Mitsubishi Electric’s “Ears for Physical AI” Featured in Nikkei Report Tied to CEATEC

The Nikkei published an article on Oct. 28, 2026, titled “Mitsubishi Electric’s ‘Ears for Physical AI’: Autonomous Robot Technologies Roll Out One After Another, CEATEC,” covering technology for Mitsubishi Electric’s autonomous robots.

The implementation method and effects of the technology covered in the article cannot be assessed from the public information provided. The headline confirms only that technology for Mitsubishi Electric’s autonomous robots was described as “ears for physical AI” and linked to CEATEC. It is unclear whether the technology recognizes speech and converts it into instructions, uses ambient sounds to understand work conditions, or serves another purpose. Therefore, it cannot be treated as a single function spanning voice input through task execution. Our Oct. 13 report, Mitsubishi Electric shows screw-tightening AI at CEATEC, covered a demonstration related to autonomous control of screw-tightening work. In addition, our Oct. 21 report, Mitsubishi Electric to demonstrate processing system linking humanoid and CNC machine tools at JIMTOF2026, described a system that creates work plans from natural-language instructions and assigns tasks to machines and robots. These were separate subjects of reporting, and it cannot be confirmed that the current article covers the same system. However, how inputs and instructions for robot tasks connect to actual process control is a shared issue to consider. Our Oct. 19 report, HoloLab joins Mitsubishi Electric’s MELFA robot partner program, launches “arm-pilot” remote-control software, also described remote control using hand tracking on XR devices and the use of operation data to train physical AI. Different types of input, such as voice, hand movements and machine status, carry different meanings in the data collected and require different approaches to linking them to control. Future evaluations will focus on what is treated as sensor input, under what conditions instruction-recognition accuracy is measured, and how robots are stopped and resumed after a misrecognition. These points cannot be confirmed from the information provided about the article, and there is no basis to draw conclusions about its implementation method or effects. Refer to the original source for details of the article.