Traffic-aware network slicing for smart cities: a machine learning framework for GBR and non-GBR traffic classification and resource optimization

Smart cities are central to driving national progress. Integrates a wide range of applications, including vehicle-to-everything (V2X) communication, surveillance, healthcare, entertainment, and so on. Effective implementation of these applications and seamless user experience within a smart city framework hinges on two critical requirements: accommodating diverse smart city use cases within specific QoS requirements and prioritizing essential services over lower, lesser-priority services. Network slicing, a core 5G capability, offers an impressive solution to meet these stringent demands by customizing network resources. To assign priority based traffic through network slicing, the proposed framework follows a two-fold approach to assigning priority based network resources. In the initial step, smart city traffic was classified into Guaranteed Bit Rate (GBR) and non-GBR (NGBR) categories using an ensemble-based Bagged Decision Trees (BDT). Then we refine classification by further classifying GBR and NGBR traffic into three critical service categories: Enhanced Mobile Broadband (eMBB), Massive Machine-Type Communications (mMTC), and Ultra-Reliable and Low Latency Communications (uRLLC) and achieved 98.45% accuracy to ensure priority handling of essential smart city services. Classified traffic was then utilized for resource distribution using a priority based heuristic approach. To assess the efficacy of this classification, we designed and compared a framework for two scenarios: a Best Effort Scenario (BES) and a Network Slicing Scenario (NSS). The proposed NSS showcased an improvement of 63.77% to optimize network resources. This approach demonstrates an effective solution for resource optimization for urban services, particularly for prioritized critical smart city applications.