Midea has built the world's first intelligent agent factory with factory brain scheduling, Agent command, and robot collaborative operations.

Author | Huang Nan

Editor | Yuan Silai

Recently, the Jingzhou factory of Midea Washing Machines has been certified by the world - record certification body WRCA as "the world's first intelligent agent factory with excellent multi - scenario coverage".

As the industry's first intelligent agent factory, based on the "Midea Factory Brain", it consists of 14 intelligent agents that cover 38 core production business scenarios for collaborative operations. It has end - to - end capabilities from perception, decision - making, execution, feedback to continuous optimization. According to the practical results, in multiple core manufacturing scenarios, intelligent agents can complete tasks that used to take hours for traditional manual labor in seconds, with an average efficiency improvement of over 80%. Among them, the scheduling response speed has increased by 90%.

The right - hand screen shows the Midea Factory Brain

Among them, the Midea Factory Brain is the "nerve center" of the intelligent agent factory. It adopts a highly available and scalable distributed multi - agent architecture, realizes autonomous collaboration of intelligent agents through Agent - to - Agent (A2A) communication, and integrates a large - model inference engine in the industrial field to enhance intelligent decision - making capabilities.

To put it simply, if the intelligent agent factory is regarded as a person, then the factory brain is the central hub responsible for task scheduling, the intelligent agents are neural networks for specific tasks, and the intelligent terminals are the limbs responsible for actual task execution.

Take the common quality inspection task in the Midea Jingzhou Washing Machine Factory as an example. The AI glasses connected to the "Midea Factory Brain" can remind workers of error - prone points based on market problems and first - inspection historical data. Based on vision technology, it connects the R & D and quality intelligent agents, automatically obtains the drawings from the R & D system according to the material information, compares them with the photographed physical objects, and returns the inspection results to the quality intelligent agent to form a complete closed - loop. The efficiency of the first inspection has been reduced from 15 minutes to 30 seconds.

At the automatic locking station for the dryer back cover, when the KUKA "iico collaborative robot" works in collaboration with the planning intelligent agent, it can automatically analyze the model features and quickly match multiple product models through the pictures uploaded to the factory brain in real - time. Even in the mixed - flow production scenario of different washing machine models, the factory brain can complete model recognition in real - time, automatically update the screw - locking program, and achieve flexible and efficient flexible operations like humans to improve the accuracy and efficiency of mixed - line production.

KUKA iico collaborative robot

Zhang Xiaoyi, the Chief Digital Officer of Midea Group, said that in the intelligent agent factory, under the command of the factory brain, every production factor such as humans, machines, materials, methods, and the environment is no longer an isolated unit. Various intelligent terminals such as embodied robots, robotic arms, AMRs (Autonomous Mobile Robots), injection molding machines, cameras, sensors, etc., all have the capabilities of perception, understanding, decision - making, and action.

In recent years, the humanoid robot industry has witnessed significant growth and has become one of the core layout directions for future industries. Different from some industrial manufacturing enterprises that are mostly in the concept exploration stage, Midea Group, relying on its own manufacturing factories and production - line resources, has moved its humanoid robots from the laboratory to actual application scenarios. The robot has completed phased debugging in the injection - molding workshop and has established a business closed - loop with the factory business system.

For industrial scenarios, the Humanoid Robot Innovation Center of Midea Group has independently developed the "Meiluo" humanoid robot. As the core execution unit of the factory's embodied intelligent agent, under the unified scheduling of the factory brain, it deeply collaborates with factory intelligent agents such as quality, DMS (Daily Management System), TPM (Total Productive Maintenance), and EHS (Environment, Health, and Safety Management). Combining multi - modal perception technology and embodied intelligent operation technology, "Meiluo" can independently perform high - frequency tasks such as quality first inspection, DMS patrol meetings, TPM inspections, and EHS inspections, achieving real - time response and full - process autonomous decision - making.

The "Meiluo" humanoid robot

When performing the first - inspection submission task, the "Meiluo" humanoid robot will transport the injection - molded rear barrel to the intelligent first - inspection station. The integrated high - precision self - developed 3D camera will independently conduct size and appearance inspections. The quality inspection data will be synchronized to the quality intelligent agent in real - time to generate a closed - loop first - inspection work order. Only when the system confirms that the part is qualified will it dispatch the robot to return and pick up the next part. If the inspection fails, it will further collaborate with the process intelligent agent and the TPM intelligent agent to optimize and adjust the parameters of the injection molding machine.

Currently, the humanoid robot industry is still in the exploration stage of true large - scale application. Xu Yi, the Dean of the AI Research Institute of Midea Group, pointed out that the primary task is to solve the scenario problem. The richness of scenarios can provide a test environment for technological iteration and promote the evolution of the overall system. Specifically in the factory environment, under the unified scheduling of the factory brain, multiple intelligent agents work together to continuously strengthen the operation capabilities of robots and various embodied intelligent terminals and effectively improve work efficiency.

Midea Jingzhou Washing Machine Factory

Zhang Xiaoyi revealed that in addition to the Jingzhou Washing Machine Factory, Midea Group will further expand the number of scenarios covered by intelligent agents and quickly replicate them to other factories of the group globally.