Mobility prediction-based resource reservation and restorability enhancement in cellular networks

In recent years, we have witnessed an unprecedented growth of research and development in the field of cellular networks. There is currently a vision that future wireless technology will provide mobile users with at least similar services as those available to the fixed hosts. As the dependence on mobile terminals increases, mobile users will likely demand the same reliable service as of today's wireline telecommunications and data networks. This dissertation addresses two connection-level reliability issues in cellular networks: (1) how to limit the forced termination probability of handoff calls without sacrificing wireless resource utilization unnecessarily, and (2) how to improve the restorability of existing radio access networks (RANs) cost-effectively.;To address the first issue, we propose two novel mobility prediction schemes that incorporate real-time mobile positioning information. Unlike previous attempts, which have either assumed circular or hexagonal cell boundaries, our schemes are the first to take the irregular nature of cell boundaries into consideration. We also propose a dynamic resource reservation scheme that utilizes the outputs of our mobility prediction schemes. Both incoming and outgoing handoff predictions at each cell are utilized for computing the reservation target, so as to achieve more efficient handoff prioritization.;To address the second issue, we develop cost-effective design methods to enhance the restorability of existing RAN structures against single element failure scenarios. Mesh-restorable RAN structures are built from existing star and tree based topologies by adding new spans, and allocating spare capacity. Five different variants of span and path restoration schemes are considered; we develop both optimization formulations and heuristic algorithms to generate RAN designs for each restoration scheme. In our design methods, each span can have a different set of candidate transmission capacities that are determined by the actual deployment constraints. The costs associated with the candidate capacities can also be specified individually for each span; this allows true costs to be used, which are likely to be both location-dependent and capacity-dependent. It also allows the design methods to take advantage of the presence of any economy-of-scale effects, which are ignored by many existing spare capacity allocation methods.