Dynamic in-network filtering of sensor readings for query processing in wireless sensor networks

Abstract

In recent years, rapid advances in sensing and wireless communication technologies have resulted in the availability of wireless sensor networks. Consisting of several small micro-electronic devices which can collect information of surrounding areas and communicate with each other through radio wave, these networks are widely used in a variety of monitoring applications such as habitat monitoring, environmental monitoring, and military surveillance. In these applications, sensor devices are distributed in the environment, periodically sensing and reporting relevant data to the base station to answer queries submitted by the users. While these applications require high accuracy query results and long-term performance of the network, they suffer from resource constraints that do not appear in more traditional wired computing environments. To save energy and prolong the network lifetime, a common approach, which is called in-network filtering, is often used. The basic idea of the approach is to install a filter at each sensor node to suppress newly observed values at the node from being sent to the base station if the values satisfy some certain conditions that guarantee the required precision of the query results. In this thesis, I study in-network filtering in the context of two popular types of queries, namely approximate aggregation queries and exact top-k queries. For the first type of queries, I propose an effective filter setting method that can improve the performance of filters when sensor readings periodically show similar change patterns. Unlike the previous methods, the proposed method dynamically adjusts the filters by predicting the change of sensor readings in the near future based on historical sensor data. In the second type of queries, I propose two improvements on an already well-known approach, called FILA, to make its filter setting strategy more efficient. I customize the centralized query reevaluation process in FILA so that it can work ...