPT多模态大模型人机交互单元里，机器人不仅拥有理解自然语言、图像等多模态信息的能力，还能够识别任意物体、理解物体关系、进行逻辑和常识推理，并根据任务目标自主完成分类、整理、抓取等复杂多样的操作。自然语言直接转化为物理动作，这极大降低了机器人的操作门槛。双人形机器人协同上下料和料筐搬运单元，是现场展示的人形机器人工业场景多机协作应用。两台搭载眼脑手的人形机器人可自主协同完成上料、装配、料筐搬运等全流程任务。梅卡曼德自研超高速动态AI成像技术，可对100至500毫米每秒快速移动的物体精准成像，0.7米处动态精度可达0.1毫米。
In the Mech-GPT multimodal large model human-robot interaction unit, robots not only understand multimodal information including natural language and images but can recognize any object, understand object relationships, perform logical and common-sense reasoning, and autonomously complete complex operations including classification, organizing, and grasping based on task goals. Natural language directly converts into physical actions, greatly lowering robot operation thresholds. The dual humanoid robot collaborative loading/unloading and basket handling unit is the on-site display of humanoid robot industrial multi-machine collaboration. Two humanoid robots equipped with "eye-brain-hand" can autonomously collaborate on full-process tasks including loading, assembly, and basket handling. Mech-Mind's self-developed ultra-high-speed dynamic AI imaging technology can precisely image objects moving at 100 to 500 mm/s, with dynamic precision of 0.1 mm at 0.7 meters.
柔性精密装配部分，FPC软排线高精度装配单元采用全新机器人3D眼睛Mech-Eye UHP-040-5M，XY分辨率达0.017毫米、Z向重复精度达±1微米，可清晰呈现微小装配特征，结合AI识别技术，精准定位软排线末端及连接器。机器人大脑实时计算插接位置与角度，结合自适应运动控制完成柔顺插接，插装成功率超过99.7%。显卡电源线柔性抓取与精密装配单元，基于机器人大脑的任务理解与AI规划能力，机器人可以自主完成从电源线抓取到插头插接的自动化闭环。多规格钣金件高精度上料与装配单元，基于机器人大脑的任务理解和自主规划能力，机器人能够识别多种规格的钣金工件，自主完成抓取和装配任务，并结合3D加2D视觉协同感知与精细化操作能力，实现亚毫米级装配精度。
In flexible precision assembly, the FPC flexible flat cable high-precision assembly unit uses the new robot 3D eye Mech-Eye UHP-040-5M, with XY resolution of 0.017 mm and Z-direction repeatability of ±1 micron, clearly presenting tiny assembly features, combined with AI recognition technology to precisely locate cable ends and connectors. The robot brain calculates insertion position and angle in real time, combined with adaptive motion control for compliant insertion, with insertion success rate exceeding 99.7%. The graphics card power cable flexible grasping and precision assembly unit, based on the robot brain's task understanding and AI planning capabilities, allows robots to autonomously complete an automated closed loop from power cable grasping to plug insertion. The multi-specification sheet metal parts high-precision loading and assembly unit, based on task understanding and autonomous planning capabilities, allows robots to recognize multiple specifications of sheet metal workpieces, autonomously complete grasping and assembly tasks, and combine 3D plus 2D visual collaborative perception with fine operation capabilities to achieve sub-millimeter assembly precision.
智能涂胶与胶路检测部分，眼镜腿五轴点胶与胶路检测单元，针对眼镜腿这类外形弯曲、胶路要贴合曲面的点胶场景，通用眼加脑能够精准采集曲面工件三维轮廓，自动生成点胶轨迹，并结合五轴协同控制，实现曲面工件的多角度连续点胶。汽车钣金件在线涂胶质量检测，通用眼加脑质检系统可在胶枪涂胶过程中实时检测断胶、胶偏、漏涂等各类缺陷，检测速度最高可达1500毫米每秒，支持360度全视角检测。
In intelligent gluing and glue path inspection, the eyeglass temple five-axis dispensing and glue path inspection unit, targeting dispensing scenarios like eyeglass temples with curved shapes requiring glue paths to fit surfaces, the general "eye + brain" precisely collects curved workpiece 3D contours, automatically generates dispensing trajectories, and combines five-axis collaborative control for multi-angle continuous dispensing on curved workpieces. Automotive sheet metal online gluing quality inspection: the general "eye + brain" inspection system can real-time detect various defects including glue breaks, deviations, and missed coating during the glue gun application process, with detection speed up to 1,500 mm/s, supporting 360-degree full-view inspection.
高难度场景稳定作业部分，透明物体泛化抓取单元基于自研透明物体AI成像技术，以及经海量真实数据训练构建的AI大脑，机器人可有效应对高透材质三维成像难题，泛化识别常见透明物体。传送带动态抓取上料单元，基于具身智能眼脑手，机器人可有效应对传送带供料过程中工件连续移动、位置与姿态实时变化、工件抖动等核心难题。黑色五角螺母高速无序上料单元，抓取节拍可达2.4秒每件，清筐率可达99%。高反光齿轮高精度机床上下料单元，机器人能够精准定位和抓取工件，满足亚毫米级的入机精度要求。
In stable operation in high-difficulty scenarios, the transparent object generalized grasping unit, based on self-developed transparent object AI imaging technology and an AI brain trained on massive real data, allows robots to effectively handle high-transparency material 3D imaging challenges and generalized recognition of common transparent objects. The conveyor belt dynamic grasping and loading unit, based on embodied intelligence "eye-brain-hand," allows robots to effectively handle core challenges including continuous workpiece movement, real-time position and posture changes, and workpiece jitter during conveyor feeding. The black pentagon nut high-speed disordered loading unit achieves grasping cycle time of 2.4 seconds per piece, with basket clearing rate up to 99%. The high-reflectivity gear high-precision machine tool loading/unloading unit allows robots to precisely position and grasp workpieces, meeting sub-millimeter machine entry precision requirements.
梅卡曼德从创立之初就坚持全球化战略。目前，梅卡曼德在慕尼黑、底特律、首尔、东京均设立子公司，构建起销售、交付、培训、售后的全流程服务网络，业务覆盖近50个国家和地区，服务超100家《财富》世界500强企业。梅卡曼德的物理AI解决方案入选世界经济论坛AI应用之星项目。此次亮相工博会，梅卡曼德通过全场景智能应用，集中展现了眼脑手跨机器人本体、跨行业、跨场景的应用能力。随着产品通用性、可靠性与易用性的持续提升，具身智能正加速落地千行百业，走向更广阔的全球市场。
Mech-Mind has adhered to a globalization strategy since founding. Currently, Mech-Mind has subsidiaries in Munich, Detroit, Seoul, and Tokyo, building a full-process service network of sales, delivery, training, and after-sales, with business covering nearly 50 countries and regions, serving over 100 Fortune Global 500 enterprises. Mech-Mind's physical AI solution was selected for the World Economic Forum's "AI Application Stars" (MINDS) program. At this Industry Fair, Mech-Mind through full-scenario intelligent applications collectively demonstrated "eye-brain-hand" application capabilities across robot bodies, industries, and scenarios. As product generality, reliability, and ease of use continue improving, embodied intelligence is accelerating deployment across thousands of industries, moving toward broader global markets.
