PADRoute: Priority-Aware Enhanced Routing Algorithm in Dynamic 5G/B5G IIoT Networks

The evolution of 5G and Beyond 5G (B5G) networks has led to a substantial increase in data transmission demands, particularly in Industrial Internet of Things (IIoT) scenarios. However, $5 \mathrm{G} / \mathrm{B} 5 \mathrm{G}-\mathrm{IIoT}$ networks face significant challenges, such as dynamic topology changes and network resource fluctuations, which severely degrade Quality of Service (QoS) guarantees. Existing routing methods often suffer from excessive computational overhead or obtaining suboptimal solution. To address these issues, this paper proposes PADRoute, which is a Priority-Aware enhanced Routing algorithm in Dynamic 5G/B5G-IIoT networks based on Graph Attention Network (GAT) and Deep Reinforcement Learning (DRL). PADRoute algorithm first calculates the bandwidth weighted priority according to service characteristics, thereby providing prioritize routing assurance for bandwidth intensive services. Then, PADRoute employs GAT to enhance network state awareness and utilizes Deep Q-Network (DQN) for intelligent routing optimization. Experimental results show that PADRoute improves network throughput, link load balancing and network robustness by $22.2 \%$, $16.9 \%$ and $28.5 \%$.