KMoPSO TE : Advancing Self-Driving Networks with a Knowledge-Driven Multi-Objective Particle Swarm Optimization Algorithm for Traffic Engineering

As communication networks become increasingly complex and expansive, Internet Service Providers (ISPs) face significant challenges in sustaining network efficiency and responsiveness. Traditional traffic engineering approaches, such as local search and linear programming, often struggle to swiftly adapt to changing network conditions and can be computationally intensive for large-scale networks. Therefore, this paper proposes the Knowledge-driven Multi-objective Particle Swarm Optimization algorithm for Traffic Engineering (KMoPSOTE), an innovative solution that addresses the limitations of current approaches by simultaneously considering multiple network performance objectives. The KMOPSOTEoffers three key benefits. First, it integrates segment routing with equal-cost multi-path (ECMP) into a multi-objective optimization framework for dynamic routing control. Second, it employs an adaptive particle encoding strategy that efficiently determines routing paths aligned with network topologies. Third, a dynamic knowledge-driven update strategy intelligently adjusts link weights based on network attributes, distinguishing it from static optimization methods. These features collectively enhance network performance and adaptability. KMoPSOTEis designed to maintain excellent performance in larger network sizes and has undergone comprehensive evaluations using real-world network topologies from the Internet Topology Zoo dataset. The experimental results indicate that when compared with state-of-the-art algorithms, KMoPSOTEachieves an average reduction of 30.58% in Maximum Link Utilization, thereby showcasing its enhancements in network performance, adaptability, and generalization capabilities within large-scale network environments.