PhD Dissertations

The rapid growth of wireless technologies such as 5G and beyond demands higher network capacity and reliability. To meet these needs, solutions like millimeter wave (mmWave) bands and Ultra-Dense Networks (UDNs) are being explored. However, these advances also bring challenges, particularly in managing seamless user mobility between cells, known as handovers. Poorly managed handovers can lead to dropped connections or delays. At the same time, ensuring fair and efficient use of limited network resources, especially for users at the edge of coverage areas remains a critical concern. This research presents an approach to address both mobility management and resource allocation. First, it develops a Self-Optimizing algorithm to improve handover decisions by tuning Handover Control Parameters (HCPs) under different network conditions. The study evaluates both fixed and adaptive strategies and further enhances them using Deep Learning (DL) models such as Levenberg-Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG). In parallel, the study introduces a Utility Function (UF)-based scheduling method to ensure fair and efficient data transmission, particularly for users at the cell edges. Simulations in MATLAB validate the proposed methods, showing improved handover performance and better overall network resource utilization. Results indicate that increasing small cell density degrades handover performance. Fixed HCP 3 (HOM = 4 dB, TTT = 640 ms) provide a more balanced performance. In the adaptive algorithm, Adaptive 4, the proposed approach is the best option, offering lower Handover Probability (HOP), acceptable Radio Link Failure (RLF), and improved overall performance. Among DL algorithm, LM approach with Dataset 2 is selected for its highest Reference Signal Received Power (RSRP), Signal-to- Interference-plus-Noise Ratio (SINR), and Handover Success Rate (HOSR), despite trade-offs in HOPP, HIT, and HOT. Comparing the fixed, adaptive, and DL algorithms, the adaptive algorithm provides the best balance of signal quality with high RSRP and SINR, along with low HOPP and RLF. However, this comes at the cost of slightly higher HIT, HOT, and HOP than the fixed algorithm, indicating a minor trade-off in handover efficiency. Additionally, a new UF based scheduler enhances cell-edge and overall throughput by 1.13–1.87x and 1.09–1.30x, respectively, while maintaining fairness across the network. Future research may explore integration of deep learning-based handover optimization with real-time resource scheduling algorithms, validate performance using industrystandard datasets or testbeds, and assess scalability across various user densities, mobility speeds, and frequency bands.