AI-Driven Resource Management for Heterogeneous Industrial IoT: Challenges and Opportunities

Abstract

The heterogeneous Industrial Internet of Things (IIoT) is emerging as a cornerstone of intelligent manufacturing, integrating diverse radio access technologies to meet the stringent and varied demands of industrial applications. However, the substantial complexity of resource management also is introduced by the heterogeneity, where conventional optimization-based methods struggle to cope with high-dimensional, dynamic, and partially observable environments. Recent advances in artificial intelligence (AI) have opened new avenues for achieving autonomous, adaptive, and data-driven resource management in IIoT networks. Therefore, this paper presents a comprehensive study on AI-driven resource management for heterogeneous IIoT systems. Particularly, typical industrial applications and their distinct performance requirements are first analyzed to highlight the need for heterogeneous network integration. Then, a unified AI-driven framework is presented to integrate the sensing, prediction, and optimization through synergistic AI paradigms such as deep learning, deep reinforcement learning, federated learning, large language models, as well as generative AI. Building upon this framework, three key enabling techniques including AI-driven random access, AI-driven resource scheduling and AI-driven heterogeneous collaboration are systematically investigated. Finally, several open challenges and potential research directions are discussed to inspire future innovations toward scalable, interpretable, and resource-efficient AI-native industrial networks.