Less is more

Abstract

Wireless sensor networks are monitoring systems consisting of many small, low-cost and low-power devices called sensor nodes. A large number of sensor nodes are deployed in an environment to monitor a physical phenomenon, execute light processes on collected data, and send either raw data or processed information to the base station. Energy consumption is the main challenge of data collection in a wireless sensor network. Several energy efficient strategies are developed to ensure the longevity of a network. Data reduction is one of the most significant energy management strategies in sensor networks. It concentrates on reducing the volume of the data collected, processed or communicated within the network. Proposed data reduction based energy management techniques assume the radio communication is the most significant energy consumption parameter. However, there are applications in which the computational and sampling energy costs are comparable to or even higher than the communication cost. Therefore, besides the communication level, there is a need to reduce energy consumption costs on sensing and computation levels as well. The main focus of this thesis is to study quality aware data reduction techniques that improve data accuracy and energy efficiency in sensing and computation. Reducing the amount of data in these levels consequently reduces data transmission costs. Data reduction in sensing level is addressed by the adaptive sampling techniques which minimizes the number of sensing operations while data quality metrics are met. In the computation and communication levels, we use compressive sensing techniques to simplify data encoding in the sensor node level, efficiently compress data and reconstruct accurate data in the base station. These two objectives in turn reduces data transmission costs as well.