Real-time congestion management in 6G networks via GNN-based detection and queue-aware mitigation

The rapid densification and dynamic nature of 6G networks bring significant challenges in real-time congestion detection and mitigation. To address this, we propose a congestion management framework that combines Graph Neural Network (GNN)-based detection with a queue-aware mitigation strategy optimized through a Genetic Algorithm (GA). The GNN captures spatial and temporal traffic correlations using dynamically generated network graphs, while the GA selects optimal mitigation actions such as rerouting, load balancing, or queue adjustment based on real-time network states. Simulation results in ns-3 show that the proposed approach achieves 95.4 % detection accuracy and approximately 91.3 % mitigation success rate. Compared to a non-mitigated baseline, it reduces average queueing delay by about 40 % while maintaining an average mitigation latency of 185 ms. These results confirm the framework's capability to provide adaptive, low-latency congestion control suitable for next-generation wireless environments.