An Optimized Machine Learning and Reinforcement Learning based Framework for Edge–Cloud Computing in Industrial Automation

Abstract

Industrial automation systems generate a large volume of real-time data through Industrial Internet of Things (IIoT) devices, requiring efficient prediction and resource management for uninterrupted operations. Traditional cloud-centric models often face challenges such as increased latency, limited scalability, and inefficient resource utilization. To address these issues, this work proposes a Machine learning and Reinforcement learning-based Edge-Cloud computing framework (EC-IA) for optimized prediction and resource allocation in industrial environments. Initially, real-time industrial data are collected from distributed sensors and undergo preprocessing for data cleaning and feature extraction. A Machine Learning model is employed at the edge layer to predict the status of resources and classify them into normal and abnormal conditions. For normal conditions, the processed data are stored in the cloud for historical analysis. In abnormal situations, resource availability at the edge is examined. If sufficient resources are unavailable, tasks are offloaded to the cloud, where Reinforcement Learning-based resource allocation is performed for intelligent scheduling and allocation. The global framework is continuously updated to improve learning efficiency. The proposed framework aims to enhance prediction accuracy, reduce Energy consumption, improve Resource Utilization, and support efficient decision-making in industrial automation systems.