Robots from Yaskawa Electric Corporation used in the demo, Source
The Japanese text in the image reads: Humanoid Robot × Collaborative Robot × MEC AI
In December 2025, at the International Robot Exhibition in Japan, the mobile operator SoftBank Corp. and industrial robot manufacturer Yaskawa Electric Corporation unveiled a Physical AI demonstration combining AI-RAN (MEC) and AI robotics: MEC AI managed and controlled a fleet of robots within a building.
Demo scenario: In a new employee training situation, the system detects missing items in the office (detecting that one training smartphone is missing), then performs intelligent decision-making and instruction (directing a robot to locate the missing smartphone in the storage room).
<Figure. The edge-based MEC AI communicates with the Robot AI inside the robot and instructs the robot’s movements>
|
Previously, for a robot to make decisions and move autonomously, a high-performance computer had to be embedded within the robot itself. However, this increased the robot’s cost and significantly raised battery consumption. On the other hand, if all of the robot’s intelligence (AI) was processed in the cloud, communication latency made it difficult to perform instantaneous decision-making. This created a fundamental dilemma.
In this demonstration, the robot’s control intelligence was divided into two parts: MEC AI (the “brain”), which generates and instructs the tasks the robot should perform, and Robot AI (the “cerebellum”), which controls the robot’s joint movements in real time based on those instructions. The MEC AI is placed at a nearby base station rather than on the robot, while the Robot AI is embedded within the robot itself.
As a result, the robot becomes lighter and battery consumption is reduced, while still being able to perform advanced decision-making with the assistance of MEC AI.
This is a concrete example that demonstrates the future being shaped by AI-RAN, robotics, Physical AI, and 5G/Private 5G.
|
What impressed me most is the use of the building management system and inventory data to help the AI identify missing items and assign tasks to different robots. This goes beyond simple robot automation and moves toward intelligent coordination across multiple systems.
A question for the team: How many robots can a single MEC AI platform manage simultaneously, and what are the main scalability challenges when expanding to hundreds or thousands of robots? @level devil