Efficient data collection in wireless sensor networks: modeling and algorithms
Lorenzo A. Rossi · University of Southern California Digital Library · 2011
This dissertation focuses on data gathering for wireless sensor networks. Data gathering deals with the problem of transmitting measurements of physical phenomena from the sensor nodes to one or more sinks in the most efficient manner. It is usually the main task performed by a sensor network and therefore the main cause of energy depletion for the nodes. The research efforts presented here propose insightful models for the phenomena sampled by sensor networks with the purpose of designing more energy efficient data gathering schemes. ? We first focus on phenomena that can be characterized by a diffusive process. We propose to model the data via discretized diffusion partial differential equations (PDEs). The rationale is that few equation coefficient plus initial and contour conditions may have the potential to completely describe such spatio-temporal phenomena in a compact manner. We propose and study an algorithm for the in-network identification of the diffusion coefficients. Then, we adopt a spatially non stationary correlation model and we study how this impacts correlation based data gathering and, in particular, the problem of optimally placing a sink node in a sensor network region. Finally, we view each round of sensor measurements as a still image and we represent it via intensity histograms. This way, we can adopt image content analysis tools (intensity histograms matching) to analyze the data and determine which rounds of measurements are of interest to the final users. Therefore energy can be saved by transmitting only some rounds of measurements to the base station. We study the above models and the performance of data collection algorithms via analysis and experiments on synthetic and real data.