Using Tiny-ML Models on Edge Devices to Improve Industrial Control Systems in Real Time

Abstract

Tiny Machine Learning (Tiny-ML) lets machine learning models run on edge devices with very few resources. This opens up new possibilities for smart Industrial Control Systems (ICS) that don't need to be connected to the cloud all the time. Traditional Programmable Logic Controllers (PLCs) and Supervisory Control and Data Acquisition (SCADA) systems predominantly depend on static, rule-based logic and face challenges in adapting to dynamic, data-intensive industrial settings that necessitate predictive maintenance, anomaly detection, and real-time process optimization. This paper examines the implementation of highly efficient Tiny-ML models on edge devices, including ARM-based micro-controllers and single-board computers, to attain low-latency inference, bandwidth optimization via local processing, and enhanced energy efficiency and security for industrial Internet of Things (IIoT) sensors. The paper talks about hardware and software limits, how to connect to old PLC/SCADA systems using industrial protocols, and gives an overview of contemporary industrial projects that show how edge AI is useful in modern workplaces.