Some studies of routing and signal processing in sensor networks

This dissertation focuses on selected topics in signal processing and communications for wireless networked system. Specifically, we investigate three problems of interest: the impacts of route discovery on the capacity of wireless networks, energy efficient detection and estimation in wireless sensor networks, and the detection performance of a network of radar sensors in non-Gaussian noise-plus-clutter.;The lack of infrastructure inherent to wireless ad hoc networks leads to the problem of distributed route discovery and maintenance. We introduce an analytical model and perform a quantitative analysis of the route discovery process (RDP) in wireless ad hoc networks. Bounds on RDP performance in terms of pertinent system parameters are determined. We apply our analytical RDP model to specific system models and compare analytical results with those obtained by numerical simulations. Our results give insight into the sustainable level of RDP in an ad hoc network.;Throughput capacity of large ad hoc networks has been shown to scale adversely with the size of network n. However the need for the nodes to find or repair routes has not been analyzed in this context. We explicitly take route discovery into account and obtain the scaling law for the throughput capacity under general assumptions on the network environment, node behavior, and the quality of route discovery algorithms. We also discuss a number of possible scenarios and show that the need for route discovery may change the scaling for the throughput capacity.;Significant research efforts have attempted to improve the energy efficiency of the information processing in wireless sensor networks (WSNs). We study energy efficient sensor selection for target detection in WSNs under the Neyman-Pearson criterion, and in particular, we acknowledge the unreliability of connections. We propose two sensor selection schemes which attempt to minimize the total energy consumption for the desired detection performance. Optimization problems are formulated for both schemes, and we show that slightly suboptimal solutions can be found by a low complexity greedy approach. Simulation results demonstrate that the proposed schemes achieve a better energy efficiency than a scheme where sensors closest to the location of interest are selected.;We also consider energy efficient estimation of an unknown scalar parameter in Gaussian noise in a sensor network, and discuss a new energy-efficient approach to obtain an approximate of Maximum likelihood (ML) estimate. In our approach, sensor transmissions are ordered according to the magnitude of measurements. Sensors with large magnitude measurements will transmit earlier, and those with small magnitude measurements, smaller than a threshold, will not transmit. Compared to the ML estimate, our approach saves energy by reducing the number of sensor transmissions. We also derive a bound on the approximation error which can be utilized to dynamically determine the threshold such that an appropriate trade-off between the energy savings and the accuracy of approximation can be maintained. Numerical results show that our approach can be very energy efficient with only a negligible error introduced.;Many previous investigations of MIMO radar focused on cases with Gaussian noise-plus-clutter. In particular, the optimum detector for this case, called the Gaussian detector here, has been well-established. The performance of the Gaussian detector in cases with non-Gaussian noise-plus-clutter has received much less attention. In this paper, we evaluate the detection performance of the Gaussian detector under non-Gaussian noise-plus-clutter for a non-coherent MIMO radar system. Two different classes of statistical models of non-Gaussian noise-plus-clutter are employed: one employing the generalized Rayleigh distributed envelope distribution with uniform distributed phase, and the other employing the complex spherically invariant random vector (SIRV) distribution. Simulations are carried out which illustrate the receiver operating characteristics (ROCs) and miss probability versus SNR curves of the Gaussian detector in a non-Gaussian environment. The impacts of the numbers of antennas and the SNR are also investigated. The results show that non-Gaussian noise-plus-clutter has no impact on the diversity gain of a MIMO radar system although it degrades the detection performance in some other ways. We also verify some known results on the optimality of the Gaussian detector for SIRV noise-plus-clutter models, while showing this optimality does not generally hold true for non-SIRV models.