Mitsubishi Electric Develops Smart-control AI Technology that Adapts Rapidly and Nimbly to Changing Conditions

Enables industrial robots to optimize motions in real time, even as objects move

PDF Version (PDF:275.7KB)

TOKYO, February 8, 2018 - Mitsubishi Electric Corporation (TOKYO: 6503) announced today that it has developed a smart-control artificial-intelligence (AI) technology under its Maisart* AI brand that enables devices such as industrial robots to rapidly grasp and nimbly adapt in real time to changing conditions of target objects. The application of this AI technology in devices will simplify automation tasks, even in the case of dramatically changing conditions, such as adapting to the changing shape of a non-rigid object.

* Mitsubishi Electric's AI creates the State-of-the-ART in Technology

Main Features

Uses AI technology and multiple sensors to grasp changes in objects in real time
The technology grasps the state of an object via multiple sensors and then recognizes any change by applying the company's Maisart-brand smart-learning AI. Through repeated estimation based on deep learning, tests have shown that the technology can decrease learning time and grasp changes in conditions in just 3.5ms

Technology redesigns control algorithms autonomously in real time
The automatic generation of optimal control algorithms through deep reinforcement-learning frees designers from having to redesign complex control algorithms. Application of this technology in industrial robots, etc. makes it possible for such devices to adapt to objects that are conventionally difficult to adapt to, such as flexible objects that change shape or objects for which conditions can change dramatically.

Development Objectives

  Learning/Control Method Target
New Grasp conditions immediately using multiple types of sensors and real-time control Adapt to objects such as flexible items that change their shapes or for which conditions can change dramatically
Conventional Learning and controlling based on predetermined conditions Only objects for which conditions change in a predictable way

Note that the releases are accurate at the time of publication but may be subject to change without notice.