An LSTM-based Path Switching Approach for Edge Gateways in Industrial Heterogeneous Networks

Traditional path-switching methods are primarily designed for static or conventional network environments, ren-dering them inadequate for the dynamic nature of heterogeneous industrial networks. This paper aims to tackle the dual challenges of efficient path switching and accurate traffic prediction in such environments. First, an LSTM-based predictive model is developed to forecast critical network performance metrics, such as delay and load, facilitating proactive path adjustments. Second, a dynamic path-switching strategy is introduced, taking advantage of these predictions to minimize path oscillations and maintain stable network performance under varying traffic conditions. Finally, the performance of the proposed method is evaluated using key metrics, including latency, load distribution, and path stability. Simulation and experimental results confirm the effectiveness of the approach, demonstrating significant improvements in latency reduction, load balance, and overall network efficiency within industrial heterogeneous networks.