A Novel Machine Learning Architecture for Traffic Grooming and Resource Optimization in 5G Optical Fronthaul

Abstract

The emergence of 5G networks has emerged as innovative solutions for traffic grooming and resource management in optical fronthaul networks. Traditional methods are often incapable of managing the complexity of different traffic patterns, low latencies, and high bandwidth consumption, which leads to suboptimal resource allocation and, consequently, high operating costs. The objective is to develop an innovative machine learning (ML) architecture called Intelligent Multi-Attentive Generative Adversarial Networks (IMAGAN) for maximizing resource utilization and traffic grooming (TG) in 5G optical fronthaul networks. The suggested IMAGAN-based architecture consists of a multi-attentive model for identifying spatiotemporal traffic patterns combined with a generative adversarial model to provide synthetic network scenarios. The findings indicate that the IMAGAN-based architecture enhances the performance of energy management systems in terms of resource utilization ratio, bandwidth utilization ratio, rejection ratio, MAE, and RMSE. The findings of the study offer a strong foundation for further improvements in intelligent 5G network design and management.