| Sensor nodes with sensing, computing and wireless communication capabilities can form pervasive networks which provide seamless instrumentation of the physical world. Wireless sensor networks have been envisioned for many applications including surveillance, wildlife tracking, structural health monitoring, health care and etc.;The goal of this dissertation is to design an integrated data gathering architecture in mobile sensor networks adapted to application requirements. The dissertation achieves this goal by solving the following problems.;First, due to device imperfection and deterioration, the calibration function, i.e., the relationship between measurements and true measured values, for each sensor device needs to be determined. Instead of relying on traditional schemes to calibrate each sensor individually, which requires tremendous time and efforts, blind calibration schemes, which rely exclusively on measurements from devices in operation without using gold devices or controlled environment, are developed. The schemes have the advantage of small communication overhead and low computation complexity. To work with a wide range of sensor devices, calibration solutions are devised for a broad class of calibration functions including linear, polynomial nonlinear and piecewise-linear relationships. The performance of the proposed schemes is verified through experiments with off-the-shelf MICA2 sensor nodes from Crossbow Technology.;Second, due to size and cost constraints, sensors are typically equipped with limited battery supply. This dissertation provides applications the mechanism to trade-off sensor lifetime, data latency and data loss due to buffer overflow. Specifically, a parameterized data gathering scheme is designed as the instrument to realize this tradeoff. In addition, to help select design parameters in the scheme, an optimization framework is developed that allows sensors to operate at any feasible point in the above trade-off space depending on application requirements. The effectiveness of the proposed framework is validated through implementation on MICA2 sensor nodes. Moreover, to better serve application purposes, the framework can be customized to specific application scenarios. In particular, the framework is adapted to the context of a surveillance application detecting the presence of a target.;The dissertation concludes by identifying open issues that need further investigation. |
