Japanese semiconductor manufacturer ROHM has announced the 'ML63Q253x-NNNxx/ML63Q255x-NNNxx' AI microcomputer that can learn and infer without a network connection.


This is the industry's first microcomputer capable of performing AI learning and inference on its own, offering an innovative approach that is differentiated from existing solutions.

ROHM has implemented technology that can operate AI without the cloud through its proprietary AI solution, 'Solist-AI™'.

This technology boasts a processing speed approximately 1,000 times faster than existing software-based AI models by using a simple 3-layer neural network algorithm and AI accelerator 'AxlCORE-ODL'.

This allows for real-time detection and analysis of abnormal conditions of equipment, such as signs of failure, and allows for rapid learning according to the field environment, enabling a variety of functions.It can be applied flexibly to the machine.

The AI microcomputer features a low-power design (power consumption of approximately 40 mW) and features an ARM Cortex®-M0+ core, CAN FD, and 3-phase motor control PWM to support efficient operation of industrial equipment, residential facilities, and home appliances.

In addition, ROHM plans to build an ecosystem with simulation tools to enable users to evaluate the performance of AI microcontrollers before introduction, and to provide related model development and support.

With this development, ROHM is presenting a new paradigm for detecting signs of failure and improving maintenance efficiency, and is expected to contribute to the stable operation of devices and cost reduction.