AI-Driven Network Slicing Optimization for Low-Latency and Energy-Efficient 5G/6G Services

Abstract

The rapid proliferation of ultra-reliable low-latency communication, massive Internet of Things, extended reality, autonomous systems, and intelligent industrial applications is placing increasingly stringent demands on 5G and emerging 6G networks. Network slicing provides a flexible mechanism for creating isolated logical networks with application-specific quality-of-service requirements; however, conventional slicing strategies often struggle to simultaneously minimize latency, energy consumption, resource wastage, and service-level violations under highly dynamic traffic conditions. This paper presents an AI-driven network slicing optimization framework that integrates machine learning-based traffic prediction, intelligent slice admission, dynamic resource allocation, and multi-objective optimization. The proposed approach continuously analyzes traffic demand, latency requirements, computational load, and energy conditions to determine appropriate slice configurations and resource assignments. AI-based decision mechanisms are employed to predict congestion and proactively reconfigure slices, while energy-aware optimization reduces unnecessary resource activation without compromising latency and reliability requirements. The framework is designed for heterogeneous 5G/6G environments incorporating cloud, edge, and radio resources. Performance is evaluated using latency, energy consumption, resource utilization, throughput, slice acceptance rate, and SLA violation rate as principal metrics. The proposed architecture establishes an intelligent and adaptive foundation for autonomous network slicing, supporting efficient operation of latency-sensitive and energy-constrained future wireless services.