Attention-Aided Channel Prediction for Efficient Resource Management in Industrial IoT Subnetworks

Channel State Information (CSI) is critical for optimizing wireless communication systems, particularly in Industrial Internet of Things (IIoT) networks where real-time performance and reliability are paramount. In high-density 6G-enabled factory environments, the mobility of subnetworks and dense deployments exacerbate interference, necessitating proactive and efficient resource management. This paper introduces a dual attention-based channel prediction framework, combined with an AI-driven resource allocation model, to mitigate the challenges of outdated CSI in IIoT subnetworks. By employing spatio-temporal attention mechanisms, the proposed framework effectively predicts CSI and integrates sub-band allocation with power control for optimized resource utilization. Simulation results highlight significant gains in spectral efficiency, enhanced Quality of Service (QoS) adherence, and superior robustness to delay, outperforming state-of-the-art methods. These advancements make the proposed solution a scalable and reliable choice for next-generation IIoT deployments.