Artificial Intelligence for MPLS and Segment Routing Traffic Engineering: A Systematic Review of Machine Learning, Reinforcement Learning, and Deep Reinforcement Learning

The growing complexity and variability of network traffic have increased the demand for adaptive traffic engineering in Multiprotocol Label Switching (MPLS) and Segment Routing networks. Conventional approaches, including constraint-based routing, mathematical programming, and heuristic optimization, provide established mechanisms for path selection and resource allocation but may encounter computational and operational limitations under dynamic network conditions. Artificial Intelligence (AI), particularly Machine Learning (ML), Reinforcement Learning (RL), and Deep Reinforcement Learning (DRL), offers alternative approaches to traffic prediction, adaptive routing, congestion management, and policy optimization. This paper presents a systematic literature review framework for AI-assisted traffic engineering in MPLS and Segment Routing networks, guided by PRISMA(Preferred Reporting Items for Systematic Reviews and Meta-Analyses.) 2020. The review aims to classify intelligent optimization techniques, examine their integration with forwarding and control architectures, and critically evaluate their effectiveness against conventional routing approaches. Key evaluation dimensions include link utilization, throughput, latency, packet loss, computational overhead, routing stability, generalization, and constraint satisfaction. Particular emphasis is placed on experimental methodologies, baseline fairness, reproducibility, and the distinction between simulation-based improvements and practical deployment feasibility. The proposed synthesis seeks to identify methodological limitations, unresolved technical challenges, and future research opportunities for developing scalable, reliable, and operationally feasible AI-assisted traffic engineering solutions.