Date of Award
Level of Access
Master of Science (MS)
Second Committee Member
Third Committee Member
Wireless Sensor Networks (WSN) are becoming increasingly ubiquitous and have broad applications in scientific data gathering, environmental monitoring, surveillance, and localization. This thesis focuses on two critical issues in WSN: methods of localizing in noisy environments and energy modeling. Localization is a fundamental system-level function that can provide a crucial context for measurements taken in an environment. Since sensor nodes are battery-powered, another critical issue in WSN is to understand the energy consumption characteristics of individual sensor nodes. This can provide insight into minimizing power consumption. This thesis demonstrates that neural networks are a viable and efficient option for localization in noisy environments. This thesis qualitatively compares the performance of four different categories of neural networks: Radial Basis Function (RBF), Multi-Layer Perceptron (MLP), Recurrent Neural Networks (RNN), and Reduced Radial Basis Function (RRBF). The performance of these four networks is compared against two variants of the conventional Kalman Filter that is widely used in WSN. The resource requirements in terms of computational and memory resources for all of these methods are compared as well. Experiments indicate that these neural networks strike different tradeoffs between localization accuracy and resource requirements. The RBF neural network has the best accuracy in localizing, however it also has the greatest computational and memory resource requirements. The MLP neural network, on the other hand, has the least computational and memory resource requirements although it is less accurate than the RBF neural network in localization accuracy. In addition, due to the mobile nature of these devices and their reliance on batteries, minimizing energy consumption is a pressing concern in WSN. This thesis develops a detailed statistical model based on Petri nets to evaluate the energy consumption of a wireless sensor node. The model factors critical components of a sensor node, including processors with emerging energy-saving features, wireless communication components, and an open or closed workload generator. Experimental results show that this model is more flexible and accurate than Markov models. The model provides a useful simulation platform to study energy-saving strategies in WSN. For example, it can be demonstrated that by carefully selecting a parameter as subtle as the powering-down threshold that determines when a processor of a sensor node enters a low-power state can potentially result in very large power savings.
Shareef, Ali, "Localization and Energy Modeling in Wireless Sensor Networks" (2008). Electronic Theses and Dissertations. 878.
Files over 10MB may be slow to open. For best results, right-click and select "save as..."