| Recent advances in wireless communication technologies and the explosive growth of the number and variety of mobile devices and multimedia applications motivate the development of next generation wireless networks, i.e., beyond 4G mobile systems. Indeed, the compelling demand for extensive coverage and high capacity has brought about a challenging problem to design adaptive and scalable network architectures. Some broadband wireless technologies, such as WiMAX, millimeter-wave, and free space optics (FSO), have been developed to meet this demand.;Free space optics have emerged as a promising technology for next generation wireless broadband networks [1] [2]. FSO is a wireless telecommunication system that uses free space as transmission medium to transmit optical data at high bit rates. Compared with traditional wireless technologies, such optical wireless links have many advantages, including cost effectiveness, long transmission distance, license-free operation, interference immunity, high bandwidth, and so on.;In this dissertation, we propose to design and optimize broadband wireless networks based on the FSO technology. We first provide a comprehensive survey of prior work. We classify prospective global FSO networks into three categories, and present current state of the art in the field and discuss the challenging issues and open problems. The objective is to achieve a fundamental understanding of the context and importance of our research and proposed solutions.;Next, we investigate the problem of building robust spanning trees for FSO networks. We adopt the notion of the algebraic connectivity from spectral graph theory as a measure of network robustness, and formulate it as a 0-1 integer linear programming (ILP) problem. The formulated problem is NP-hard. We then present a fragment selection and merging (FSM) algorithm, which is executed in a distributed and asynchronous fashion, to obtain a strongly connected spanning tree. FSM can be used not only for FSO network bootstrapping, but also for auto-reconfiguration in response to network dynamics during operation.;We also investigate the design and optimization of a wireless access network, i.e., wireless mesh network, that exploits free space optical (FSO) communications. In order to improve the scalability of wireless mesh networks, we propose a hierarchical architecture: (i) the lower tier consists of mesh routers that are clustered based on traffic demand and delay requirements, and (ii) the cluster heads are equipped with wireless optical transceivers and form the upper tier FSO network. We develop the plane sweeping and clustering (PSC) and greedy edge-appending (GEA) algorithms for the lower and upper tiers respectively. The proposed algorithms are analyzed with regard to complexity and performance bounds, and evaluated via simulations.;Finally, we study the problem of joint optimization of topology design and load balancing in FSO networks. We consider FSO physical characteristics, cost constraint, traffic demand, and QoS requirements in the formulation, along with various objective functions including network-wide average load and delay. We first propose Reformulation-Linearization Technique (RLT)-based branch and bound (BnB) algorithm, which can produce highly competitive solutions with performance guarantees in the form of the bounded optimality gap. We then develop a fast heuristic algorithm to provide feasible solutions with significantly reduced computation time. This algorithm iteratively perturbs the current topology and computes optimal network flows for the new topology, thus progressively improves the configuration and load balancing of the FSO network. Our simulation results show that the fast heuristic algorithm can also achieve an optimality gap close to that of the BnB algorithm proposed in our prior work.;We conclude this dissertation with a discussion of potential directions for future work, including distributed topology control, Wavelength Division Multiplexing (WDM) FSO networks, and topology transformation. |
