| Summary: | 碩士 === 國立臺灣大學 === 資訊管理研究所 === 88 === In this thesis, research shall be conducted to study relative topics in wireless network planning and management. The objective of this thesis is to provide planners and managers of an integrated wireless network with both decision support and operation support with respect to network planning and management, so that the total network deployment and operation cost can be minimized under these constraints of QoS (quality of service).
The proposed functionality of this thesis contains the following five topics: (i) MTSOs configuration and capacity assignment problem; (ii) the wired backbone circuit-switching network topology design problem; (iii) base station allocation, antenna configuration of each base station and power control problem; (iv) channels assignment problem; (v) high reliability of system problem. These problems have positive and negative effects on one another. Because of the complexity of this problem, we use Lagrangean relaxation and subgradient method as our main methodology.
The conclusions in this thesis are presented in terms of formulation, sectorization and performance. In terms of formulation, we model a mathematical expression to describe the fixed sectorization problem. At the same time, we consider not only non-uniform size cell but also non-uniform traffic demand. In this point, our model is more generic. Because of the complexity of this problem, we use Lagrangean relaxation and subgradient method as our main methodology. When using these mathematical tools, they can provide us some hints to improve our heuristics. In terms of sectorization, we find that sectorization is less useful when one base station needs fewer channels. In other words, the wastage of resource fragmentation is more significant than the gains of sectorization. By increasing the number of channels required by one base station, the advantage of sectorization is more evident. In terms of performance, our Lagrangean relaxation based solution has more significant improvement than other intentional algorithms.
|
|---|
