| Summary: | Wireless sensor networks (WSNs) are becoming increasingly important as they provide an ideal platform for monitoring and controlling physical environments. Starting from small match-box sized nodes to tiny coin-like sensors a WSN promises to be the most ubiquitous information gathering system produced.
Being tiny enables ubiquitous and pervasive sensing. However, this form factor comes at the cost of severe resource constraints for the sensor nodes. The nodes can accommodate only a certain amount of energy in the form of batteries and can store and process only a small amount of data due to its crippled size. Due to this reason, sensor networks cannot host more sophisticated applications and the mean time to failure, due to nodes running out of energy, is also low. These are probably the main reasons why sensor networks have not reached their expected potential. This thesis is an effort to alleviate the energy problem of sensor nodes. We attempt to solve the problem using different data and user centric models which can lead to a multi-fold increase of life for the sensors.
Most of the research till date has aimed at micro-adjustments in the sensor hardware and software in order to improve performance. These techniques, though beneficial, increase complexity, and are sometimes difficult to implement. The thesis demonstrates simple techniques that can significantly improve energy-savings over and above the micro-adjustment techniques. The thesis takes a radical point of view and looks at higher level primitives that can be modified for certain applications.
This thesis provides two energy reduction techniques. The first technique involves aggressive duty-cycling of sensors while maintaining connectivity and coverage followed by reconstruction at the base station. Significant gains can be achieved when the sensed environment has correlation between sensor readings. The second technique involves sampling interval scheduling depending on the utilization of the sampled data based on user queries. While the first method ensures correct reproduction of the sensed environment, while reducing the burden on individual sensors, the second method provides an optimal sampling frequency that regulates energy consumption depending on user demands.
|
|---|
