Training machine learning (ML) models directly in the embedded environment is now possible following MicroAI’s collaboration with Renesas to integrate the former’s MicroAI AtomML technology in the latter’s RA microcontroller range.
It is claimed to result in the first time microcontrollers can train ML models directly in the embedded environments.
The MicroAI-powered microcontrollers can be used in industrial, commercial and consumer systems and devices to adopt edge AI (artificial intelligence). Embedding intelligence at the source of the data, lowers operational costs as well as the expenses around connectivity in ‘smart’ machines and IoT devices, says MicroAI.
Mohammed Dogar, senior director of global business development, Renesas, said: “The industry has been asking to bring more insight and intelligence into the performance of their assets closer to the source of the data, and, working with MicroAI, we have a solution.”
MicroAI is a patented ML algorithm that lives directly on a machine or IoT device, providing deep insight into the behaviour, health and performance of the equipment or devices. Typical examples are robotic welding arms used in automotive assembly lines or for greenhouse gas efficiency in agriculture. These can often face unexpected downtime and static maintenance schedules, resulting in unnecessary costs in terms of lost production and service charges because operators can only when a problem occurs. Creating more visibility into the operation of manufacturing lines, allows asset owners and manufacturers can make adjustments to reduce those events to keep equipment and operations running smoothly.
The Renesas RA (Renesas Advanced) microcontrollers are 32-bit devices with Arm Cortex-M33, -M23 and -M4 processor cores. They are PSA-certified for IoT hardware, software and devices.
In response to calls for predictive insights into machinery, equipment and production systems, MicroAI’s chief executive officer, Yasser Khan, says: “Working with Renesas, MicroAI is delivering that capability by utilising our technology to bring machine learning to microcontrollers, providing the ability to train machine learning models directly in the embedded environment.”
