| Conventional applications of wireless networks feature point-to-point based communications with central control. Recent emerging applications of those networks require more flexible system infrastructure and go beyond the point-to-point prototype to achieve better performance. Implementation of these wireless networks requires guidelines from network information theory, a topic which is still emerging. In this dissertation, we investigate four aspects of communications theory which have impact on these new applications of wireless networks, with the hope of shedding more light on this promising area. The first topic is to study the Multiple Input Multiple Output (MIMO) enhanced Gaussian broadcast channel with the objective of maximizing the instantaneous system sum rate. We exploit multi-user diversity based on Dirty Paper Coding (DPC) to maximize the sum rate by a simple greedy scheduling algorithm. Numerical results show the proposed scheme achieves near optimum sum rate for widely accepted fading channel models. The second topic is to design practical coding schemes for the binary distributed source coding (DSC) problem. We propose a capacity-achieving algebraic binning scheme based on turbo codes for the lossless compression of distributed binary sources.; Simulation results reveal excellent performance, among the best reported so far, which is very close to the theoretic limit. The third topic is to design scalar quantizers for the continuous DSC problem where one source acts as side information at the decoder. We propose an iterative algorithm to find optimum quantizers for this new scenario. The new algorithm generalizes the well-known Lloyd type I algorithm. It will, in general, yield different scalar quantizers than the traditional approaches and these new quantizers provide some advantages. In the fourth topic, we consider the design of a sensor network for detecting an emitter whose exact position is unknown. We seek to minimize the total system power consumption subject to detection performance constrains by carefully choosing the thresholds and positions of the sensors. Toward this goal, we propose an iterative algorithm for the optimization problem. |
