A Theoretical Framework for Adaptive Resource Management in Heterogeneous 5G Networks Under Stochastic Traffic Conditions in Nigerian Urban Environments

Resource management in fifth generation heterogeneous networks must contend with highly dynamic traffic conditions arising from the coexistence of diverse service types with disparate quality of service requirements, operating across multiple radio access technology layers. This doctoral study develops an original theoretical framework for adaptive resource management in heterogeneous 5G networks under stochastic traffic conditions, grounded in the propagation and traffic characteristics of Nigerian urban environments. The framework integrates Lyapunov optimization theory with multi-armed bandit reinforcement learning to derive online resource allocation policies that maximize network utility while maintaining per-user quality of service constraints with probability guarantees. The theoretical framework is parameterized using stochastic traffic arrival models fitted to empirical traffic measurement data from a major Nigerian operator's core network traffic logs covering twelve months of operations in Lagos and Kano. Theoretical convergence proofs are provided for the proposed algorithm under the derived traffic model family. Simulation evaluation against state-of-the-art baseline algorithms under the parameterized traffic scenarios demonstrates that the proposed framework achieves 31% higher network utility and 44% lower QoS violation probability compared to proportional fair scheduling, while maintaining computational complexity within real-time implementation bounds. The study constitutes an original theoretical contribution by deriving the first resource management framework analytically grounded in empirically characterized Nigerian urban mobile traffic processes, providing a novel algorithm with proven convergence properties applicable to 5G network deployment in West African urban contexts. Keywords: resource management, heterogeneous network, 5G, Lyapunov optimization, reinforcement learning.