Field -gathering wireless sensor networks: Throughput scaling laws and network lifetime.

Abstract

This thesis attempts to understand certain performance limits of field-gathering wireless sensor networks, and to use such understanding toward the energy efficient design of these networks. A field-gathering wireless sensor network is a network where sensor nodes are deployed over a sensing field, which can be one, two or three-dimensional, with the purpose of taking spatial and temporal measurements of a given set of parameters about the field. Many interesting and challenging research issues subsequently arise from a field-gathering sensor network. This thesis focuses on two aspects, both related to the understanding of fundamental performance limits of such networks. The first aspect concerns the throughput scaling laws of such networks as the size of the network grows. The throughput measures how fast each source can deliver data to the sink node, and scaling reveals how the throughput changes as the size of the network grows. The second aspect concerns the lifetime of such a network under certain energy consumption models on sensing and communication. Network lifetime measures the longevity of a network, and reveals the energy efficiency of the network in terms of its architecture, algorithms, and operations. In the throughput scaling study, we derive upper and lower bounds on the per-node throughput achieved for the many-to-one communication scenario under both a flat and hierarchical network architecture. We show how the introduction of clustering can improve the throughput of the network. These scaling results are next compared with those obtained using a different propagation model that directly impacts the extent to which simultaneous transmissions can be successful. In the network lifetime study, we develop a fluid-flow based computational framework that provides estimates of the maximum network lifetime optimized over all possible routing choices. Different from earlier modeling work that typically requires knowledge of precise locations of nodes in the network, our scheme views a specific network deployment as an instance (sample path) from an underlying <italic>distribution</italic> of sensor node. Finally we present an approach to the joint design of data compression and data dissemination by combining linear programming with distributed data compression.