Collaborative Cloud-Edge Channel Prediction for Dynamic Industrial Wireless Networks: An Obstacle-Aware Channel Transformative Architecture

In the era of Industry 4.0, predicting channel states in Industrial Wireless Networks (IWNs) poses significant chal-lenges due to factors like communication delays, device mobility, and obstacle interference. Traditional methods fail to accurately capture the dynamic changes in channel states, particularly in environments with multiple obstacles and rapid device movement, which significantly reduces prediction accuracy. To overcome these challenges, this paper proposes a high-precision channel prediction method based on a cloud-edge-end collaborative architecture. The architecture leverages real-time obstacle detection and multi-dimensional feature extraction to enhance prediction accuracy. At its core, the Obstacle-aware Channel Transformative Architecture (OCTA) model integrates spatial and temporal feature modeling, enabling robust adaptability to dynamic en-vironments. The model is continuously optimized via system feedback and regularly updated to maintain high prediction accuracy and adaptability. Experimental results demonstrate that OCTA significantly improves the accuracy and generalization of channel state prediction in complex industrial environments, outperforming traditional approaches, especially when facing dynamic changes and various types of obstacles.