Abstract

The increased intricacy related to the architecture of smart cities necessitates sophisticated and intelligent adaptation of Quality of Service (QoS) techniques to satisfy the distinct needs of different mobile users and applications. The coexistence of diverse latency-sensitive, bandwidth-intensive, and mission-critical services in smart cities imposes dynamically changing QoS requirements that cannot be met with traditional static or best-effort policies. This paper proposes an Adaptive QoS Policy Framework based on Software-Defined Networking (SDN), Mobile Edge Computing (MEC), and Machine Learning (ML) based prediction models for context-aware specific service QOS delivery in real-time. This framework is centered on the Adaptive QoS Policy Engine (AQPE), which executes dynamic traffic classification alongside predictive resource allocation and policy change employing reinforcement learning. The system was simulated in NS-3 with an OpenFlow SDN controller integrated with MEC modules developed in Python. It was tested against industry standards benchmarks of latency, jitter, overall throughput, and packet delivery ratio (PDR). The results suggest the model achieves, on average, up to 46% reduction in latency, 25% improvement in throughput, and over 10% increase in PDR compared to baseline approaches. The results show the value of employing learning intelligence and edge computing within mobile network architectures intended for smart cities. As noted earlier, the adaptive metropolitan approach provides an extendable, controllable, real-time, and service-aware QoS configuration policy fulfilling the requirements for next-generation urban communication systems.