| Wireless networking has emerged as an important area of research in the design of modern communications. Applications based on wireless networks span from ubiquitous voice communication to broadband wireless Internet and environment monitoring. As there is no "one size fits all" network, many different network architectures have been proposed in practice. Examples include the cellular architecture, the mesh architecture, and the ad hoc network architecture.;This thesis provides an optimization framework for wireless network design. It puts the application needs as the optimization objective, and takes into account various design challenges across the network layers as optimization constraints. By adopting a "divide-and-conquer" strategy, this framework decomposes a complex optimization problem into several smaller subproblems, thus making efficient and distributed implementation possible.;There are three main results in this thesis. First, we apply the optimization framework in wireless sensor network design by formulating a joint optimization of source coding, routing, and power allocation. Due to partial and correlated observations among the sensors, source coding distortion becomes a nonlinear and nonseparable function of all sensors' rates. We propose a novel source coding model, which allows efficient design of practical quantization schemes. We then present a dual algorithm, together with a technique called the column generation method, that allows an efficient solution for the overall sensor network design.;Second, we address the throughput maximization problem in wireless mesh networks for broadband multimedia applications. In particular, we apply the optimization framework as a crosslayer strategy of joint multicast routing and interference management. We emphasize the network coding technique for multicast routing and a game theoretic method for interference management, for which efficient and distributed solutions are derived and illustrated.;Third, we tackle the inter-cell interference problem in wireless cellular networks. We present efficient numerical solutions from either a (centralized) optimality perspective, or a (distributed) game theoretic perspective, or a combination of both. Our solutions dynamically balance the spectrum among multiple users with a taxation mechanism, and achieve significant improvement in spectral efficiency. |
